JP2009157492A - Vehicle detection device, vehicle detection system, and vehicle detection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a vehicle detection device, a vehicle detection system and a vehicle detection method for detecting a vehicle whose prescribed display light is turned on/off which used to be difficult to detect in conventional image processing. <P>SOLUTION: A time difference image generation part 13 generates the time difference image of two pickup images whose pickup time is different. A line image extraction part 14 extracts a line image having the characteristic section of a luminance value from the time difference image. A space-time image generation part 16 generates a space-time image by time-sequentially arranging a plurality of line images whose imaging time is different. A space-time luminance waveform generation part 18 generates a space-time luminance waveform by adding the luminance values at the same imaging time of the space-time image, and a Fourier conversion part 19 carries out Fourier transformation to the space-time luminance waveform. A stop vehicle decision part 21 decides the presence/absence of a stop vehicle based on a power spectrum after Fourier transformation. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a vehicle detection device that detects a vehicle blinking a display lamp such as a hazard lamp or a revolving light, a vehicle detection system including the vehicle detection device, and a vehicle detection method.

  In order to realize safe and smooth urban traffic, it is necessary to quickly and accurately grasp traffic events, accidents, or other outbreaks that occur on a daily basis. It is necessary to implement a solution. For this reason, a technology has been developed in which vehicles such as roads and intersections are imaged with an imaging device such as a video camera, and image data obtained by imaging is image-processed to detect the operation state of the vehicle (for example, a stopped vehicle). ing.

  For example, an image for classifying vehicle lamps based on hue information and lightness information obtained from image data, and discriminating the operation status of the vehicle (for example, traffic jam conditions, stopped vehicles, etc.) based on the operation status of the classified lamps A processing apparatus is disclosed (see Patent Document 1).

  Also, the video signal of the area including the right turn lane at the intersection is image-processed to obtain the average luminance, the video signal is binarized based on the average luminance, and the case where the presence or absence of the binarized output is alternately repeated at a predetermined cycle is detected. And the right-turn vehicle detection method which determines the presence or absence of the vehicle which is stopping trying to turn right in the intersection is disclosed (refer patent document 2).

Further, in the conventional image processing, for example, when detecting a stopped vehicle, the vehicle region is extracted and labeled by the difference between the captured image when the vehicle is present and the captured image when the vehicle is not present, and is labeled. The lower shadow position of the obtained vehicle is detected as the position of the vehicle, the temporal change of the detected position is specified as a locus, and the stopped vehicle is detected based on the movement distance.
JP 2001-216597 A Japanese Patent Laid-Open No. 6-195590

  However, when a stopped vehicle is detected by Patent Document 1, Patent Document 2, and conventional image processing, if the stopped vehicle is stopped by blinking an indicator lamp (for example, a hazard lamp), the luminance is increased by blinking the hazard lamp. The change becomes large, and the detected coordinates of the vehicle on the captured image vary greatly from frame to frame. For this reason, it is difficult to determine whether the detected vehicle is the same vehicle, and there is a problem that it cannot be detected whether the vehicle is running or stopped.

  Conversely, in order to reduce the luminance change due to blinking of the hazard lamp, it is necessary to adjust the imaging conditions according to the characteristics of each video camera or the road conditions where the video camera is installed. However, the adjustment cost for adjusting each place where the video camera is installed becomes enormous and is not practical. For this reason, an image processing system that does not depend on a video camera or road conditions has been desired.

  In addition, when the imaging condition of the video camera is adjusted so as to reduce the luminance change due to blinking of the hazard lamp, there is a problem that the vehicle cannot be detected when the hazard lamp is extinguished. Furthermore, in addition to the hazard lamp, when there is a blinking object such as a visual guidance indicator that repeatedly blinks (blinks) at a different period from the blinking period of the hazard lamp, there is a problem that the vehicle is erroneously determined as a stopped vehicle. .

  The present invention has been made in view of such circumstances, and a vehicle detection apparatus capable of detecting a vehicle blinking a predetermined indicator lamp, which has been difficult to detect by conventional image processing, and the vehicle An object of the present invention is to provide a vehicle detection system and a vehicle detection method including a detection device.

  A vehicle detection apparatus according to a first aspect of the present invention is a vehicle detection apparatus that detects a vehicle blinking a predetermined indicator light based on a captured image obtained by capturing an image of a road, and is configured to capture two images at different imaging time points. A time difference image generating means for generating a time difference image obtained by a difference between pixel values of each image, and one or a plurality of pixel lines along a predetermined direction from the time difference image generated by the time difference image generating means A line image extracting means for extracting the line image to be read, a storage means for storing the line image extracted by the line image extracting means in association with an imaging time point, and a plurality of imaging time points stored in the storage means. Time-series image generation means for generating a time-series image arranged two-dimensionally in a series, and temporal changes related to characteristic pixels that are characteristic pixels of the time-series image generated by the time-series image generation means A frequency component calculating means for calculating a corresponding frequency component, characterized in that it comprises detecting means for detecting a vehicle on the basis of the frequency component calculated by the frequency component calculation unit.

