JP5827892B2 - Structure crime prevention system and elevator equipped with the same - Google Patents

Description

  The present invention relates to a security system for a structure for preventing crime by recognizing an abnormal behavior of a person in a structure such as an elevator through a surveillance camera.

  In a system that visually monitors the video of a surveillance camera, if the supervisor constantly monitors the video visually, the burden on the supervisor is heavy. In order to reduce such a burden, the behavior of a person in the captured video is measured by a certain index (for example, the degree of movement of a dynamic element in the video, that is, the movement amount or movement speed in a predetermined time), and the measured value Surveillance systems that determine abnormal behavior by means of this are becoming widespread. Such a monitoring system is typically used in a crime prevention system that detects dangerous abnormal behavior of a person by image recognition to prevent an accident in advance or identifies a person who has performed abnormal behavior. ing.

  Non-Patent Document 1 discloses a technique for detecting an abnormal behavior of a person in a captured image by using a local appearance on the image called high-order local autocorrelation and a feature amount of motion. In order to improve the safety of passengers in the elevator, the monitoring system is used to detect abnormal behaviors in which a malicious passenger damages the equipment in the car or harms other passengers. Is required.

  In the monitoring system, a normal behavior that is difficult to distinguish from an abnormal behavior from image recognition may be mistakenly recognized as an abnormal behavior, and it is a problem to eliminate such false information. For example, the gesture of straightening hair in front of the surveillance camera is actually a trivial action taken in the elevator car, but when viewed from the surveillance camera, the visual movement becomes a large behavior. This is a typical example of an action that leads to false reports. In addition, in an elevator with a window, when the light inserted from the outside of the car changes as the elevator moves up and down or a shadow moves, it may lead to false alarms.

  In order to improve the false alarm of abnormal behavior detection caused by image recognition through the monitoring camera as described above, it is considered effective to use a sensor with a detection principle different from that of image recognition in combination with image recognition (monitoring camera). .

  Patent Document 1 discloses a technique for detecting an abnormal behavior of a passenger using various sensors such as an infrared sensor and a vibration sensor provided in a car. However, when detecting abnormal behavior using these sensors, considering that there are variations in elevator car dimensions and structures such as floors and walls, even if the physical quantities applied to the sensors are the same, Since the output may change depending on specifications and structural characteristics, it is necessary to set an appropriate sensor threshold value according to the car. Patent Document 1 does not disclose a method for setting the threshold value of the sensor.

  Patent Document 2 discloses two methods for setting a threshold value (reference level) for determining an abnormal behavior with respect to a sensor output signal when an abnormal behavior is detected using a vibration sensor and a volume sensor. One of them is a method of setting a threshold value by learning the amount of change in the output signal of the sensor when an abnormal behavior occurs. Since this method sets the threshold value by learning, the accuracy of the threshold value can be increased. However, before learning, the initial values of the vibration level and the volume level that should be threshold values are determined by the environment where the elevator is attached. Although the method of setting the initial value in this case is not disclosed, it must be individually set for each elevator according to the environment, and the burden of setting work is large, so that the work is simplified. It is desirable.

  The other is a method of learning the amount of change in the output signal of the sensor during normal (normal) boarding, and adjusting the threshold value so as not to misreport frequently during normal boarding. In this method, if the threshold value is increased to suppress the false alarm rate, the degree of detection of abnormal behavior, the so-called false alarm rate, tends to increase, and the threshold value is set so as to suppress the false alarm significantly while suppressing the false alarm rate. It was difficult.

JP-A-63-12579 JP 2004-149287 A

Takuya Minamisato, Nobuyuki Otsu, “Detection of abnormal motion from multiple moving images”, computer vision and image media, ”p. 43-50, October 2005

  The main object of the present invention is to provide a security system for detecting abnormal behavior in a structure having various dimensions and structures such as an elevator car using various sensors in combination with a surveillance camera. An object of the present invention is to provide a system that can easily set a threshold for determining abnormal behavior of a sensor according to a feature with a small amount of work.

  Further, on the second hand, on the premise of a crime prevention system that achieves the main purpose as described above, a threshold for detecting abnormal behavior that significantly suppresses false alarms of abnormal behavior while considering suppression of the false alarm rate of the sensor is set. It is to provide a system that can be configured.

In order to achieve the main object described above, the present invention is basically configured as follows.
(1) One is a monitoring camera that is installed in a structure to be monitored and images a person in the structure, and an image that recognizes the person imaged by the monitoring camera and determines an abnormal behavior of the person A determination unit, a sensor installed in the structure for detecting an abnormal behavior of a person in the structure, and a signal obtained by processing the output signal of the sensor are compared with a threshold for determining the abnormal behavior. In a security system for a structure including a sensor determination unit that performs determination,
A sample signal collection unit that captures an output signal of the sensor as a sensor sample signal when a signal source having a predetermined size is added to the structure;
A reference value that is registered in advance and is a reference value that is a reference value that is a peak value of a change amount of the output signal of the sensor when the signal source is added to a representative structure. A threshold value for determining the abnormal behavior based on the reference threshold value set based on the frequency distribution of the degree of abnormality in the representative structure and the peak value of the change amount of the sensor sample signal. It is calculated, before the larger the ratio of peak value of the sensor sample signal for hexane Telč increases to increase the threshold value, so as to reduce the threshold as the ratio decreases, registered in advance and the ratio And a threshold value calculation unit that calculates the threshold value according to the structure by a predetermined calculation formula using a reference threshold value as a parameter.
(2) The other is a security system including a monitoring camera similar to the above, an image determination unit, a sensor for detecting abnormal behavior, and a sensor determination unit.
A structure information input unit for designating at least one of the model of the structure and the specification of the vibration isolating rubber of the structure;
A data storage unit having data in which a plurality of types or specifications are registered with respect to at least one of the type and specification, and a plurality of threshold values for detecting abnormal behavior are associated with them, and
Have a, a threshold setting unit for setting by searching the threshold from the data according to at least one of the type and specification specified by the structure information input unit,
The threshold for detecting the abnormal behavior is
Reference value for comparison registered in advance, the reference value that is a peak value of the amount of change in the output signal of the sensor when a signal source of a predetermined size is added to the representative structure, Reference threshold values that have been registered, the reference threshold value set based on the frequency distribution of the degree of abnormality in the representative structure, and the output signal of the sensor when the signal source is added to the structure It is a threshold value calculated based on the peak value of the change amount, and is a threshold value associated with each of the plurality of types of the structures .
(3) In order to achieve the secondary object of the present invention, the following configuration is also proposed.

The at least one crime prevention system of (1) to (2) is based on a sensor output signal based on the amount of change in the output signal of the sensor obtained in advance under specific conditions relating to a typical structure that represents the structure in advance. It is registered as a change amount, and the abnormality frequency distribution processed by the sensor determination unit from the sensor output signal change amount related to the representative structure at normal time is registered as a normal abnormality frequency distribution.
The sample signal collecting unit collects, as a normal sensor sample signal, the same condition as the reference sensor output signal with respect to the sensor output signal when the structure is normal,
The threshold calculation unit includes a ratio of a peak value of the sensor output signal change amount of the normal sensor sample signal to a peak value of the reference sensor output signal change amount, a normal abnormality degree frequency distribution of the reference, and the sensor From the threshold value for determining the abnormal behavior of the sensor, a normal frequency error frequency distribution of the normal sensor sample signal and an error rate of the abnormal behavior determination by the sensor determination unit are also calculated, and the error rate is displayed on the display. It is comprised so that it may display via.

  According to the present invention, it is possible to easily set a threshold for determining abnormal behavior of a sensor according to the characteristics of each structure with a small work amount.

  Also, secondarily, the threshold value for detecting an abnormal behavior that suppresses the false alarm rate and significantly suppresses the misreporting of the abnormal behavior with reference to the false alarm rate displayed on the display while exhibiting the main effects described above. Can do.

