JP6226368B2 - Vehicle monitoring apparatus and vehicle monitoring method - Google Patents

Description

  Embodiments described herein relate generally to a vehicle monitoring apparatus and a vehicle monitoring method for reading information written on a license plate of a vehicle from an image captured by a camera.

  2. Description of the Related Art An apparatus that captures a road or the like with a camera and reads information described on a vehicle number plate in the captured image is known (for example, see Patent Document 1). This type of system is attached to, for example, an entrance or exit of a toll road and used to recognize a passing vehicle.

Patent No. 4901676

In the conventional apparatus, there are cases where the recognition accuracy cannot be maintained in an environment where the position and angle of the license plate vary from image to image.
The problem to be solved by the present invention is to provide a vehicle monitoring apparatus and a vehicle monitoring method capable of improving recognition accuracy.

The vehicle monitoring apparatus according to an embodiment includes a vehicle detection unit, a geometric processing unit, a reading unit, and an output unit . A vehicle detection part detects a vehicle in the image imaged by the imaging part attached to the position which can image a vehicle. The geometric processing unit reads the correction parameter corresponding to the vehicle position detected by the vehicle detection unit from the storage unit that stores the correction parameter associated with the vehicle position in the image, and uses the read correction parameter. The geometric processing that has been performed is performed on the image. The reading unit reads information described on the license plate of the vehicle from the image subjected to the geometric processing. The output unit scores with respect to the correction parameter such that the smaller the angle for correcting the angle and size information corrected in the geometric processing, the higher the score, and the number plate in the image captured by the imaging unit Is extracted for a plurality of coordinates at the coordinate position including the boundary portion where the visual field can be visually recognized, and the fact that the visual field setting is appropriate is output from comparison with a standard score that enables license plate recognition.

It is the figure which illustrated a mode that the cameras 10 and 20 contained in the vehicle monitoring system 1 of one Embodiment were attached to the gantry GT. It is a figure which shows an example of a function structure of the vehicle monitoring apparatus 100 in an operation stage. It is the figure which illustrated a mode that the number plate NP of a vehicle inclines with respect to an image. It is the figure which illustrated a mode that the image was divided | segmented into the areas 1-3. It is a figure for demonstrating an example of the measuring method by the vehicle type determination part. It is the figure which illustrated a mode that a vehicle moved for every time. It is the figure which illustrated the scene where a license plate is difficult to visually recognize by the shield of a vehicle body. It is a figure which shows an example of the image which imaged the vehicle which the attachment position of a license plate is other than the center part of a vehicle. 6 is a diagram illustrating an example of information registered in a correction parameter database 142. FIG. It is a figure which shows an example of the image information which the information registered into the correction parameter database 142 represents. It is a figure which shows typically a mode that an affine transformation is performed with respect to the image of a license plate, and a template is applied. It is a figure which shows an example of the information which the recognition result output part 128 outputs (it memorize | stores in the memory | storage part 140). It is a figure which shows an example of a function structure of 100 A of vehicle monitoring apparatuses in the correction parameter collection stage. It is the figure which illustrated a series of images which picturized a vehicle which runs on the right side of a road, a vehicle which runs on the center, and a vehicle which runs on the right side, respectively. It is a figure which shows typically the process performed by the vehicle number candidate detection part 112 and the NP candidate geometric correction | amendment part 118 in a correction parameter production | generation stage. It is a figure which shows an example of the captured image recognized by dividing | segmenting at the time of recognizing a vehicle on the 1st driving lane side and the 2nd driving lane side, respectively.

  Hereinafter, embodiments of a vehicle monitoring device and a vehicle monitoring method will be described with reference to the drawings. FIG. 1 is a diagram illustrating a state in which cameras 10 and 20 included in a vehicle monitoring system 1 according to an embodiment are attached to a gantry GT. The gantry GT is, for example, an automatic fee collection system gantry. The cameras 10 and 20 are cameras including an image sensor such as a CCD (Charge Coupled Device) or a CMOS (Complementary Metal Oxide Semiconductor).