  The vehicle detection apparatus according to a second aspect of the present invention is the vehicle detection apparatus according to the first aspect, wherein the time series image generated by the time series image generation unit is scanned in the direction in which the time of imaging is scanned, Fluctuation number calculating means for calculating the number of fluctuations in which the fluctuation range of the pixel value of the series image exceeds a predetermined first threshold value, and determining whether or not the fluctuation number calculated by the fluctuation number calculating means is greater than a predetermined second threshold value Determination means, and the detection means is configured to detect a stopped vehicle or a low-speed vehicle based on a determination result of the determination means.

  According to a third aspect of the present invention, the vehicle detection device according to the first or second aspect of the present invention is based on the frequency component calculated by the frequency component calculation means, and the first frequency of the first frequency band and the first frequency band. A frequency calculating means for calculating the second frequency of the second frequency band in the lower frequency range, and a ratio calculating means for calculating the ratio of the first frequency and the second frequency calculated by the frequency calculating means. The detecting means is configured to detect a stopped vehicle or a low-speed vehicle based on the ratio calculated by the ratio calculating means.

  According to a fourth aspect of the present invention, there is provided a vehicle detection device according to the third aspect, further comprising frequency band setting means for presetting a predetermined frequency band, wherein the frequency calculation means excludes the frequency band set by the frequency band setting means. It is configured to calculate the frequency.

  The vehicle detection device according to a fifth aspect of the present invention is the vehicle detection apparatus according to any one of the first to fourth aspects, further comprising: a sorting unit that sorts the captured image into a plurality of regions, wherein the time difference image generating unit is the sorting unit. The time-difference image of the segmented captured image is generated.

  A vehicle detection device according to a sixth aspect of the present invention is the vehicle detection device according to any one of the first to fifth aspects, further comprising noise removal means for removing noise from the time-series image generated by the time-series image generation means, and the frequency component The calculating means is configured to calculate the frequency component of the pixel value of the time-series image from which noise has been removed by the noise removing means.

  According to a seventh aspect of the present invention, in any one of the first to sixth aspects, the vehicle detection device includes a color component extraction unit that extracts a color component of a specific color from the captured image, and the time difference image generation unit includes: A time difference image of the color component extracted by the color component extraction means is generated.

  According to an eighth aspect of the present invention, there is provided a vehicle detection apparatus according to any one of the first to seventh aspects, further comprising an adding unit that adds pixel values at the same imaging time point of the time-series image generated by the time-series image generating unit. The frequency component calculating means is configured to calculate a frequency component corresponding to a temporal change in the pixel value obtained by adding by the adding means.

  A vehicle detection device according to a ninth aspect of the present invention is a vehicle detection device that detects a vehicle blinking a predetermined indicator light based on a captured image obtained by capturing an image of a road, and is configured to capture two images at different imaging time points. A time difference image generating means for generating a time difference image obtained by a difference between pixel values of each image, and one or a plurality of pixel lines along a predetermined direction from the time difference image generated by the time difference image generating means A line image extracting means for extracting the line image to be read, a storage means for storing the line image extracted by the line image extracting means in association with an imaging time point, and a plurality of imaging time points stored in the storage means. A time-series image generating means for generating a time-series image arranged two-dimensionally in the series, a scanning means for scanning the time-series image generated by the time-series image generating means in the direction in which the image was captured, and the running When the scanning is performed by the means, the fluctuation number calculating means for calculating the number of fluctuations in which the fluctuation range of the pixel value of the time-series image exceeds a predetermined first threshold value, and the fluctuation number calculated by the fluctuation number calculating means is a predetermined second value. It is characterized by comprising a determination means for determining whether or not it is larger than the threshold value, and a detection means for detecting a stopped vehicle or a low-speed vehicle based on the determination result of the determination means.

  A vehicle detection system according to a tenth aspect of the present invention is a vehicle detection system comprising one or more imaging devices and the vehicle detection device according to any one of the first to ninth aspects described above, wherein the vehicle detection The apparatus includes an acquisition unit that acquires a captured image obtained by imaging a road with the imaging device, and an output unit that outputs information about the vehicle detected by the detection unit to the outside.