BRIEF DESCRIPTION OF THE DRAWINGS Schematic which applied the crime prevention system in Example 1 of this invention to the elevator car. 1 is a block diagram illustrating a configuration of a crime prevention system according to a first embodiment. FIG. 6 is a flowchart showing the operation of the sensor determination unit 4 in the first embodiment. FIG. 4 is a graph showing an example of an output signal of a sensor 53 in the first embodiment. The graph which shows the example of the abnormality degree of the elevator car which the sensor determination part 4 in Example 1 determines. The figure which shows the example of the sample signal source which gives the output signal of the sensor 53 to the sample signal collection part 5. FIG. The figure which shows the example of the output signal of the sensor 53 when a sample signal source is given. FIG. 6 is a flowchart of the threshold value calculation unit 6 according to the first embodiment. The graph which shows an example of the abnormality frequency distribution of the abnormal behavior in the representative cage used when setting the threshold for abnormality detection in Example 1. FIG. The graph which shows an example of the abnormal frequency distribution of the abnormal behavior in the representative car and the car 50. The block diagram which shows the structure of the crime prevention system in Example 2 of this invention. The flowchart which shows the method of calculation of the false alarm rate of the threshold value calculation part 16 in Example 2. FIG. The graph figure of the frequency distribution of the scale (abnormality) which shows the abnormality calculated | required from the sensor output signal variation | change_quantity at the time of the normal time created beforehand, when calculating the false alarm rate of the threshold value calculation part 16 in Example 2. FIG. The block diagram which shows the structure of the crime prevention system in Example 3 of this invention. An example of a screen of the user interface 200. The figure which shows the example of the threshold value of abnormality determination of the sensor determination part 4 according to the model of the cage | basket | car 50. FIG.

  Hereinafter, specific embodiments of the present invention will be described with reference to the following examples and drawings.

  Example 1 of the present invention illustrates an elevator as a structure to which the present invention is applied. FIG. 1 shows a schematic configuration diagram of the elevator car provided with a security system.

  In FIG. 1, a surveillance camera 51 is installed inside an elevator car (structure) 50. On the outer wall of the car 50, for example, the outer wall of the ceiling, a signal processing device 52 for performing signal processing of the crime prevention system and determining abnormal behavior of the passenger is installed. In addition, a sensor 53 for detecting abnormal behavior used in combination with the monitoring camera 51 is installed in the lower part of the car 50. In this embodiment, a load sensor is illustrated as the sensor 53, but the present invention is not limited to this as will be described later. Reference numeral 56 denotes an elevator lifting rope from which the car 50 is suspended.

  The surveillance camera 51 is provided on the ceiling or inner wall of the car 50 and is installed so that the movement of the passenger in the car 50 can be captured. The monitoring camera 51 may be a near-infrared camera or a distance image camera such as a Time Of Flight method in addition to a visible camera.

  In this embodiment, for example, an eddy current type load sensor is used as the sensor 53. However, the present invention is not limited to this, and various types such as a well-known strain gauge method can be applied.

  When an eddy current type load sensor is used, the sensor 53 is constituted by a sensor coil that generates high-frequency magnetic flux, and this sensor 53 is connected to the lower surface of the support base 55 of the car floor 54 (the eddy current by the high-frequency magnetic flux of the sensor coil). The conductive plate is disposed so as to be opposed to each other through a gap. The distance between the lower surface of the support base 55 and the sensor 53 changes due to the impact or vibration caused by the load of the passenger or cargo on the car floor 54 or the impact or vibration generated when the wall of the car 50 is hit. The change is detected from the change in the eddy current value on the lower surface 55 and the change in impedance of the sensor (sensor coil) 53. The sensor 53 measures a change in load at a predetermined cycle. This predetermined period is sufficiently shorter than the period of the passenger's movement, and when the car floor 54 is vibrated by the passenger's movement, the output signal of the sensor 53 changes. The video output from the monitoring camera 51 and the signal output from the sensor 53 are respectively input to the signal processing device 52.

  The signal processing device 52 is a device that inputs a video signal from the monitoring camera 51 and an output signal from the sensor 53 and performs signal processing necessary to determine whether or not there is an abnormal behavior of a person in the elevator car. Any calculator can be applied. Although FIG. 1 shows the case where the signal processing device 52 is composed of one device, the signal processing device 52 may be composed of two or more devices connected. In addition, the signal processing device 52 including the processing device inside the monitoring camera 51 or the sensor 53 may be used. Alternatively, the information terminal 59 may be connected to the signal processing device 52 through a connection line not shown in FIG. The information terminal 59 refers to a device having a function of accepting an input operation from an operator such as a keyboard, a mouse, and a touch panel, a signal processing function, and a display function such as a display. The information terminal 59 can be realized by a notebook personal computer, a tablet personal computer, a PDA (personal digital assistant), a mobile phone, or the like.

  FIG. 2 is a block diagram illustrating an example of functional configurations of the monitoring camera 51, the sensor 53, and the signal processing device 52 according to the first embodiment of the present invention.

  Each function of the image determination unit 2, the sensor determination unit 4, the sample signal collection unit 5, the threshold value calculation unit 6, the comprehensive determination unit 7, and the reporting unit 8 in FIG. 2 is configured by a computer program and hardware that executes the computer program. Is done.

  The outline of these functions will be described. The image determination unit 2 measures and measures a predetermined index in the video of the camera 51 (for example, a moving amount and a moving speed of a dynamic element in the video for a predetermined time), The measured value is compared with an image threshold value (reference value) to determine whether or not an abnormal behavior of the passenger has occurred in the car 50.

  The sensor determination unit 4 uses a time series of output signals (sensor output signals) of the sensor 53 for detecting abnormal behavior from a time series, for example, an index indicating a degree of change (processing value: a specific example will be described later). ) Is calculated (processing), and if this index exceeds a threshold value π for determining abnormal behavior of the sensor, it is determined that there is abnormal behavior. The threshold value π for the sensor 53 is calculated by the threshold value calculation unit 6. An example of the threshold calculation formula will be described in detail later.

  The overall determination unit 7 combines the results of the abnormality determinations of the image determination unit 2 and the sensor determination unit 4 and finally determines whether or not the behavior of the passenger in the car 50 is abnormal. The reporting unit 8 generates an alarm signal when the general determination unit 7 determines that an abnormality has occurred, and generates an alarm via an alarm device provided in the car 50.

  The sample signal collection unit 5 collects the sensor sample signals used when the threshold calculation unit 6 calculates the threshold value π for determining abnormal behavior for the sensor 53 as follows. That is, a predetermined signal source (for example, mechanical vibration or sound) is added in the car 50, the vibration or impact by the signal source is detected by the sensor 53, and the sensor output signal is collected as a sensor sample signal. As shown in FIG. 6A, a signal source for generating such a sensor output signal includes a weight 60 that applies an impact to the car floor 54 by dropping, and a surface of the car floor 54 as shown in FIG. 6B. It is only necessary to place at least one of the vibrator 61 that vibrates the floor and the movement of the worker. Such a signal source is prepared and used only when setting the threshold value π. The threshold value π is calculated by the threshold value calculation unit 6 of the signal processing device 52 when the elevator car equipped with the sensor 53 is installed in the building, when the sensor is installed in the existing elevator car, or when the car equipped with the sensor is maintained. The

  When the signal source is added to the car 50, the sensor 53 detects it.

  The threshold calculation unit 6 acquires a sensor output signal obtained by using the signal source as a sensor sample signal. When the amount of change in the output signal of the sensor 53 (difference from the output signal before the input to the sensor 53) is defined as the amount of change in the sensor output signal, the reference value ωo for comparison prepared (registered) in advance is used. The ratio α = ω / ωo of the peak value ω of the sensor output signal change amount of the sensor sample signal is calculated, and the threshold value is increased as the ratio α increases, and the threshold value is decreased as the ratio α decreases. And a threshold value π corresponding to the car (structure) to be installed with the sensor is calculated by a predetermined calculation formula using the reference threshold value registered in advance as a parameter. As this calculation formula, in this embodiment, the sensor threshold value π for determining abnormal behavior according to the specification and structure of the car is calculated by multiplying the reference threshold value πo prepared in advance by this ratio α. However, as will be described later, the predetermined calculation formula may be another calculation formula using the ratio α and the reference threshold as parameters.

  Hereinafter, details of each function of the image determination unit 2, the sensor determination unit 4, the sample signal collection unit 5, the threshold value calculation unit 6, and the comprehensive determination unit 7 will be described.

  The image determination unit 2 inputs the video of the monitoring camera 51 at a predetermined cycle, calculates a video motion abnormality index by applying signal processing to the video, and the video motion abnormality index is calculated. If the threshold value for the image for abnormality determination is exceeded, it is determined to be abnormal, and if not, it is determined to be normal. Since the method of Non-Patent Document 1 can be applied to the method of calculating the index of the abnormal motion of the video, the detailed description thereof is omitted. Further, the method for calculating the index of the abnormal motion of the video is not limited to that described in Non-Patent Document 1, and other methods may be used. The image determination unit 2 outputs an abnormality determination result at a predetermined output cycle.