  The camera 10 images the front surface of the vehicle MB passing close to the gantry GT from an obliquely upward direction. In the figure, A (10) indicates an imaging region of the camera 10. Further, the camera 20 captures an image of the rear surface of the vehicle MB passing through the gantry GT from an obliquely upward direction. In the figure, A (20) indicates an imaging region of the camera 20. For example, the cameras 10 and 20 perform imaging at a depression angle of about 45 [degrees] from 7 [m] above the road. Further, the cameras 10 and 20 are controlled so as to repeatedly perform imaging at a predetermined cycle. The installation locations of the cameras 10 and 20 are not limited to those shown in FIG. 1, but may be any locations where the front and rear surfaces of the vehicle can be imaged. Images captured by the cameras 10 and 20 are transmitted to the vehicle monitoring apparatus 100.

  In this way, by imaging from the front and rear of the vehicle MB, the license plate can be recognized even for a vehicle having a license plate attached only to the rear, such as a two-wheeled vehicle. Further, when the road is congested, even if the number plate of the subsequent vehicle MB is hidden in the image captured from the front of the vehicle MB, the number of the subsequent vehicle MB in the image captured from the rear of the vehicle MB. It may be possible to recognize the plate.

[Operation stage]
The vehicle monitoring apparatus 100 can be installed in any place. FIG. 2 is a diagram illustrating an example of a functional configuration of the vehicle monitoring apparatus 100 in the operation stage. The “operation stage” refers to a stage where the vehicle monitoring apparatus 100 actually monitors the vehicle after the correction parameters are collected. Processing at a stage where correction parameters are collected will be described later.

Vehicle monitoring apparatus 100, for example, a vehicle detection unit 102, a vehicle state Determination unit 104, the recognition position determination unit 106, a vehicle type determination unit 108, a preprocessing unit 110, a vehicle number candidate detecting section 112, Number Plate (may be abbreviated as NP in the following description and drawings) candidate creation unit 114, correct candidate parameter reference unit 116, NP candidate geometric correction unit 118, NP candidate type determination unit 120, and NP correct candidate A selection unit 122, an NP recognition unit 124, a recognition result comparison unit 126, a recognition result output unit 128, a correct candidate parameter registration unit 130, and a storage unit 140 that stores a correction parameter database 142 are provided.

  Each functional unit other than the storage unit 140 is, for example, a software functional unit that functions when a CPU (Central Processing Unit) of the vehicle monitoring apparatus 100 executes a program stored in the storage unit 140. Alternatively, each functional unit may be a hardware functional unit such as an LSI (Large Scale Integration) or an ASIC (Application Specific Integrated Circuit). The storage unit 140 is, for example, an HDD (Hard Disk Drive), a flash memory, a RAM (Random Access Memory), a ROM (Read Only Memory), or the like.

The vehicle detection unit 102 determines whether or not a part (part) of the vehicle exists in the images captured by the cameras 10 and 20. The vehicle detection unit 102, for example, for each part constituting the vehicle such as a front part, a body part, and a rear part of the vehicle, feature quantities that are edge-enhanced by Sobel filter processing, gradient feature quantities, luminance feature quantities, etc. A discriminator composed of these patterns is constructed (see the following reference). And the vehicle detection part 102 pinpoints the kind and position of the site | part of the vehicle which exist in an image using the comprised discriminator.
Reference: “Coarse-to-fine approach for fast deformable object detection”, M.Pedersoli, A.Vadaldi, J.Gonzalez, CVPR2011

  Moreover, the vehicle detection part 102 tracks the lower end of the position of a vehicle, estimates the advancing direction of a vehicle from a tracking result, and estimates the present vehicle position and the inclination of a license plate. FIGS. 3A and 3B are diagrams illustrating a state in which the license plate NP of the vehicle is tilted with respect to the image. At this time, if the traveling direction of the vehicle curves more than a certain angle with respect to the optical axis of the camera, the angle of the license plate also increases in conjunction with the movement. When the magnitude of the angle is equal to or larger than the predetermined angle, the vehicle detection unit 102 stores the travel locus of the vehicle in the storage unit 140 and saves it. This is because when the traveling direction of the vehicle is greatly inclined with respect to the optical axis of the camera, there is a high possibility that the vehicle is skewed and the necessity of monitoring the vehicle is high.

  The vehicle state determination unit 104 virtually divides the images captured by the cameras 10 and 20 into, for example, three division lines in the horizontal direction of the image, and determines a rough position of the vehicle (in which area). FIG. 4 is a diagram illustrating a state in which the image is divided into regions 1 to 3.