  A vehicle detection method according to an eleventh aspect of the present invention is a vehicle detection method for detecting a vehicle blinking a predetermined indicator lamp based on a captured image obtained by capturing an image of a road, and two imagings having different imaging time points Generate a time difference image obtained by the difference between pixel values of each image, extract a line image composed of one or more pixel lines along a predetermined direction from the generated time difference image, and extract the extracted line image Temporal changes related to characteristic pixels, which are characteristic pixels of the generated time-series image, are generated in association with the imaging time point, generate a time-series image in which the stored line images at a plurality of imaging time points are two-dimensionally arranged in time series The frequency component corresponding to is calculated, and the vehicle is detected based on the calculated frequency component.

  The vehicle detection method according to a twelfth aspect of the present invention is the vehicle detection method according to the eleventh aspect, wherein the generated time-series image is scanned in the direction in which time elapses, and the fluctuation range of the pixel value of the scanned time-series image exceeds a predetermined first threshold value. The number of times is calculated, it is determined whether or not the calculated number of fluctuations is greater than a predetermined second threshold, and a stopped vehicle or a low-speed vehicle is detected based on the determination result.

  In the first invention and the eleventh invention, the vehicle detection device generates a time difference image obtained by a difference between pixel values of two captured images having different imaging time points. In the time difference image, the running vehicle has a pixel value (for example, brightness, density) change point (for example, brightness, density) on the vehicle head side and the vehicle tail side due to the difference in pixel values between the vehicle and the road (bright portion: dark to bright, dark portion: From light to dark). On the other hand, in a vehicle that is stopped by blinking (flashing) a predetermined indicator lamp (for example, a hazard lamp or a rotating lamp), a pixel value change portion appears according to the blinking state of the indicator lamp. The vehicle detection device extracts a line image composed of pixel lines along a predetermined direction (for example, the traveling direction of the vehicle) from the generated time difference image, and temporarily stores the extracted line image in association with the imaging time point. . The extracted line image is composed of one or a plurality of pixel lines, and has a characteristic portion where the pixel value changes.

  The vehicle detection device generates a time-series image in which a plurality of line images at the time of imaging are two-dimensionally arranged in time series. In the time-series image, for example, the horizontal axis indicates the time axis, and the vertical axis indicates the traveling direction of the vehicle. In the time-series image, the running vehicle appears in a belt shape in which each of the dark part and the bright part continuously moves in the traveling direction of the vehicle along the time axis. On the other hand, in the stopped vehicle, the dark part and the bright part repeatedly appear at the same position along the time axis. The vehicle detection device calculates a frequency component (a power spectrum amount with respect to a frequency) corresponding to a temporal change related to the feature pixel on the time axis of the time-series image. A feature pixel is, for example, a pixel that can be a feature point on a time-series image, such as a pixel whose pixel value exceeds (or does not exceed) a predetermined threshold value, or a pixel at a boundary where the pixel value rapidly changes. Refers to a pixel having a value. The temporal change relating to the feature pixel refers to a temporal change such as the total number of feature pixels or the total number of pixel values of the feature pixels, their standard deviation or variance, and is a numerical value related to the feature pixel. Focusing on the frequency component of the time variation of the vehicle, the behavior of the vehicle is grasped. For example, complex discrete Fourier transform can be used to calculate the frequency component. The vehicle detection device detects a vehicle with a blinking indicator lamp when there is a predetermined power spectrum in a predetermined high frequency band with the calculated frequency component.

  Thereby, it is possible to detect a vehicle blinking (flashing) with a display lamp such as a hazard lamp or a rotating lamp. In addition, there is no need to adjust the imaging conditions such as fine adjustment of the road surface distance and fine adjustment of the threshold value of the brightness value for detecting the speed of the vehicle or specifying the size of the vehicle. Adjustment costs can be reduced. In addition, since it is configured to detect blinking of the indicator light, it does not depend on characteristics such as a video camera or road conditions, and it is not necessary to limit the imaging area according to the characteristics, and the vehicle can be used in a wider area than in the past. Can be detected.

  In the second invention and the twelfth invention, the vehicle detection device scans the time-series image in the elapsed direction (time axis direction) at the time of imaging, and the fluctuation range of the pixel value of the time-series image has a predetermined first threshold value. Calculate the number of fluctuations exceeding. For example, when there is a vehicle blinking the indicator light, the bright part and the dark part, which are the pixel value changing portions, appear repeatedly, so the difference (variation width) between the pixel value of the dark part and the pixel value of the bright part By using a small first threshold value, it is possible to obtain the number of repetitions (the number of fluctuations) between the bright part and the dark part. The vehicle detection device determines whether or not the calculated number of fluctuations is greater than a predetermined second threshold, and detects a stopped vehicle or a low-speed vehicle based on the determination result. For example, a vehicle that is traveling while blinking the indicator light is moving in the traveling direction with time, so compared to a vehicle that is blinking the indicator light and is stopped at a low speed, The number of fluctuations becomes smaller.