  The sensor determination unit 4 determines the abnormal behavior of the passenger in the car 50 from the sensor output signal by applying the flow of FIG. 3 to the sensor output signal at a predetermined cycle.

  That is, the sensor determination unit 4 first reads the output signal of the sensor 53 at a predetermined cycle (S1). FIG. 4 is a graph showing an example of the output signal of the sensor 53 read in S1. In the graph of FIG. 4, the horizontal axis T represents time, the vertical axis W represents the output signal of the sensor 53 (signal strength of the output signal), and the output signal W is substantially constant until time T0, but after time T0. It has fluctuated greatly. This is transmitted to the car floor 54 when the output signal W of the sensor 53 shown in FIG. 4 is measured, for example, because two passengers are in the car 50 and stay still until T0. The force is constant with the weights of the two passengers, but since the abnormal behavior of the balance occurred after T0, the movement of the passenger accompanying this abnormal behavior becomes vibration, and the force transmitted to the car floor 54 changes with time. Yes. The amount of change in the output signal of the sensor 53 (difference from the output signal before the input to the sensor 53) is referred to as a sensor output signal change amount ΔW. Note that the sensor output signal change amount ΔW is not limited to the maximum and minimum values of the output signal waveform in FIG. 4, and the magnitude of the output signal W at that time and the sensor 53 are also in the middle of the position. It is a value calculated by the difference from the magnitude of the output signal before the input is applied to. Next, the sensor determination unit 4 calculates the strength of the vibration from the change amount (sensor output signal change amount) ΔW of the output signal of the sensor 53 in S1, and measures the strength of the vibration as an abnormality in the car 50. Output as (abnormality) (S2). The graph of FIG. 5 is an example of the degree of abnormality calculated from the output signal W of FIG. 4, and the horizontal axis indicates time T and the vertical axis indicates abnormality degree P. The degree of abnormality P in FIG. 5 is a statistic of the sensor output signal change amount ΔW within the time window τ shown in FIG. 4 at each time. That is, in this embodiment, as an example, the statistical amount (abnormality) P of the sensor output signal change amount ΔW is used as the processed value of the output signal of the sensor 53, but the present invention is not limited to this. Other processing values may be adopted. The degree of abnormality P can be calculated by, for example, the standard deviation of the sensor output signal change amount ΔW within the time window τ. Since the standard deviation of the sensor output signal change amount ΔW is the same as the standard deviation of the output signal W of the sensor 53, it may be calculated there. In FIG. 5, the degree of abnormality P takes a small value close to 0 until T0 when the passenger is still, but takes a value larger than 0 after T0 when the passenger behaves abnormally. Next, the sensor determination unit 4 compares the degree of abnormality P calculated in S2 with the threshold value π, and determines that the abnormality degree P is greater than the threshold value π and is abnormal (S4), and otherwise normal (S5) ( S3). In the example of the graph of FIG. 5, it is determined that the abnormality degree P is normal until T1 where the abnormality degree P is equal to or less than the threshold value π, and the abnormality is determined after T1 where the abnormality degree P is larger than π.

  In the above description, the standard deviation of the sensor output signal change amount ΔW within the time window τ is used as the degree of abnormality P of S2. However, other degree of proportionality to the magnitude of the sensor output signal change amount ΔW within the time window τ is used. An index (processing value) may be used. The intensity of the spectrum of Fourier transform is an example. Alternatively, the time window τ may be divided into several sections, and after calculating the standard deviation and spectrum intensity within the sections, the statistical value such as the median or average may be obtained to obtain the degree of abnormality P.

  As described above, the sample signal collecting unit 5 collects the output signal of the sensor 53 when the car 50 generates vibrations from a predetermined signal source (simulated behavior: weight, vibrator, etc.) on the car floor 54. To do. FIG. 6A shows a case where the weight 60 is dropped from a predetermined height L onto the car floor 54 as an example of the signal source. The mass of the weight 60 is a predetermined value determined in advance. The falling of the weight 60 may be performed by a machine or manually. FIG. 6B shows a case where the vibrator 61 is vibrated by placing it on the car floor 54 as an example of the signal source. The vibrator 61 is a device that repeats vibration by applying force to an object at a predetermined cycle. The falling of the weight 60 and the vibration by the vibrator 61 are transmitted to the sensor 53 through the car floor 54 and the support base 55.

  Examples of output signals of the sensor 53 collected by the sample signal collecting unit 5 are shown in FIGS. 7A and 7B, the horizontal axis indicates time T, and the vertical axis indicates the output signal (signal strength of the output signal) W of the sensor 53. In the waveform 71 when the weight 60 is dropped in FIG. 7A, the peak value ω1 of the sensor output signal change amount appears at time T2. A waveform 72 in FIG. 7B is a waveform when the vibrator 61 is vibrated, and continues regular vibration having a peak value ω2 of the sensor output signal change amount after time T3.

  The threshold calculation unit 6 calculates the threshold value π for abnormal behavior determination used in the sensor determination unit 4 from the output signal of the sensor 53 of the sample signal collection unit 5, that is, the sensor sample signal using the signal source, according to the flow of FIG. To do.

  The threshold calculation unit 6 first calculates the sensor output signal change amount from the sensor sample signal collected by the sample signal collection unit 5, that is, the output signal of the sensor 53 to which the signal source (the weight 60 or the vibrator 61) is applied. The peak value ω is calculated (S10). The peak value ω of the sensor output signal change amount is calculated by an appropriate method according to the type of signal source given to the sensor 53. When the weight 60 is dropped as a signal source, the threshold value calculation unit 6 measures the peak value ω1 of the sensor output signal change amount at the peak at time T2 of the waveform 71 in FIG. The extraction of the peak can be calculated by, for example, the maximum value of the change in the waveform 71 immediately after the weight 60 is dropped. When the vibrator 61 is vibrated as a signal source, the peak value ω2 of the sensor output signal change amount at the regular vibration peak of the waveform 72 is extracted. The peak value ω2 of the sensor output signal change amount can be obtained from the regular vibration of the waveform 72 by extracting the maximum frequency component when Fourier transformed. However, the present invention is not limited to this example.

The threshold value calculation unit 6 calculates the ratio α of the peak value ω of the sensor output signal change amount obtained in S10 with respect to the reference value ωo for comparison registered in advance by Equation 1 (S11). Here, the reference value ωo for comparison gives the signal source to the sensor 53 in the representative car (representative structure) (specifically, the signal source is added in the representative car (representative structure), The peak value of the change amount of the sensor output signal given by detecting the vibration or the impact by the signal source with the sensor 53 provided in the representative car and collecting the sensor output signal as the sensor sample signal in the representative car) It is. Here, the representative car is a car having a predetermined size and structure provided with the sensor 53 like the car 50, and an arbitrary one of cars having different structures and specifications is defined as a reference. It is a basket. The representative car is used to obtain a reference value (standard value) ωo to be compared with the peak value ω of the sensor output signal change amount in advance, and is a standard used when the threshold calculation unit 6 calculates the threshold π. Used to determine the threshold value πo. A method for obtaining the reference threshold value πo will be described later.
[Equation 1]
α = ω / ωo
As described below, the ratio α of the peak value of the sensor output signal change amount using the signal source (the peak value of the sensor output signal change amount of the car 50 and the representative car) is set between the car 50 and the car floor 54 of the representative car. A description will be given of the ratio of the sensor output signal change amount ΔW when the same force F is applied. First, when a force of a magnitude F is applied to the car floor 54 in the car 50, the difference between the sensor output signal change amount ΔW is treated as a spring model of Formula 2.
[Equation 2]
ΔW = κ′F
Κ ′ in Equation 2 is the reciprocal (1 / κ) of the spring constant κ of the car 50. The reason why the spring model of Formula 2 is applied is that a sensor (here, an eddy current type load sensor) 53 is generated when the vibration force generated by the movement of the passenger is transmitted to the car floor 54 and the support base 55 in the car 50. This is because the phenomenon in which the gap between the support 55 and the support base 55 changes has the same structure as the phenomenon in which the weight is placed on the spring and shaken. Similarly, a representative spring model of Formula 3 is set up. In Equation 3, κ′o is the reciprocal (1 / κo) of the spring constant κo of the representative car.
[Equation 3]
ΔW = κ'oF
Here, if the maximum value of the force applied to the sensor 53 by the signal source (weight 60, vibrator 61, etc.) is Fs, the signal source force is the same for both the car 50 and the representative car. Fs. At this time, the sensor output signal change amount has a peak value. Therefore, Formula 4 and Formula 5 are established between the peak value ω of the sensor output signal change amount by the maximum value Fs of the force applied from the signal source in the car 50 and the representative car, and between ωo and Fs.
[Equation 4]
ω = κ'Fs
[Equation 5]
ωo = κ'oFs
By dividing equation 4 from equation 4, it can be seen that the relationship of equation 6 is established between the ratio α of equation 1 and the reciprocals κ ′ and κ′o of the spring constant.
[Equation 6]
α = ω / ωo = κ ′ / κ′o
Here, if the force F is constant in Equation 2, the sensor output signal change amount ΔW is proportional to the reciprocal κ ′ of the spring constant. That is, when the force F applied to the car floor 54 is constant, ΔW increases as the reciprocal κ ′ of the spring constant increases. Therefore, the ratio α is a ratio of the sensor output signal change amount ΔW with respect to the representative car of the car 50 when a force F having a certain magnitude is applied to the car floor 54.