  Based on the result of the vehicle state determination unit 104, the recognition position determination unit 106 detects a portion corresponding to the lower end of the vehicle in each region. The recognition position determination unit 106 performs straight line extraction by, for example, Hough transform, and extracts the most prominent straight line existing around the front part of the vehicle as a representative sample corresponding to the lower end of the vehicle.

  The vehicle type determination unit 108 first measures the vehicle width from the outer contour of the vehicle in the image. FIG. 5 is a diagram for explaining an example of a measurement method performed by the vehicle type determination unit 108. For example, the vehicle type determination unit 108 measures the vehicle width based on the road width L (y) for each vertical position y of the image stored in the storage unit 140 in advance and the inclination θ in the road width direction in the image. That is, if the lower vector of the vehicle outer contour is set to → W (→ indicates a vector), the vehicle type determination unit 108 maps this to a vector → W * parallel to the road width direction. Then, the vehicle type determination unit 108 obtains a ratio between the vector → W * and the road width l (y) on the image at the vertical position y, and the road width L () where the starting point of the vector → W * is the vertical position y. The vehicle width is measured by multiplying y) by the obtained ratio.

  Further, the vehicle type determination unit 108 determines the vehicle type of the vehicle shown in the image by integrating the shape pattern of the front part and the like detected by the vehicle detection unit 102 and the vehicle width measured as described above. The vehicle type determination unit 108 classifies the vehicle type into “two-wheeled vehicle”, “light vehicle”, “normal vehicle”, “large vehicle”, “other”, “indeterminate”, and the like. Among these, “two-wheeled vehicle” and “four-wheeled vehicle” are determined based on the shape pattern of the front portion, etc., and the type of “four-wheeled vehicle” is determined based on the shape pattern of the front portion and the vehicle. A determination is made based on the width.

  Here, based on the determination results of the vehicle state determination unit 104 and the vehicle type determination unit 108, the recognition position determination unit 106 waits for the vehicle to move to the region 1 or the region 2 shown in FIG. To do. Further, since the vehicle moves at each time, the recognition position determination unit 106 waits until the license plate moves to an appropriate position. FIG. 6 is a diagram illustrating a state in which the vehicle moves for each time. Here, the recognition position determination unit 106 regards the “large vehicle” in the case where it is difficult to visually recognize the license plate due to the shielding of the vehicle body (front bumper or the like), or when the license plate is attached at a position other than the central portion in the vehicle width direction. Since there is a case, the region 1 shown in FIG. FIG. 7 is a diagram exemplifying a situation where the license plate is difficult to visually recognize due to the shield of the vehicle body. In the figure, the license plate of the vehicle A is difficult to visually recognize from the cameras 10 and 20, but the license plate of the vehicle B is highly likely to be visually recognized from the cameras 10 and 20. FIG. 8 is a diagram illustrating an example of an image obtained by capturing a vehicle in which the number plate (NP) is attached at a position other than the central portion of the vehicle. Further, the recognition position determination unit 106 may set the region 3 shown in FIG. 3 as the recognition target frame for the “two-wheeled vehicle” from the viewpoint of the angle and resolution of the license plate.

  In addition, the recognition position determination unit 106 uses an image used for license plate recognition based on the type of vehicle determined by the vehicle type determination unit 108, an image showing the front of the vehicle captured by the camera 10, and the camera 20. The image may be switched between an image showing the rear surface of the vehicle imaged by.

  The preprocessing unit 110 performs extraction processing, binarization processing, and the like of components below a specific frequency by expansion / reduction processing. The vehicle number candidate detection unit 112 extracts a plurality of circumscribed rectangles of the character candidates extracted by the binarization process, and the circumscribed rectangles of appropriately sized character candidates that form a four-digit serial number are in a straight line and at regular intervals. Are extracted to create a plurality of combinations constituting the license plate (see Patent Document 1 by the present applicant).

  The NP candidate creation unit 114 creates a plurality of combinations of character candidates constituting the license plate from the above-described combinations arranged on a straight line at regular intervals in consideration of the interval and size of the number candidates.

  The correct candidate parameter reference unit 116 extracts, from the correction parameter database 142, correction parameter candidates when the NP candidate geometric correction unit 118 performs affine transformation from the combinations constituting the license plate created by the NP candidate creation unit 114. To do. The correct candidate parameter reference unit 116 extracts, from the correction parameter database 142, correction parameters for cases where the positions of the vehicles in the image are the same, for example.