  Thereby, it is possible not to detect the vehicle that is running by blinking the indicator light, and it is possible to detect the stopped vehicle or the low-speed vehicle with high accuracy. In addition, for example, when the night illumination in the tunnel is reflected by the traveling vehicle and the light and darkness is seen repeatedly, or in the semi-underground tunnel where part of the tunnel ceiling is open, Even when a vehicle traveling in a shaded place appears to blink, it can be distinguished from a stopped vehicle or a low-speed vehicle, and erroneous detection of a stopped vehicle can be prevented.

  In the third aspect of the invention, the vehicle detection device, based on the frequency component of the time-series image, has a first frequency (for example, a power spectrum of the first frequency band (for example, a predetermined indicator light target setting frequency)). Maximum frequency) and a second frequency (for example, an average value of a power spectrum) in a second frequency band (for example, a predetermined frequency that is not subject to indicator lights) that is lower than the first frequency band. The vehicle detection device calculates a ratio of the calculated first frequency and second frequency (for example, maximum value of power spectrum / average value of power spectrum), and detects a stopped vehicle or a low-speed vehicle based on the calculated ratio. To do. For example, when the calculated ratio is larger than a predetermined ratio threshold, a stopped vehicle or a low-speed vehicle is detected on the assumption that a high-frequency component due to blinking of the indicator lamp exists. By using the ratio, it is possible to detect a stopped vehicle or a low-speed vehicle with high accuracy without being affected by the change in the amount of power spectrum that varies according to the contrast of the captured image, and it is not necessary to adjust the imaging conditions. Thus, adjustment costs can be reduced.

  In the fourth invention, the vehicle detection device calculates the frequency by excluding a predetermined frequency band set in advance. For example, the predetermined frequency band is a frequency band due to an unnecessary blinking (flickering) object such as a gaze guidance mark or a road fence installed on the center line of a road or on a pedestrian crossing. Thereby, it is possible to accurately detect a stopped vehicle or a low-speed vehicle by eliminating the influence of an unnecessary blinking (blinking) object.

  In the fifth invention, the vehicle detection device divides the captured image into a plurality of regions (for example, for each lane of the road), and generates a time difference image of the divided captured image. A vehicle detection apparatus detects a stop vehicle or a low-speed vehicle for every divided area based on the generated time difference image. As a result, when there is a vehicle running in the adjacent lane, the stop vehicle or the low-speed vehicle is erroneously recognized as a vehicle traveling at a normal speed due to the influence of the vehicle. The situation where it cannot be detected can be prevented.

  In the sixth aspect of the invention, the vehicle detection device removes noise from the time series image and calculates the frequency component of the time series image from which the noise has been removed. For example, among pixel values of a time-series image, a pixel value (or a range of pixel values) constituting a bright part, a pixel value (or a range of pixel values) constituting a dark part, and a background part (an area where no vehicle exists) Noise (pixel value) that is unnecessary for detection of a stopped vehicle or a low-speed vehicle is performed by performing a filtering process such that a pixel value that does not belong to any of the constituent pixel values (or pixel value ranges) is set as a pixel value in the background portion. ) Is removed. Thereby, a stop vehicle or a low-speed vehicle can be detected accurately.

  In the seventh invention, the vehicle detection device extracts a color component of a specific color from the captured image and generates a time difference image of the extracted color component. Instead of extracting only the luminance signal (Y) from the image data obtained by imaging with an imaging device (video camera), for example, by extracting an RGB signal, YCbCr (CbCr is a color difference signal) signal, a specific color component is obtained. Extract. By bringing the specific color component close to the color component of the hazard lamp or the rotating lamp, it is possible to detect blinking of the indicator lamp with high accuracy and detect a stopped vehicle or a low-speed vehicle.

  In the eighth invention, the vehicle detection device adds the pixel values at the same imaging time point of the time series image, and calculates a frequency component corresponding to the temporal change of the pixel value obtained by the addition. As described above, in the time-series image, the horizontal axis indicates the time axis, and the vertical axis indicates the traveling direction of the vehicle. On a time-series image, a traveling vehicle produces the same amount of time difference at any position on the time axis. Therefore, by adding the pixel values at the same imaging time point of the time-series image, the time distribution of pixel values The presence of the traveling vehicle can be deleted. Thereby, the influence of the traveling vehicle can be eliminated, the blinking of the indicator light can be detected with high accuracy, and the stopped vehicle or the low-speed vehicle can be detected.