Next, the threshold value calculation unit 6 calculates the threshold value π of the car 50 from the ratio α obtained in S11 and the reference threshold value πo for the sensor determination unit 4 obtained in advance by the representative car (S12). In S12, it is assumed that the degree of abnormality P of the car 50 conforms to Equation 7 in the calculation of the threshold value π. The reference threshold value πo is also based on the premise that the degree of abnormality Po of the representative car follows the equation (8).
[Equation 7]
P = βκ'Fa
[Equation 8]
Po = βκ′oFa
In Equations 7 and 8, β is a proportionality coefficient, and Fa is a representative value of the force exerted on the car floor 54 by the movement of the abnormally behaving passenger. The establishment of Equations 7 and 8 is based on the following two preconditions. The first precondition is that the force Fa follows Formula 2. This precondition is based on the fact that the vibration force caused by the movement of the passenger in the car 50 is transmitted to the sensor 53 through the car floor 54 and the support base 55 in the same manner as the signal source (weight, vibrator). The second precondition is that the abnormalities P and Po are proportional to the sensor output signal change amount ΔW. As described in the description of the sensor determination unit 4, this precondition is satisfied because the abnormalities P and Po are statistical amounts of the sensor output signal change amount ΔW such as standard deviation. By satisfying the above two preconditions, the relationship of the formula 2 is established between the force Fa applied to the car floor 54 when the abnormal behavior occurs and the sensor output signal change amount ΔW, and the abnormalities P and Po are Equations (7) and (8) are satisfied by being proportional to ΔW.

In S12, the peak value ω of the sensor output change amount using the vibration source (the peak value of the sensor output fluctuation amount in the car 50) and the peak value ωo of the sensor output change amount that is the reference value (the sensor output change amount in the representative car). The threshold value π for determining abnormal behavior in the car 50 is calculated by using the ratio α of the peak value) and the reference threshold value πo, and this ratio α represents the degree of abnormality P of the car 50 in Expression 7 as a representative car. This also agrees with the value divided by the degree of abnormality Po in equation (8) (see equation (9)).
[Equation 9]
P / Po = κ ′ / κ′o = α
Here, how to obtain the reference threshold value πo in the representative car will be described with reference to the graph of FIG. The fact that the threshold value π in the car 50 is obtained by multiplying the reference threshold value πo by the ratio α will be described with reference to the graph of FIG.

  FIG. 9 gives various abnormal behaviors that a passenger can take to a representative car, collects the degree of abnormality (statistics) Po obtained from the sensor output signal change amount of the sensor 53 obtained thereby, and the magnitude of the degree of abnormality Po. 3 is a graph showing the frequency distribution 81 of the error (graph of the abnormality frequency distribution). The horizontal axis of this graph is the degree of abnormality Po, and the vertical axis is the frequency H. As described above, the degree of abnormality Po is obtained from the statistic of the sensor output signal change amount in the representative car, and is obtained by the sensor determination unit 4 calculating the degree of abnormality as shown in FIG. In the frequency distribution of the degree of abnormality, as the degree of abnormality decreases, there is a higher possibility that the abnormal behavior will be missed (hereinafter also referred to as “missing information”). In FIG. 9, when a reference threshold value πo for determining abnormal behavior is set for an arbitrary degree of abnormality in the abnormality degree frequency distribution 81, a region 82 that is smaller than the reference threshold value πo in the abnormality degree frequency distribution 81 is a region in which a false alarm occurs. Thus, the area of the region 82 corresponds to the unreported rate when viewed from the entire frequency distribution 81. If the reference threshold value πo is excessively reduced in order to reduce the misreporting rate to zero, the degree of erroneous reporting of normal behavior as an error is increased, so the reference threshold value πo should be set in consideration of both the misreporting rate and the misreporting rate. It is decided. In other words, the reference threshold value πo is set so that the unreported rate is equal to or less than the target value. The false alarm rate is estimated and the reference threshold value πo may be set so that the false alarm rate is equal to or less than the target value. However, as described in Example 2 described later, when πo and thus π are set as the threshold values, It may be displayed by calculation how much the false alarm rate becomes so that the threshold setting person can understand.

The reference threshold value πo obtained from the abnormality frequency distribution 81 in this manner is stored in advance in the database of the signal processing device 52 of the actual car 50 together with the threshold value calculation expression of several tens, regardless of whether it is a representative car. It is registered.
[Equation 10]
π = απo
When the above-described ratio α is obtained, the sensor determination unit 4 in the car 50 calculates the threshold value π in the car 50 by using the threshold value α of Formula 10 using the ratio α and the reference threshold value πo. This can be represented in a graph as shown in FIG.

  The vertical axis and the horizontal axis of the graph of FIG. 10 are the same as those of FIG. In FIG. 10, reference numeral 81 is a frequency distribution graph of the degree of abnormality (statistics of the sensor output signal change amount described in the description of the sensor determination unit 4) in the representative car similar to FIG. 9, and reference numeral 91 denotes the car 50. It is an abnormality degree frequency distribution graph. The abnormality degree frequency distribution 91 is moved by α times from the position of the abnormality degree frequency distribution 81 by multiplying the abnormality degree frequency distribution 81 in the representative car by the ratio α of Equation 9 (obtained by Equation 1). . That is, the car 50 has a different structure and specifications than the representative car, and therefore the spring constants of Formula 2 are different, so that the abnormality frequency distributions 81 and 91 are also shifted from each other. Therefore, the threshold value π in the abnormality degree frequency distribution 91 is also shifted from the reference threshold value πo. By setting the threshold value π, the unreported rate (indicated by reference numeral 92) below the threshold value π becomes substantially equal to the unreported rate of the representative car (indicated by reference numeral 82).

When the ratio α cannot be calculated accurately, a calculation formula with a margin may be used. Formula 11 is a case where the threshold value π is calculated by multiplying the right side of Formula 10 by the margin coefficient η. The margin coefficient η is a number between 0 and 1.
[Equation 11]
π = ηαπo
For example, when the ratio α is assumed to vary up to 20% from the true value, if the right side of Equation 10 is multiplied by the margin coefficient η by 0.83 = (100/120), the ratio α is greater than the true value. Is offset by 20%, and the threshold value π can be prevented from being excessively increased by mistake. If the threshold value π can be prevented from becoming excessively large, the unreported rate (region indicated by reference numeral 92) can be prevented from becoming excessively large. Note that the above-described method of giving a margin to the threshold value π is merely an example, and the threshold value π may be modified to be smaller by another method. For example, when the ratio α is larger than 1, the threshold π is smaller than that calculated by Equation 10 by taking the square root of the ratio α and converting the square to the square when the ratio α is smaller than 1. As described above, the threshold value calculation formula may be changed depending on whether the ratio α is larger or smaller than 1.

  The comprehensive determination unit 7 comprehensively performs abnormality determination based on the abnormality determination results of the image determination unit 2 and the sensor determination unit 4. In the simplest determination method of the comprehensive determination unit 7, the logical product of both of the sensor determination unit 4 of the image determination unit 2 is taken. That is, when both of the sensor determination units 4 of the image determination unit 2 determine that there is an abnormality, it is determined that there is an abnormality, either one or normal. In addition, the logical product may be obtained after processing the sensor determination unit 4 so as to continue the abnormality until a predetermined time after the sensor determination unit 4 determines that the abnormality is once within a certain time range.