  Here, the correction parameter database 142 registers candidate correction parameters having high scores in license plate recognition at each point on the road as database information. The score is, for example, a score with a higher score as the angle for correcting the angle and size information corrected in the geometric processing is smaller. FIG. 9 is a diagram illustrating an example of information registered in the correction parameter database 142, and FIG. 10 is a diagram illustrating an example of image information represented by information registered in the correction parameter database 142. “Coordinate x” and “coordinate y” in FIG. 9 indicate the position of the license plate NP in FIG. Further, “inclination dx” and “inclination dy” in FIG. 9 indicate the inclination of the lower end line FL of the vehicle MB in FIG. Further, “enlarged scx” and “enlarged scy” in FIG. 9 indicate the horizontal width ratio and vertical width ratio of the license plate NP to the image, respectively. In addition, “vehicle type” in FIG. 9 indicates the determination result by the vehicle type determination unit 108, and “vehicle center” indicates the horizontal coordinate of the center point (center of gravity) of the vehicle MB.

  Here, there may be a case where the correction parameter corresponding to the vehicle position is not registered in the correction parameter database 142. In this case, the correct candidate parameter reference unit 116 may estimate and apply the correction parameter from the difference between the registered coordinates and the coordinates. For example, the correct candidate parameter reference unit 116 may estimate the correction parameter by linear interpolation processing or the like using two or more data having the closest vehicle positions.

  The NP candidate geometric correction unit 118 performs affine transformation correction on the license plate image based on the correction parameter extracted by the correct candidate parameter reference unit 116, and performs coordinate conversion so as to face the optical axis of the camera.

  The NP candidate type discriminating unit 120 applies large / medium / small plate templates to license plate images that have been subjected to affine transformation, and serial numbers such as land branches, classification numbers, etc. for four-digit serial numbers. The candidate that the circumscribed rectangle of the character candidate other than is applied to the assumed template is determined. FIG. 11 is a diagram schematically illustrating how the template is applied after affine transformation is performed on the image of the license plate.

  The NP correct answer candidate selection unit 122 recognizes the license plate for each item and outputs the combination having the highest score for the image that has been coordinate-transformed so as to face the optical axis of the camera by the NP candidate geometric correction unit 118. To do.

  When the NP correct answer candidate selection unit 122 selects a candidate, the NP recognition unit 124 outputs the selected candidate as a license plate recognition result. In addition, when the NP correct candidate selection unit 122 does not select a candidate, the NP recognition unit 124 performs, for example, an OCR process independently, and derives and outputs a license plate recognition result.

  The recognition result comparison unit 126 determines whether or not the correction parameter used by the NP candidate geometric correction unit 118 is extracted from the correction parameter database 142 for the output as the license plate recognition result. The result is output to the correct candidate parameter registration unit 130.

  The recognition result output unit 128 outputs, for the license plate recognition result having the highest score, the license plate recognition position, the affine transformation parameters, and the large / medium / small plate template information in addition to the license plate recognition result. . FIG. 12 is a diagram illustrating an example of information output from the recognition result output unit 128 (stored in the storage unit 140). As shown in the figure, the recognition result output unit 128 outputs information described in the vehicle type and the license plate in association with the vehicle image.

  The correct candidate parameter registration unit 130 refers to the outputs of the recognition result comparison unit 126 and the recognition result output unit 128 to replace or update the correction parameter.

  By such processing, the vehicle monitoring apparatus 100 of the present embodiment can improve the license plate recognition accuracy. In a situation where the vehicle moves at various speeds, when the position of the vehicle is determined from the image processing result and used as a trigger for license plate recognition, a variation factor related to the angle of the license plate due to the vehicle position deviating from the image. Will increase. In contrast, the vehicle monitoring apparatus 100 according to the present embodiment recognizes the position of the vehicle in the image, reads out the correction parameter associated with the position of the vehicle in the image, and performs a geometric process using the read correction parameter. Since the information written on the license plate is read from the image that has been subjected to the image processing and subjected to the geometric processing, the recognition accuracy of the license plate can be improved.

  Note that the vehicle monitoring apparatus 100 in the operation stage scores the correction parameters so that the smaller the angle for correcting the angle and size information corrected in the geometric processing, the higher the score is, and the image is captured by the imaging unit. A function unit that extracts a plurality of coordinates of the coordinate position including the boundary part where the license plate can be visually recognized in the image, and outputs that the visual field setting is appropriate from comparison with a standard score that enables license plate recognition May be provided.