  In the ninth aspect of the invention, the vehicle detection device generates a time difference image obtained by the difference between the pixel values of two captured images at different imaging time points. In the time difference image, the running vehicle has a pixel value (for example, brightness, density) change point (for example, brightness, density) on the vehicle head side and the vehicle tail side due to the difference in pixel values between the vehicle and the road (bright portion: dark to bright, dark portion: From light to dark). On the other hand, in a vehicle that is stopped by blinking (flashing) a predetermined indicator lamp (for example, a hazard lamp or a rotating lamp), a pixel value change portion appears according to the blinking state of the indicator lamp. The vehicle detection device extracts a line image composed of pixel lines along a predetermined direction (for example, the traveling direction of the vehicle) from the generated time difference image, and temporarily stores the extracted line image in association with the imaging time point. . The extracted line image is composed of one or a plurality of pixel lines, and has a characteristic portion where the pixel value changes.

  The vehicle detection device generates a time-series image in which a plurality of line images at the time of imaging are two-dimensionally arranged in time series. In the time-series image, for example, the horizontal axis indicates the time axis, and the vertical axis indicates the traveling direction of the vehicle. In the time-series image, the running vehicle appears in a belt shape in which each of the dark part and the bright part continuously moves in the traveling direction of the vehicle along the time axis. On the other hand, in the stopped vehicle, the dark part and the bright part repeatedly appear at the same position along the time axis. The vehicle detection device scans the time series image in the elapsed direction (time axis direction) at the time of imaging, and calculates the number of fluctuations in which the fluctuation range of the pixel value of the time series image exceeds a predetermined first threshold. For example, when there is a vehicle blinking the indicator light, the bright part and the dark part, which are the pixel value changing portions, appear repeatedly, so the difference (variation width) between the pixel value of the dark part and the pixel value of the bright part By using a small first threshold value, it is possible to obtain the number of repetitions (the number of fluctuations) between the bright part and the dark part. The vehicle detection device determines whether or not the calculated number of fluctuations is greater than a predetermined second threshold, and detects a stopped vehicle or a low-speed vehicle based on the determination result. For example, a vehicle that is traveling while blinking the indicator light is moving in the traveling direction with time, so compared to a vehicle that is blinking the indicator light and is stopped at a low speed, The number of fluctuations becomes smaller.

  Thereby, it is possible to detect a vehicle blinking (flashing) with a display lamp such as a hazard lamp or a rotating lamp. In addition, there is no need to adjust the imaging conditions such as fine adjustment of the road surface distance and fine adjustment of the threshold value of the brightness value for detecting the speed of the vehicle or specifying the size of the vehicle. Adjustment costs can be reduced. In addition, since it is configured to detect blinking of the indicator light, it does not depend on characteristics such as a video camera or road conditions, and it is not necessary to limit the imaging area according to the characteristics, and the vehicle can be used in a wider area than in the past. Can be detected. Further, it is possible not to detect the vehicle that is running by blinking the indicator light, and it is possible to detect the stopped vehicle or the low-speed vehicle with high accuracy. In addition, for example, when the night illumination in the tunnel is reflected by the traveling vehicle and the light and darkness is seen repeatedly, or in the semi-underground tunnel where part of the tunnel ceiling is open, Even when a vehicle traveling in a shaded place appears to blink, it can be distinguished from a stopped vehicle or a low-speed vehicle, and erroneous detection of a stopped vehicle can be prevented.

  In the tenth aspect of the invention, the vehicle detection device acquires a captured image obtained by imaging a road with one or a plurality of imaging devices, and blinks a predetermined display lamp (blinks) based on the acquired captured image. ) Detecting a stopped vehicle or a low-speed vehicle, and outputting information on the detected stopped vehicle or low-speed vehicle to the outside (for example, a traffic control device, a traffic control device, etc.). As a result, the person in charge of operation of the external device can acquire information on the stopped vehicle or the low-speed vehicle stopped by blinking the hazard lamp or the revolving light in real time, and an accident or failure that occurs suddenly. It is possible to respond quickly to these events.

  In the present invention, it is possible to detect a vehicle blinking (flashing) with a display lamp such as a hazard lamp or a rotating lamp. In addition, there is no need to adjust the imaging conditions such as fine adjustment of the road surface distance and fine adjustment of the threshold value of the brightness value for detecting the speed of the vehicle or specifying the size of the vehicle. Adjustment costs can be reduced. In addition, since it is configured to detect blinking of the indicator lamp, it does not depend on characteristics such as video cameras or road conditions, and it is not necessary to limit the imaging area according to those characteristics, and a wider area than conventional It becomes possible to detect the vehicle.