  The reporting unit 8 issues a report after the comprehensive determination unit 7 determines that there is an abnormality. A method of calling attention to passengers who are abnormally behaving by a speaker in the car 50 (not shown in FIG. 1), or stopping an elevator on the nearest floor and opening a door so that passengers damaged by abnormal behavior can escape A method of recording the video of the camera 51 in a video recording device (not shown in FIG. 1) and leaving evidence of abnormal behavior, or sending the video to an external monitor via a communication path (not shown in FIG. 1). The transmission method is an example, but is not limited thereto.

  In the first embodiment of the present invention, with the configuration described above, it is not necessary to collect a large number of samples of the output signal of the sensor 53 for the abnormal behavior even in the car 50 having a different size and structure from the representative car. By collecting at least one output signal (sensor sample signal) of the sensor 53 of the signal source such as the falling of the weight 60 collected by the unit 5 or the vibration of the vibrator 61, the representative rate of the false alarm for the abnormal behavior can be represented. The abnormal behavior detection threshold using the camera 51 and the sensor 53 can be realized by setting a threshold value for determining the abnormal behavior for the sensor 53 of the car 50 so as to be substantially equivalent to the above.

  In the description of the threshold value calculation unit 6 according to the first embodiment, since Equation 10 is used for calculating the threshold value π, the calculation formula and the reference threshold value πo are registered in advance in a database or the like. In addition, the background of the threshold calculation formula of Formula 10 uses the concept of the abnormality frequency distribution (81, 91) and the unreported rate (82, 92) as shown in FIG. 9 and FIG. As a way of thinking, the reference threshold value πo may be calculated by replacing with a method other than the abnormality frequency distribution. For example, the reference threshold value πo may be calculated by a calculation method in which a predetermined margin coefficient is multiplied from the representative value of the abnormality degree Po of the abnormal behavior. The representative value of the degree of abnormality Po can be calculated using, for example, an abnormality degree statistic such as an average or median degree of abnormality. In this example, since the representative value of the degree of abnormality changes according to the ratio α between the representative car and the car 50 according to Equation 9, the calculation method of multiplying the average value of the degree of abnormality of the abnormal behavior by a predetermined margin coefficient The threshold value π in the car 50 when the above is applied is α times the threshold value πo of the representative car, and can be calculated using Equation (10).

  FIG. 11 shows a functional configuration of the second embodiment of the present invention. 11 also includes a monitoring camera 51, a sensor 53 for detecting abnormal behavior, an image determination unit 2, a sensor determination unit 4, a comprehensive determination unit 7, and a reporting unit 8, similar to the example of FIG. These functions are the same as those in the first embodiment. Moreover, you may have the display part 17 connected with the threshold value calculation part 16. FIG. The sample signal collection unit 15 and the threshold calculation unit 16 are realized by a signal processing device 52 (see FIG. 1), and the display 17 is realized by an information terminal 59 (see FIG. 1).

  The sample signal collecting unit 15 and the threshold value calculating unit 16 have the following functions in addition to the functions described in the sample signal collecting unit 5 and the threshold value calculating unit 6 of the first embodiment.

  That is, the sample signal collecting unit 15 collects sensor sample signals for threshold calculation (for calculating the ratio α) performed by the sample signal collecting unit 5 of the first embodiment, that is, from a signal source (weight 60, vibrator 61, etc.). In addition to collecting the output signal of the sensor 53 when receiving the force of the sensor, the normal condition different from the signal source for the threshold calculation (weight 60, vibrator 61, etc.) The output signal of the sensor 53 obtained in (operation) is collected as a normal sensor sample signal.

  Similar to the threshold calculation unit 6 of the first embodiment, the threshold calculation unit 16 uses the sensor sample signal of the signal source (weight 60, vibrator 61, etc.) to calculate the threshold π for determining abnormal behavior. The explanation is omitted here. In addition to the calculation of the threshold value π, the threshold value calculation unit 16 calculates a normal false alarm rate by using the normal time sensor sample signal. Alternatively, the threshold value for determining at least one abnormal behavior of the sensor determination unit 4 and the image determination unit 2 is calculated so that the false alarm rate at the normal time becomes as small as possible. Here, typical conditions for normal operation include vibrations caused by the raising and lowering operation of the car 50 when unattended. The sample signal collecting unit 15 collects the output signal of the sensor 53 generated by such normal vibration of the lifting operation as a normal sensor sample signal. The reason why the sensor signal collecting condition is set when the car is lifted or lowered when the vehicle is unattended is as follows. In normal times, passengers can be broadly divided into unmanned times when the car 50 is not on board and manned people with one or more passengers on board. In the unattended state, vibration transmitted to the sensor 53 when the car 50 is shaken during ascent and descent is almost all of the sensor output signal change amount. In the manned state, in addition to the swing of the car 50 during ascending / descending, vibrations due to the movement of the passenger are transmitted to the sensor 53 through the car floor 54 and the support base 55. The shaking is dominant. From the above, the swing of the car 50 during the elevation is dominant in both the manned time and the unmanned time during normal operation. The reason why the signal source at the normal time is unattended is that it is selected as a representative case having a high frequency during the normal time.

  The details of the function for calculating the false alarm rate and minimizing the false alarm rate will be described below.

The threshold calculation unit 16 calculates the false alarm rate in the flow of FIG. First, the normal condition (for example, the output signal of the sensor 53 when the car 50 is lifted or lowered) is collected by the sample signal collecting unit 15, and the procedure similar to S10 in FIG. 8 is performed based on the collected data. The peak value ρ of the change amount of the sensor output signal (normal sensor sample signal) is calculated (S20). The normal sensor sample signal is collected in advance in the same manner as the car 50 even in the representative car. The normal in the car 50 (structure) with respect to the peak value ρo of the change amount of the normal sensor sample signal (that is, the reference normal sensor sample signal) of the representative car (representative structure) registered in the database in advance. The peak value ρ of the change amount of the hour sensor sample signal is calculated as the ratio γ as shown in Equation 12 (S21).
[Equation 12]
γ = ρ / ρo
Next, the false alarm rate is calculated from the frequency distribution of normality of the car 50 (normal abnormality frequency distribution) and the ratio γ (S22). Here, the “abnormality at normal time” is described above by calculating the statistical value of the strength of car vibration at normal times (in this embodiment, the amount of change in the output signal of the sensor 53). It is processed and shown on a scale (abnormality) representing anomaly, and does not mean anomalous behavior. This normal frequency distribution of abnormalities, that is, data of normal abnormality frequency distributions, is created by collecting sensor output signal changes during normal lifting operation, and is representative of the cage (see Fig. 13 (shown by reference numeral 181 of FIG. 13) is previously stored in the database as a reference normal-time abnormality degree frequency distribution. FIG. 13 shows the frequency distribution 181 of the abnormality level when the representative car is normal (normal frequency frequency distribution when normal) 181 and the frequency distribution 191 of the abnormality level when the car 50 is normal. The frequency distribution 191 of the car 50 is represented by the graph of FIG. 13 by multiplying the frequency distribution 181 of the representative car registered in advance by the above-described ratio γ (where γ is the formula 12 or the formula 13). Shift to the indicated position.

The vibration force of the car 50 accompanying the raising and lowering of the car 50 is transmitted to the support base 55 through the housing of the car 50 and appears as the amount of change in the output signal of the sensor 53. If such a process is regarded as a normal spring model, γ in Equation 12 is considered to be a ratio of the spring constant in the normal spring model between the car 50 and the representative car. Further, as in Equations 7 and 8, the normal car 50 and the normal car abnormalities P and Po when normal can be considered to be proportional to the normal spring constants of the car 50 and the representative car, respectively. it can. Accordingly, it is assumed that Equation 13 is established between the normal car 50 and the normal car abnormalities P and Po.
[Equation 13]
P / Po = γ
With reference to FIG. 13, the process for obtaining the region 192 in which the false alarm occurs in S22 will be described. In FIG. 13, since the abnormality degree in all cases in the normal state follows the equation (13) and the frequency distribution 181 of the abnormality degree in the normal state of the representative car is registered in advance, the abnormality in the normal state of the car 50. The frequency distribution (degree of normality abnormality frequency distribution) 191 can be calculated by multiplying the frequency distribution (normal normal quantity abnormality degree frequency distribution) 181 by a ratio γ. In the abnormality degree frequency distribution 191, when the threshold π for determining abnormal behavior described above of the sensor determination unit 4 is applied to a certain point of the abnormality degree frequency distribution 191 at the normal time, the normality of the region 192 exceeding π is normal. The time abnormality degree is erroneously determined as the abnormality degree in which the abnormal behavior occurs, so that a region 192 in which a so-called misreport occurs. The ratio of the area 192 to the total area of the abnormality frequency distribution 191 is the false alarm rate.