  Further, the vehicle monitoring apparatus 100 in the operation stage compares the distribution of information on the angle and size corrected in the affine transformation with the distribution corresponding to the recognition result within a certain time, and if the comparison result exceeds the threshold, You may provide the function part which outputs the information which shows the possibility of the position shift of 10 and 20.

[Correction parameter generation stage]
Hereinafter, the generation step of the correction parameter used as described above will be described. FIG. 13 is a diagram illustrating an example of a functional configuration of the vehicle monitoring device 100A in the correction parameter collection stage (denoted as the vehicle monitoring device 100A for convenience in order to be distinguished from the operation stage). In the correction parameter generation stage, the vehicle monitoring device 100A generates a correction parameter using the captured images of the cameras 10 and 20 installed at the same place as in the operation stage. In the correction parameter generation stage, images are taken when the vehicle travels in various positions (center, left and right, etc.) in the travel lane, and correction parameters are generated for each position of the vehicle in the image. FIG. 14 is a diagram illustrating a series of images obtained by capturing images of a vehicle traveling on the right side of the road, a vehicle traveling on the center, and a vehicle traveling on the right side.

The vehicle monitoring apparatus 100A, for example, a vehicle detection unit 102, a vehicle state Determination unit 104, the recognition position determination unit 106, a vehicle type determination unit 108, a preprocessing unit 110, a vehicle number candidate detecting section 112, NP Candidate creation unit 114, NP candidate geometric correction unit 118, NP candidate type discrimination unit 120, NP correct answer candidate selection unit 122, recognition result output unit 128, correct candidate parameter registration unit 130, and correction parameter generation unit 132 And a correction parameter display unit 134 and a storage unit 140 that stores the correction parameter database 142. Among these, those given the same name and reference numerals as those in the operation stage have the same functions as those in the operation stage, and thus description thereof will be omitted as appropriate.

  FIG. 15 is a diagram schematically illustrating processing executed by the vehicle number candidate detection unit 112 and the NP candidate geometric correction unit 118 in the correction parameter generation stage. The vehicle number candidate detection unit 112 extracts a large number of circumscribed rectangles of the character candidates extracted by binarization, and the circumscribed rectangles of appropriately sized character candidates constituting a 4-digit serial number are aligned on a straight line at regular intervals. Extract the combinations that line up, and create multiple combinations that make up the license plate. The NP candidate geometric correction unit 118 performs affine transformation correction of the license plate based on the correction parameter, and performs coordinate conversion so as to face the optical axis of the camera. At this time, assuming templates of large, medium, and small plates, determine the candidate that the circumscribed rectangle of the character candidate other than the serial number such as the land branch, classification number, etc. applies to the assumed template for the 4-digit serial number To do.

  The recognition result output unit 128 performs license plate recognition for each item on the image that has been coordinate-transformed so as to face the optical axis of the camera by the NP candidate geometric correction unit.

  The correction parameter generation unit 132 basically uses a correction parameter when the position of the vehicle in the image is affine transformed so as to face the optical axis of the camera as a correction parameter registered in the correction parameter database 142. Generate. Further, the correction parameter generation unit 132 estimates a correction parameter at a position where the detection frequency is low based on the recognition position of the license plate by performing interpolation from correction parameters at a plurality of positions located in the vicinity. This is because, for example, in the vicinity of the boundary of the visual field range, there is little traffic history of the vehicle, and there is a concern about a decrease in score. When the correction parameter for the same position is obtained from the recognition result output unit 128, the correction parameter generation unit 132 compares with the estimated parameter. If the score is high, the correction parameter generation unit 132 replaces the correction parameter or corrects the correction parameter. Re-estimate

  The correct candidate parameter registration unit 130 registers the correction parameter in the correction parameter database 142 in association with a candidate having a high score in license plate recognition for each position of the vehicle on the image. In addition, the correct candidate parameter registration unit 130 refers to the determination result by the vehicle type determination unit 108 and registers a correction parameter for each type of vehicle information such as large and medium models. In consideration of the case where the license plate exists at a position that is difficult to see from the camera depending on the vehicle type, the license plate position and the vehicle position information at this time are linked and registered.

  The correction parameter display unit 134 overlays the license plate information on the license plate information when the recognition result is output, and displays the license plate area in an overlay manner.