  Hereinafter, the present invention will be described with reference to the drawings illustrating embodiments. FIG. 1 is a block diagram showing an example of the configuration of a vehicle detection system according to the present invention. The vehicle detection system according to the present invention includes a stopped vehicle detection device 100, a video camera 200, and the like. The video camera 200 is connected to the stopped vehicle detection device 100 and outputs image data obtained by imaging with the video camera 200 to the stopped vehicle detection device 100. Note that the number of video cameras 200 is not limited to one, and a plurality of video cameras 200 may be provided.

  FIG. 2 is an explanatory diagram illustrating an example of the imaging region M of the video camera 200. As shown in FIG. 2, the video camera 200 is installed on the roadside so as to image a predetermined imaging region M (for example, about 200 m along the road direction) of a road (for example, a road in a tunnel). In the example of FIG. 2, the video camera 200 captures a road on a road 1, a lane 1, a lane 2, and two lanes on one side of the road shoulder 2. A plurality of vehicles can be installed with a separation distance of about 200 m so as to image the road in the tunnel.

  The stop vehicle detection device 100 includes a control unit 10 that controls each unit in the device, an interface unit 11, a detection region setting unit 12, a time difference image generation unit 13, a line image extraction unit 14, an image memory 15, and a spatiotemporal image generation unit. 16, the image noise removal part 17, the spatio-temporal luminance waveform generation part 18, the Fourier-transform part 19, the frequency band mask part 20, the stop vehicle determination part 21, the output part 22, etc. are provided.

  The interface unit 11 acquires image data output from the video camera 200, extracts a Y signal (luminance signal) from the acquired image data (for example, RGB signal), and outputs it. Further, the interface unit 11 extracts a Y signal (luminance signal) and a CbCr signal (color difference signal) from the acquired image data, and outputs specific color components (for example, the color of a hazard lamp, the color of a rotating lamp, or these colors) Output a signal of a combination of colors, etc.). The controller 10 can control whether to output a luminance signal or a signal of a specific color component. By bringing the specific color component close to the color component of the hazard lamp or the rotating lamp, it is possible to detect blinking of the indicator lamp with high accuracy and detect a stopped vehicle. The interface unit 11 stores image data in units of frames in the image memory 15 based on the output signal.

  The detection area setting unit 12 extracts pixels (luminance, density, specific color components, etc.) corresponding to a plurality of points in the imaging area M from image data obtained by imaging the imaging area M, and has a predetermined size. To a captured image composed of pixel blocks. Further, the detection area setting unit 12 divides the captured image obtained by mapping into a plurality of detection areas.

  FIG. 3 is an explanatory diagram illustrating an example of a detection region of a captured image. As shown in FIG. 3, the imaging region M is mapped to a captured image configured by, for example, a pixel block of 100 pixels × 60 pixels. That is, 6000 pixels are selected from the image of the imaging region M, and each selected pixel is mapped to a pixel block. The conversion from the image of the imaging region M to the pixel block is not limited to this, and projective conversion or the like may be used. The captured image is divided into, for example, four detection areas X1, X2, X3, and X4 (the captured image is divided into four). In the case of FIG. 1, the detection areas X1 to X4 correspond to the road shoulder 1, the lane 1, the lane 2, and the road shoulder 2 of the imaging area M.

  A vehicle that stops due to an accident or a failure is near the road, and a subsequent vehicle that travels from behind may pass by the side of the stopped vehicle. Therefore, the width of the detection areas X1 and X4 is set to the detection area X2, Set smaller than the width of X3. Moreover, in order to prevent the detection failure of a stop vehicle, each detection area can also set the overlapping area which overlaps with the adjacent detection area, respectively.

  By dividing the captured image and performing each process described below for each of the divided captured images, when there is a vehicle running in the adjacent lane, the vehicle is affected by the vehicle and is traveling on the stopped vehicle. It is possible to prevent a situation in which a stopped vehicle cannot be detected by misrecognizing it as a vehicle.

  The time difference image generation unit 13 generates a time difference image for each detection region X1 to X4 by obtaining a difference between pixel values (for example, luminance values) of two divided captured images at different imaging time points.