  The display unit 17 displays the false alarm rate 192 calculated by the threshold calculation unit 16. By looking at the display of the false alarm rate 192, the operator can measure the performance of the sensor determination unit 4 in the car 50. For example, when the false alarm rate 192 is close to 100% of the frequency distribution 191, even if the sensor determination unit 4 and the comprehensive determination unit 7 are added to the image determination unit 2, an improvement in performance of the image determination unit 2 from a single unit cannot be expected. . Therefore, the false alarm rate 192 becomes a material for determining whether or not the abnormality determination with the sensor 53 of the present invention is added to the car 50.

  In addition, the display unit 17 may display the false alarm rates of the image determination unit 2 and the comprehensive determination unit 7 using the same method as described above. That is, a specific index (determination index: for example, the degree of movement of the dynamic element of the image) in the video of the monitoring camera 51 obtained under the normal conditions regarding the representative car is registered in advance as a normal normal camera output signal. Keep it. In addition, the camera abnormality degree is calculated by calculating the statistic of the normal camera output signal (the specific index in the video of the surveillance camera) regarding the representative car, and the normal normal camera abnormality degree frequency distribution is calculated as the abnormality degree. It is created on the scale (abnormality) to be shown (that is, a normal normal camera abnormality degree frequency distribution based on images is created) and registered in advance. The sample signal collecting unit 15 collects, as a normal camera sample signal, a signal having the same condition as the reference camera output signal regarding the output signal of the monitoring camera 51 when the car 50 is normal. Multiply the reference normal camera abnormality degree frequency distribution by the ratio γ ′ (γ ′ is the ratio of the normal camera sample signal (specific index in the car 50) to the standard normal camera output signal (specific index in the representative car)). Thus, an abnormal frequency distribution (normal camera abnormal frequency distribution) of the normal image regarding the car 50 is obtained. When a threshold for determining abnormal behavior in an image is applied to the normality frequency distribution of the image at the normal time, an area exceeding the threshold becomes the false alarm rate of the image determination unit 2.

Here, the false alarm rate 192 of the sensor determination unit 4 of the car 50 is Ew, the false alarm rate of the image determination unit 2 is Ei, the determination of the comprehensive determination unit 7 is the abnormality determination of the sensor determination unit 4 and the abnormality determination of the image determination unit 2 Assuming that the false alarm rates of the sensor determination unit 4 and the image determination 2 are calculated independently, the false alarm rate Ec of the comprehensive determination unit 7 can be calculated by Equation 14.
[Equation 14]
Ec = Ei × Ew
By looking at the false alarm rates Ec and Ei, it is possible to easily measure the degree of improvement of the false alarm rate Ec when the sensor determination unit 4 and the comprehensive determination unit 7 are added from the single false alarm rate Ei of the conventional image determination unit 2. It becomes.

The threshold value calculation unit 16 may determine the threshold value of the image determination unit 2 by the following process. The unreported rate 92 of the sensor determination unit 4 of the car 50 is set to Mw, the unreported rate of the image determination unit 2 is set to Mi, and the determination of the comprehensive determination unit 7 is determined as an abnormality determination of the sensor determination unit 4 and the abnormality determination of the image determination unit 2 Assuming that the unreported rates of the sensor determination unit 4 and the image determination 2 are calculated independently, the unreported rate Mc of the comprehensive determination unit 7 can be calculated by Equation 15.
[Equation 15]
Mc = Mi + Mw
The threshold value calculation unit 16 also calculates the false alarm rate and the false alarm rate of the image determination unit 2 when (Ei ₁ , Mi ₁ ), (Ei ₂ , Mi) when the abnormality determination threshold value of the image determination unit 2 is changed in several ways. ₂ ) and the like, and similarly, the false alarm rate and the false alarm rate of the sensor determination unit 4 when the abnormality determination threshold value of the sensor determination unit 4 is changed in several ways are (Ew ₁ , Mw ₁ ), (Ew ₂ , Mw ₂ ) and the like, and a combination of threshold values for abnormality determination of the image determination unit 2 and the sensor determination unit 4 that gives the best evaluation of the misreporting rate Ec and the unreporting rate Mc of the comprehensive determination unit 7 is selected. May be. For the evaluation of the misreporting rate Ec and the unreporting rate Mc, it is conceivable to minimize the misreporting rate Ec while keeping the unreporting rate Mc below the target value, but the present invention is not limited to this. The threshold calculation unit 16 sets the abnormality determination thresholds of the image determination unit 2 and the sensor determination unit 4 obtained in the above process in the image determination unit 2 and the sensor determination unit 4, respectively.

  FIG. 14 shows a functional configuration of the third embodiment of the present invention. 14 also includes a monitoring camera 51, a sensor 53 for detecting abnormal behavior, an image determination unit 2, a sensor determination unit 4, a comprehensive determination unit 7, and a reporting unit 8, similar to the embodiment of FIG. The function is the same as in the first embodiment. In this embodiment, instead of the sample signal collection units (5, 15) and the threshold calculation units (6, 16) for the sensor determination unit 4 described in the first and second embodiments, a structure information input unit (in other words, a sensor An information input unit 1 regarding the installation target) 1 and a threshold setting unit 26 for the sensor determination unit 4 are provided.

  First, the outline of the present embodiment will be described. The structure information input unit 1 is for inputting information related to a structure (a car 50 in the present embodiment) to be installed on the sensor 53. The worker inputs information on the specifications and structural characteristics (for example, model) of the car 50 to the signal processing device 52 via the structure information input unit 1. Then, the threshold setting unit 26 in the signal processing device 52 receives the structure information from the input unit 1 and searches for a determination threshold (an abnormal behavior determination threshold) in the sensor determination unit 4 according to the received information. Calculate and set. For this threshold setting, threshold data registered in advance in a data storage unit 3 (for example, a database, hereinafter referred to as “threshold DB”) 3 is used. The threshold data will be described later with reference to FIG.

  First, the structure information input unit 1 will be described with reference to FIG. FIG. 15 shows a screen of the user interface 200 constituting the input unit 1.

  The user interface 200 includes an input unit 201 for inputting the model of the car 50, an input unit 211 for inputting the number of anti-vibration rubbers, and an input unit 212 for inputting the type of anti-vibration rubbers. Is illustrated. These input units may include at least one.

  The input units 201, 211, and 212 may use other graphical user interface components such as pull-down menus and radio buttons instead of the text boxes.

  When the worker inputs the model of the car 50 to the input unit 201, the threshold setting unit 26 receives the model information from the input unit 1, and from the threshold DB3, the model-threshold data table (data) as shown in FIG. 250 is read, a threshold value for the sensor determination unit 4 corresponding to the model of the car 50 is searched (selected) from the model-threshold data table 250, and this is output to the sensor determination unit 4.

  A table 250 in FIG. 16 shows an example of a model-threshold data table, and threshold values suitable for various types are registered. For example, π1 is associated with model 1, π2 with model 2, and π3 with models 3, 4, and 5.

  Such a table 250 is, for example, a method as described in the first embodiment for each type of typical car with respect to the reference threshold value πo of the representative car obtained in the first embodiment. The threshold value π is obtained using the equation (11), and the threshold values are created by associating these threshold values with a model in a table. Such a table 250 is registered in advance in the threshold value DB 3. The concept of the present embodiment is that the same type of car 50 has a similar structure, and therefore the variation in the installation location of each car 50 is considered to be small. Therefore, the determination threshold value of the sensor determination unit 4 is shared by the type. It is based on the idea that it may be. Further, a margin in consideration of variations in the installation location may be added to the threshold value in the table 250 as shown in Equation 11. Further, in the example of the table 250, the abnormality determination threshold value of the sensor determination unit 4 may be shared in a model having a similar structure such that π3 is output for the models 3, 4, and 5. The table 250 may be made to correspond to specifications (for example, the number of anti-vibration rubbers and the anti-vibration rubber types) instead of corresponding to the type. This is because if the number of anti-vibration rubbers and the type of anti-vibration rubbers are close, the threshold value π may be close. Alternatively, the threshold value π that is actually set may be replaced with a threshold value that is smaller than the optimum value. Setting a threshold value that is smaller than the optimum value in this way corresponds to the shift of the threshold value π to the left in FIG. 10, so that the unreported rate 92 does not deteriorate more than the target value.