  The vehicle monitoring apparatus 100A performs the above operation for each frame in which the vehicle moves within the camera field of view, and registers it in the correction parameter database 142 while checking the position of the vehicle for each frame and its correction parameters. To do.

  According to the vehicle monitoring device 100A described above, the correction parameter is generated for each position of the vehicle in the image, so that the license plate recognition accuracy can be improved in the vehicle monitoring device 100 in the operation stage.

[Application to multiple lanes]
Hereinafter, the case where the vehicle monitoring apparatus 100 monitors a plurality of lanes will be described. The cameras 10 and 20 can have a vehicle detection range (imaging range) for two lanes, for example. When the vehicle travels while maintaining the driving lane, the license plate can be recognized by the above method. On the other hand, when traveling on the boundary part of the lane or moving from one lane to another lane within the vehicle detection range, it is preferable to perform license plate recognition by the method described below.

  FIG. 16 is a diagram illustrating an example of a captured image that is recognized by being divided when the vehicle is recognized on each of the first traveling lane side and the second traveling lane side. Thus, when there are a plurality of driving lanes to be monitored, the vehicle monitoring apparatus 100 may have the functional configuration shown in FIG. 2 for each lane. As shown in the drawing, after the vehicle A travels in the first travel lane, it may enter the adjacent second travel lane. In this case, the information on the first traveling lane side is taken over to the second traveling lane side. On the second lane side, there may be a scene in which the vehicle suddenly appears in the area 2, but smooth vehicle management is possible by the inherited information.

  In some cases, a plurality of vehicles are shown in a single travel lane. For example, there may be a vehicle A parked on the road shoulder and a vehicle B passing by it due to traffic jams or breakdowns. In this case, when the state in which the vehicle B travels continues a plurality of times while the vehicle A is stopped, the system detects an abnormality and sounds an alarm. Alternatively, when the place where the vehicle A is stopped is a road shoulder, it may be determined that the vehicle A is abnormal and an alarm is sounded. On the contrary, since the state in which the vehicle A travels on the shoulder is dangerous even though the vehicle B is stopped on the lane, a configuration is considered in which an alarm is sounded in consideration of the meaning of the notification. .

  In particular, when a plurality of vehicles are shown in a single lane, a case where a vehicle D determined to be a two-wheeled vehicle runs alongside the vehicle C, or vehicles E and F that are both determined to be two-wheeled vehicles are arranged side by side. There is a case to run. In this case, the order of the two vehicles may be switched between the timing at which ETC communication is performed and the timing at which license plate recognition is performed. In this case, since the vehicle monitoring device 100 operates independently of the ETC radio system, if there are two vehicles at the same time, the above-described replacement may occur. The license plate recognition result and the time log at the time of recognition are output. The user outputs to the video management system so that the user can see and confirm the video before and after the timing when the billing is performed.

  In addition, the vehicle monitoring device 100 is independent of the vehicle detection and the associated license plate recognition algorithm when there is a vehicle traveling at a low speed on the center of the road due to traffic jams or breakdowns, and the succeeding motorcycles overtake. The recognition result may be output in parallel.

According to at least one embodiment described above, the position of the vehicle is recognized in the image, the correction parameter associated with the position of the vehicle in the image is read, and geometric processing using the read correction parameter is performed on the image. On the other hand, since the information described in the license plate is read from the image subjected to the geometric processing, the recognition accuracy of the license plate can be improved.
As mentioned above, although some embodiment of this invention was described, these embodiment is shown as an example and is not intending limiting the range of invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

10, 20 Camera 100 Vehicle monitoring device (operational stage)
100A Vehicle monitoring device (correction parameter generation stage)
102 vehicle detection unit 104 vehicle state determination unit 106 recognition position determination unit 108 vehicle type determination unit 110 preprocessing unit 112 vehicle number candidate detection unit 114 NP candidate creation unit 116 correct candidate parameter reference unit 118 NP candidate geometric correction unit 120 NP candidate type determination Unit 122 NP correct answer candidate selection unit 124 NP recognition unit 126 recognition result comparison unit 128 recognition result output unit 130 correct candidate parameter registration unit 132 correction parameter generation unit 134 correction parameter display unit 140 storage unit 142 correction parameter database

Claims (11)