  FIG. 4 is an explanatory diagram showing an example of a time difference image. The example of FIG. 4 shows an example in the case where there is a traveling vehicle on the captured image. As shown in FIG. 4, it is assumed that, for example, a white vehicle C1 and a black vehicle C2 are traveling in the captured images at the imaging times t and t-1. In the luminance distribution on the captured image, the luminance value of the portion of the white vehicle C1 is large (for example, 156), and the luminance value of the portion of the black vehicle C2 is small (for example, 0). Further, the luminance value of the road surface is a luminance value (for example, 100) that is about the middle of them.

  When the difference between the luminance value of each pixel of the captured image at time t and the luminance value of each pixel of the captured image at time t−1 is calculated, the luminance value of the white vehicle C1 changes from 100 to 156 ( Bright part: dark to bright, plus difference), and the rear side of the white vehicle C1 changes from 156 to 100 (dark part: light to dark, minus difference).

  On the other hand, when the difference between the luminance value of each pixel of the captured image at time t and the luminance value of each pixel of the captured image at time t−1 is calculated, the luminance value of the black vehicle C2 is changed from 100 to 0. It changes (dark part: difference from light to dark, minus), and the black vehicle C2 rear side changes from 0 to 100 (bright part: difference from dark to light, plus).

  In this way, a time difference image at time t can be obtained, and in the time difference image, luminance value changing portions appear on the front side and the rear side of the traveling vehicle.

  FIG. 5 is an explanatory diagram showing another example of the time difference image. The example of FIG. 5 shows an example in the case where there is a stopped vehicle blinking (flashing) the hazard lamp on the captured image. As shown in FIG. 5, it is assumed that the hazard lamp is turned off in the captured image at time t, and the hazard lamp is turned on in the captured image at time t-1. When the hazard lamp is on, the luminance value is the largest (for example, 255), and when the hazard lamp is off, the luminance value is small (for example, 0).

  When the difference between the luminance value of each pixel of the captured image at time t and the luminance value of each pixel of the captured image at time t−1 is obtained, the luminance value near the hazard lamp of the stopped vehicle changes from 255 to 0 (dark part) : Light to dark, minus difference). Conversely, if the hazard lamp is turned on in the captured image at time t and the hazard lamp is off in the captured image at time t−1, the luminance value of each pixel of the captured image at time t, When the difference from the luminance value of each pixel of the captured image at time t-1 is obtained, the luminance value near the hazard lamp of the stopped vehicle changes from 0 to 255 (bright portion: dark to bright, plus difference). Become.

  In this way, a time difference image at time t can be obtained. In the time difference image, a luminance value change point appears in the vehicle (or in the vicinity thereof) stopped by blinking the hazard lamp.

  The line image extracting unit 14 extracts a line image by extracting a characteristic luminance vertically in a row (one pixel line) from the time difference image generated by the time difference image generating unit 13, and uses the extracted line image as an imaging time. The image memory 15 stores the associated information.

  FIG. 6 is an explanatory diagram showing an example of line image extraction. As shown in FIG. 6, the luminance extracted from the time difference image at time t is the highest luminance from the luminance value of the same pixel at time t-1 in one horizontal line when the time difference image is scanned in the horizontal direction. Extract those with a difference in values. By extracting the characteristic luminance for each horizontal line, a line image composed of one pixel line along the vertical direction (vehicle traveling direction) of the time difference image is extracted.

  Note that pixels having characteristic luminance extracted for each line in the horizontal direction are not limited to pixels on the same line in the vertical direction (vertical direction) of the time difference image, and for example, as in the example of FIG. When the time difference image has a line for 20 pixels in the vertical direction, it may be extracted from any position of the 20 pixel line. As a result, even when the traveling vehicle travels slightly in the width direction of the road within the detection region, the characteristic portion can be reliably extracted.

  In the extracted line image, characteristic portions such as characteristic luminance of the traveling vehicle (that is, luminance change on the vehicle head side and vehicle tail side) and luminance change of the stopped vehicle stopped by blinking the hazard lamp appear. . In the example of FIG. 6, the line image is composed of one pixel line, but may be composed of a plurality of pixel lines.

  The spatiotemporal image generation unit 16 generates a spatiotemporal image (time series image) in which the line images extracted by the line image extraction unit 14 and temporarily stored in the image memory 15 are arranged in time series.

  FIG. 7 is an explanatory diagram showing an example of a spatiotemporal image generation method. As shown in FIG. 7, 64 line images from time t to time t-63 stored in the image memory 15 are taken out and arranged in a two-dimensional manner in a time-series manner, thereby obtaining a time-space image (time-series image). Image). In the spatiotemporal image, the horizontal axis indicates the time axis, and the vertical axis indicates the traveling direction of the vehicle. Note that the number of line images to be extracted is not limited to 64, and may be another number, for example, 2 to the Nth power. Further, the line image is not limited to one composed of one pixel line, and may be composed of a plurality of pixel lines. Also in this case, the spatio-temporal image is generated by arranging line images composed of a plurality of pixel lines adjacent to each other and two-dimensionally arranging them in the time axis direction.