  In the above description of the present embodiment, by inputting the model of the car 50 to the user interface 200, the threshold value π for the sensor determination unit 4 corresponding thereto is automatically calculated by the threshold setting unit 26 based on the table 250 (FIG. 16). In the example described above, the threshold value π setting function using the following calculation method is provided instead.

For example, when the worker inputs the number of anti-vibration rubbers and the type of anti-vibration rubbers as the specification of the structure information in the input unit 211 or the input unit 212, the threshold setting unit 26 sets the thresholds according to the specifications of the structure information. π is calculated by a predetermined calculation method. As an example of this calculation method, first, as parameters relating to the specification in advance, the number of vibration isolating rubbers of the representative car, the type of the anti-vibration rubber, the spring constant determined by them, or the inverse κ′o thereof, and further the parameters The calculation formula for obtaining the threshold value using is registered in advance. Further, when the number of anti-vibration rubbers input via the input unit 211 or the input unit 212 and the type of anti-vibration rubber for the car 50 are changed from those of the representative car described above, the increase / decrease of the anti-vibration rubber instead A model in which κ ′ changes according to the car 50 is constructed using the number and / or the car model and κ′o of the representative car. For example, when the number of input anti-vibration rubbers increases from the representative car, the sensor output signal is equivalent to the amount of force when the same force F is applied to the car floor 54 according to the number of anti-vibration rubbers. Considering that the change amount ΔW is small, correction is performed so that κ ′ is small. A specific method for correcting κ ′ is, for example, a method of correcting κ ′ to be small in inverse proportion to the increase / decrease number n of the anti-vibration rubber, assuming that the influence of the anti-vibration rubber is dominant (Equation 16). Not limited to this. In Expression 16, no is the number of vibration isolating rubbers of the representative car.
[Equation 16]
κ ′ = (no / n) κ′o
For example, as another example, when the spring coefficient h of the anti-vibration rubber varies depending on the model, and the car 50 is applied with a harder rubber than the representative car, the sensor when the same force F is applied to the car floor 54 is similarly applied. Considering that the output signal change amount ΔW is small, κ ′ is corrected to be small (Expression 17). In Expression 17, ho is the spring coefficient of the vibration isolating rubber of the representative car.
[Equation 17]
κ ′ = (ho / h) κ′o
When there is an input to both the input unit 211 or the input unit 212, the influences of both are made synergistic. Therefore, the following equation 18 is derived from the equation 6, the ratio α can be calculated, and the threshold π can be calculated using the equation 10.
[Equation 18]
α = κ ′ / κ′o = (no / n) × (ho / h)
It should be noted that the user interface 200 is provided with an input unit for other structural conditions such as the number and material of the outer frame of the car 50 in addition to the anti-vibration rubber conditions of the input unit 211 and the input unit 212. The threshold value π may be calculated by determining the margin coefficient η from the conditions of other structures in the equation. However, here, the value is not limited to the range 0 to 1 of the margin coefficient η in Equation 11, and the coefficient corresponding to the margin coefficient η may take an arbitrary value of 0 or more.

  For example, for a type that differs only in type 1 and the number of anti-vibration rubbers, type 1 is regarded as a representative car, and no and πo are replaced with the number n1 of anti-vibration rubbers of type 1 and the threshold value π1 to set the threshold value π. You may calculate.

  Further, the threshold value used for the image determination unit 2 of the camera image is registered in advance in association with the type and specification of the elevator car structure in advance, similarly to the threshold value of the sensor determination unit 4. When the information input unit 1 inputs a model or specification, the threshold value setting unit 27 for the camera may select and set a threshold value for the image determination unit 2 corresponding thereto.

  In the third embodiment described above, the structure information input unit 1 holds the specifications of the car 50 and information on the structural features (models, etc.) in the storage unit of the structure information input unit 1 in addition to using the user interface. In addition, it may be realized by a processing device that returns the information when the information is requested to the structure information input unit 1. For example, at the stage of installation work of the car 50, information on the specifications and structural features of the car 50 is recorded in the processing unit 52 in a storage unit included in the structure information input unit 1, and the threshold setting unit 26 detects the sensor determination unit 4. If the structure information input unit 1 is configured to input the specification data of the car 50 and the record data of the structural features (model type, etc.) to the threshold setting unit 26 when the threshold value of the structure is obtained, the structure information input unit 1 The same function as when the user interface is used can be realized.

  In the above description of the first to third embodiments, the case where the number of sensors 53 is one has been described. However, the number of sensors 53 may be two or more. If the threshold value for abnormality determination of the sensor determination unit 4 is calculated for each sensor 53 in the same manner as in each embodiment, the number of sensors 53 may be two or more.

  In the above description of the first to third embodiments, the sensor output signal change amount of the signal source in the sample signal collecting unit 5 is directly measured by the sensor 53, but may be replaced with a measured value measured using another sensor. good. This is, for example, the correspondence between the sensor output signal change amount and the alternative sensor signal change amount when a signal source such as the drop of the weight 60 or the vibration of the vibrator 61 is generated in advance. The change amount of the output signal of the alternative sensor with respect to the signal source in the sample signal collecting unit 5 is converted into the change amount of the sensor output signal. As an example of the alternative sensor, there is a method in which a vibration meter is attached to the weight 60 and a change amount when the weight 60 falls on the car floor 54 is read by the vibration meter. The change amount of the alternative sensor is determined by connecting the alternative sensor to the processing device 52 or the information terminal 59, recording the alternative sensor data on the medium and then outputting the data to the processing device 52, or measuring the alternative sensor measurement value. Can be used for each function in the processing device 52 by reading it and inputting it to the information terminal 59, but is not limited to this example.

  In the above description of the first to third embodiments, the drop of the weight 60 or the vibration of the vibrator 61 is used as the signal source in the sample signal collecting unit 5, but it may be more easily replaced by the operation of the worker. . The fall of the weight 60 can be realized by, for example, dropping the height of the platform when the worker jumps off the platform of a predetermined height, or dropping the length of the sole of the foot by landing with a heel from standing on the toe. it can. An alternative to the vibration of the vibrator 61 can be realized, for example, when an operator performs an operation such as bending or stretching. The influence of the variation in operation can be considered for each worker. To cope with this variation, the margin coefficient η is set according to the variation in operation for each worker as shown in Equation 11 and multiplied by the ratio α. The threshold value π for abnormality determination of the sensor determination unit 4 in the car 50 may be corrected to be small. Correcting the threshold value π to a small value corresponds to shifting the threshold value π to the left in FIG. 10 and does not increase the unreported rate 92 even if it is decreased. Further, the weight of the worker may be measured from the output signal of the sensor 53 that is a load sensor, and the influence of the variation in operation may be suppressed for each worker.

  In the above description of the first to third embodiments, the sensor 53 is a load sensor, but the sensor 53 can be a vibration sensor in the car 50 other than the load sensor. For example, the sensor 53 can be replaced with a sensor that measures the tension of the rope 56 and a sensor that measures the acceleration at the end of the rope. When applied to the sensor 53 of the vibration sensor other than the load sensor, when the load sensor is used as the sensor 53, the abnormal behavior and the force exerted by the signal source of the sample signal collecting unit 5 are transmitted to the car floor 54 and the support base 55, and two types In the same way as the sensor output signal change amount ΔW, the abnormal behavior and the force exerted by the signal source of the sample signal collecting unit 5 are transmitted through a predetermined path in the vibration sensor other than the load sensor and appear in the sensor output signal change amount. Can be represented in the same manner as the equations (1) to (3). However, in the sensor 53 of the vibration sensor other than the load sensor, it is necessary to appropriately select the signal source of the sample signal collecting unit 5 according to the process appearing in the change of the output signal of the sensor 53 in the predetermined path.