A vehicle detection unit for detecting a vehicle in an image captured by an imaging unit attached to a position where the vehicle can be imaged;
A correction parameter corresponding to the position of the vehicle detected by the vehicle detection unit is read from a storage unit that stores a correction parameter associated with the position of the vehicle in the image, and geometric processing using the read correction parameter is performed. Performing a geometric processing on the image;
A reading unit for reading information written on the license plate of the vehicle from the image subjected to the geometric processing;
Regarding the correction parameter, scoring is performed such that the smaller the angle for correcting the angle and size information corrected in the geometric process, the higher the score, and the license plate can be visually recognized in the image captured by the imaging unit. An output unit that extracts a plurality of coordinates of the coordinate position including a part and outputs that the field of view setting is appropriate from comparison with a standard score that enables license plate recognition;
A vehicle monitoring device comprising: The vehicle monitoring device according to claim 1,
A generation unit configured to generate the correction parameter and store the correction parameter on the basis of an image captured in a place where the imaging unit is installed;
Vehicle monitoring device. The vehicle monitoring device according to claim 2,
The generation unit generates correction parameters associated with vehicle positions not stored in the storage unit, based on correction parameters associated with vehicle positions already stored in the storage unit, Memorize in memory
Vehicle monitoring device. The vehicle monitoring device according to any one of claims 1 to 3 ,
If the distribution of the information of the angle and size corrected in the geometric processing is compared with the distribution corresponding to the recognition result within a certain time, and the comparison result exceeds a threshold value, the positional deviation of the imaging unit A comparison processing unit that outputs information indicating the possibility of
Vehicle monitoring device. The vehicle monitoring device according to any one of claims 1 to 4 ,
The imaging unit includes a first imaging unit that images a vehicle entering a facility, and a second imaging unit that images a vehicle leaving the facility,
A vehicle type determination unit for determining the type of vehicle imaged by the imaging unit;
Based on the vehicle type determined by the vehicle type determination unit, an image used for recognition of the license plate is an image captured by the first imaging unit and an image captured by the second imaging unit. A recognition position determination unit that switches between,
A vehicle monitoring device comprising: The vehicle monitoring device according to claim 5 ,
A vehicle type determination unit for determining the type of vehicle imaged by the imaging unit;
The recognition position determination unit selects an image to be used for recognition of the license plate based on the type of the vehicle determined by the vehicle type determination unit and the position of the part of the vehicle in the image captured by the imaging unit. ,
A vehicle monitoring device comprising: The vehicle monitoring device according to any one of claims 1 to 6 ,
The vehicle detection unit recognizes the lower end of the position of the vehicle, and when the inclination with respect to the image of the lower end is equal to or greater than a predetermined angle, the vehicle traveling locus is stored in the storage unit.
Vehicle monitoring device. The vehicle monitoring device according to any one of claims 1 to 7 ,
A first processing unit that performs processing on a first travel lane, the first processing unit including the vehicle detection unit, the geometric processing unit, and the reading unit;
A second processing unit that performs processing on a second traveling lane adjacent to the first traveling lane, the second processing unit including the vehicle detection unit, the geometric processing unit, and the reading unit; With
The first processing unit provides the second processing unit with information related to a vehicle entering the second traveling lane from the first traveling lane, and the second processing unit is configured to Providing the first processing unit with information about a vehicle entering the first travel lane from the travel lane;
Vehicle monitoring device. The vehicle monitoring device according to any one of claims 1 to 8 ,
Based on the state of the stopped vehicle in the image captured by the imaging unit, an alarm output is performed.
Vehicle monitoring device. The vehicle monitoring device according to any one of claims 1 to 9 ,
When a plurality of vehicles are included in the image captured by the imaging unit, the vehicle detection unit, the geometric processing unit, and the reading unit are operated for each.
Vehicle monitoring device. Computer
In an image captured by an imaging unit attached to a position where the vehicle can be imaged, the vehicle is detected,
A correction parameter corresponding to the detected vehicle position is read out from a storage unit storing a correction parameter associated with the vehicle position in the image, and geometric processing using the read correction parameter is performed on the image. Against
Read the information written on the license plate of the vehicle from the geometrically processed image,
Regarding the correction parameter, scoring is performed such that the smaller the angle for correcting the angle and size information corrected in the geometric process, the higher the score, and the license plate can be visually recognized in the image captured by the imaging unit. A vehicle monitoring method for extracting a plurality of coordinates of the coordinates of a coordinate position including a portion, and outputting that the visual field setting is appropriate based on comparison with a standard score that enables license plate recognition .

Images