  Even when the load sensor is the sensor 53 and the abnormal behavior and the vibration caused by the signal source of the sample signal collecting unit 5 are transmitted to the sensor 53 through a path other than the car floor 54 and the support base 55, the sensor 53 described above is not a load sensor. It can be handled in the same way as the case. For example, in an elevator in which abnormal behavior that crawls and breaks a wall occurs intensively, a case in which vibration that crawls the wall is detected by a sensor 53 using a load sensor corresponds. Further, when the load sensor is the sensor 53 and the amount of change in the output signal of the sensor 53 is modeled by the force applied to the car floor 54 other than Equation 2, the above-described sensor 53 is other than the load sensor. It can be handled equally. For example, this corresponds to the case where an attenuation term is added to Equation (2). As described above, when the load sensor is the sensor 53, abnormal behavior and vibration caused by the signal source of the sample signal collecting unit 5 are transmitted to the sensor 53 through a path other than the car floor 54 and the support base 55, and other than the formula 2 When modeling a sensor output signal change amount when a force is applied to the car floor 54, they may be combined.

  As the sensor 53, any sensor may be used as long as the threshold value can be converted by an arbitrary calculation formula using the ratio α and the reference threshold value πo as parameters, such as Formula 10 or Formula 11. . When obtaining the calculation formula, a process in which the abnormal behavior or the influence from the signal source is transmitted to the sensor 53 through a predetermined path is modeled, for example, as shown in Formula 2. For example, when the acoustic sensor is the sensor 53, a process in which a speaker with a constant volume is used as the signal source of the sample signal collecting unit 5 and the abnormal behavior and the sound generated by the signal source are diffused in the car 50 and transmitted to the sensor 53. May be modeled using parameters related to sound diffusion.

  The crime prevention system (abnormal behavior detection system) of the present invention is not limited to the structure in the elevator described in the embodiment, but can be any other structure as long as it is a monitoring system using image recognition and sensors. Can be widely applied.

1: structure information input unit, 2: image determination unit, 3: threshold database, 4: sensor determination unit, 5: sample signal collection unit, 6: threshold calculation unit, 7: comprehensive determination unit, 8: reporting unit, 51: surveillance camera, 53: sensor.

Claims (14)

A surveillance camera installed on a structure to be monitored and photographing a person in the structure;
An image determination unit that recognizes an image of a person photographed by the monitoring camera and determines an abnormal behavior of the person;
A sensor installed in the structure for detecting an abnormal behavior of a person in the structure;
In a security system for a structure comprising a sensor determination unit that performs abnormal behavior determination by comparing a signal obtained by processing the output signal of the sensor with a threshold for determining abnormal behavior,
A sample signal collection unit that captures an output signal of the sensor as a sensor sample signal when a signal source having a predetermined size is added to the structure;
A reference value that is registered in advance and is a reference value that is a reference value that is a peak value of a change amount of the output signal of the sensor when the signal source is added to a representative structure. A threshold value for determining the abnormal behavior based on the reference threshold value set based on the frequency distribution of the degree of abnormality in the representative structure and the peak value of the change amount of the sensor sample signal. It is calculated, before the larger the ratio of peak value of the sensor sample signal for hexane Telč increases to increase the threshold value, so as to reduce the threshold as the ratio decreases, registered in advance and the ratio And a threshold value calculation unit that calculates the threshold value according to the structure by a predetermined calculation formula using a reference threshold value as a parameter. In claim 1,
The threshold value for determining the abnormal behavior is at least one of when installing the structure equipped with the sensor, installing the sensor on the existing structure, or maintaining the car equipped with the sensor. A crime prevention system for structures characterized by being calculated . In claim 1 or 2,
The sensor is a load sensor installed at a lower part of the structure;
The signal source is at least one of a weight that applies an impact to the floor of the structure by dropping, a vibrator that is placed on the floor surface and vibrates the floor, and an operation of an operator. A crime prevention system for structures. In any one of Claims 1 thru | or 3,
The structure crime prevention system, wherein the predetermined calculation formula is a calculation formula obtained by multiplying the pre-registered reference threshold by the ratio as a coefficient, or a calculation formula obtained by further multiplying a margin coefficient. In any one of Claims 1 thru | or 4,
The threshold value calculation unit changes the predetermined calculation formula between when the ratio of the sensor sample signal to the reference value is greater than 1 and when the ratio is smaller, the crime prevention system for a structure. A surveillance camera installed on a structure to be monitored and photographing a person in the structure;
An image determination unit that recognizes an image of a person photographed by the monitoring camera and determines an abnormal behavior of the person;
A sensor installed in the structure for detecting an abnormal behavior of a person in the structure;
In a security system for a structure comprising a sensor determination unit that performs abnormal behavior determination by comparing a signal obtained by processing the output signal of the sensor with a threshold for determining abnormal behavior,
A sample signal collection unit that captures an output signal of the sensor as a sensor sample signal when a signal source having a predetermined size is added to the structure;
When the amount of change in the sensor output signal is defined as the sensor output signal change amount, the ratio of the peak value of the sensor output signal change amount of the sensor sample signal to the previously registered reference value for comparison increases as the sensor output signal change amount increases. Threshold calculation for calculating the threshold according to the structure by a predetermined calculation formula using the ratio and a pre-registered reference threshold as parameters, such that the threshold is increased and the threshold is decreased as the ratio is decreased. And
Further, the crime prevention system has previously registered the amount of change in the output signal of the sensor obtained under specific conditions at the normal time relating to the representative structure representing the structure as a reference sensor output signal change amount, and The frequency distribution of the degree of abnormality processed by the sensor determination unit from the amount of change in the sensor output signal related to the representative structure at the normal time is registered as a reference normal degree abnormality degree frequency distribution,
The sample signal collecting unit collects, as a normal sensor sample signal, the same condition as the reference sensor output signal with respect to the sensor output signal when the structure is normal,
The threshold calculation unit includes a ratio of a peak value of the sensor output signal change amount of the normal sensor sample signal to a peak value of the reference sensor output signal change amount, a normal abnormality degree frequency distribution of the reference, and the sensor From the threshold value for determining the abnormal behavior of the sensor, a normal frequency error frequency distribution of the normal sensor sample signal and an error rate of the abnormal behavior determination by the sensor determination unit are also calculated, and the error rate is displayed on the display. A structure crime prevention system characterized by being configured to display via An elevator comprising the security system for a structure according to any one of claims 1 to 6, wherein the structure is an elevator car. A surveillance camera installed on a structure to be monitored and photographing a person in the structure;
An image determination unit that recognizes an image of a person photographed by the monitoring camera and determines an abnormal behavior of the person;
A sensor installed in the structure for detecting an abnormal behavior of a person in the structure;
In a security system for a structure comprising a sensor determination unit that performs abnormal behavior determination by comparing a signal obtained by processing the output signal of the sensor with a threshold for determining abnormal behavior,
A structure information input unit for designating at least one of the model of the structure and the specification of the vibration isolating rubber of the structure;
A plurality of models or specifications are registered for at least one of the above models and specifications,
And a data storage unit having a data table in which a plurality of threshold values for detecting abnormal behavior are registered in association with each other,
A threshold setting unit configured to search and set the threshold from the data according to at least one of the model and specification specified in the structure information input unit,
The threshold for detecting the abnormal behavior is
Reference value for comparison registered in advance, the reference value that is a peak value of the amount of change in the output signal of the sensor when a signal source of a predetermined size is added to the representative structure, Reference threshold values that have been registered, the reference threshold value set based on the frequency distribution of the degree of abnormality in the representative structure, and the output signal of the sensor when the signal source is added to the structure A structure crime prevention system characterized in that the threshold value is calculated based on a peak value of a change amount, and is a threshold value associated with each of the plurality of structure types . In claim 8,
A structure crime prevention system characterized in that the abnormal behavior determination threshold value is corrected in accordance with at least one of the model or specification input to the structure information input unit. In claim 9,
In the data storage unit, a parameter corresponding to the specification and a calculation formula for obtaining the threshold value using the parameter are registered,
When the specification is specified by the structure information input unit, the threshold setting unit calculates the threshold by the parameter corresponding to the specification and the calculation formula for calculating the threshold using the parameter. A crime prevention system for structures. In any one of Claims 8 thru | or 10,
A security system for a structure, wherein the sensor is a load sensor and the model is an elevator model . In claim 11,
A security system for a structure, wherein the specification is at least one of the number and model of the elevator car related to the anti-vibration rubber . In claim 12,
The calculation formula includes, as the parameter, a spring constant calculated from the number and model of the anti-vibration rubber,
The threshold setting unit corrects the spring constant based on the number of the anti-vibration rubbers input to the structure information input unit or the type of the anti-vibration rubbers, and detects an abnormal behavior of the structure to be monitored. A security system for structures characterized by calculating a threshold value for use . An elevator comprising the security system for a structure according to any one of claims 8 to 13, wherein the structure is an elevator car.

Images