  As shown in FIG. 7, in the generated spatio-temporal image, the characteristic portions (luminance dark portion and light portion) of the running vehicle are continuously moved along the time axis in the traveling direction of the vehicle. Appears on On the other hand, the characteristic portions (luminance dark portion and light portion) of the stopped vehicle, which is stopped by blinking the indicator lamp such as the hazard lamp, repeatedly appear at the same position along the time axis. As described above, in the spatiotemporal image, the appearance of the characteristic portion of the brightness differs between the running vehicle and the stopped vehicle that has stopped blinking the indicator light.

  The luminance characteristic portion of the spatio-temporal image, that is, the characteristic pixel on the time axis of the spatio-temporal image (time-series image), for example, has a pixel value that exceeds (or does not exceed) a predetermined threshold value or a pixel value that suddenly increases. A pixel having a pixel value that can be a feature point on a spatiotemporal image, such as a pixel at a fluctuating boundary. The temporal change relating to the feature pixel refers to a temporal change such as the total number of feature pixels or the total number of pixel values of the feature pixels, their standard deviation or variance, and is a numerical value related to the feature pixel. The behavior of the vehicle can be grasped by paying attention to the frequency component of the time variation of the vehicle.

  The image noise removing unit 17 removes image noise indicating an unnecessary luminance value from the space-time image generated by the space-time image generating unit 16. Specifically, the image noise removing unit 17 includes, among the luminance values of the spatiotemporal image, a luminance value (or a range of luminance values) that constitutes a bright part that is a characteristic part, and a luminance value (or luminance value that constitutes a dark part). And a luminance value that does not belong to any of the luminance values (or luminance value ranges) constituting the background portion (region where the vehicle does not exist) that is not a characteristic portion of the luminance value as a luminance value of the background portion By performing the processing, noise (luminance value) unnecessary for detection of the stopped vehicle is removed. For example, assuming that the luminance value of the dark portion is 0 to 50, the luminance value of the bright portion is 150 to 255, and the luminance value of the background portion is 100, the luminance values of the pixels having the luminance values of 80, 120, and the like are set to 100. Thereby, a stop vehicle can be detected accurately.

  Further, the image noise removing unit 17 pays attention only to the temporal change in luminance value of the spatiotemporal image generated by the spatiotemporal image generation unit 16, and for example, when the luminance value changes only slightly. Assuming that there is no change in the luminance value, noise due to a slight change can be removed. Specifically, for the same pixel of the spatiotemporal image generated by the spatiotemporal image generation unit 16, when the previous luminance value is 100 and the current luminance value is 102, the luminance value of the pixel is set to 100. If the previous luminance value is 100 and the current luminance value is 120, the luminance value of the pixel is set to 120. That is, when the change in the luminance value of the pixel exceeds a predetermined threshold value (for example, 15), the current luminance value is used as it is. When the threshold value is not exceeded, the previous luminance value is set as the current luminance value. replace.

  The spatiotemporal luminance waveform generator 18 adds the luminance values at the same imaging time of the spatiotemporal image from which unnecessary noise has been removed to generate a spatiotemporal luminance waveform (time distribution waveform).

  FIG. 8 is an explanatory diagram showing an example of a method for generating a spatiotemporal luminance waveform. As shown in FIG. 8, in the spatiotemporal image, as described above, the horizontal axis indicates the time axis, and the vertical axis indicates the traveling direction of the vehicle. In a spatio-temporal image, a traveling vehicle produces the same amount of time difference at any position on the time axis, so the luminance value of each pixel at the same imaging time in the spatio-temporal image is added along the vertical direction. Thus, the presence of the traveling vehicle can be eliminated from the time distribution of the luminance value. In the generated spatio-temporal luminance waveform, only the change in the luminance of a stopped vehicle that has stopped by blinking an indicator lamp such as a hazard lamp will appear. It is possible to detect blinking of the indicator light and detect a stopped vehicle.

  The Fourier transform unit 19 performs a complex discrete Fourier transform process on the time distribution of the spatio-temporal luminance waveform (time distribution waveform) generated by the spatio-temporal luminance waveform generation unit 18 to obtain a power spectrum (frequency component of the spatio-temporal luminance waveform). Ask for.

  FIG. 9 is an explanatory diagram showing an example of a power spectrum of a spatiotemporal luminance waveform. In FIG. 9, the horizontal axis indicates the frequency, and the vertical axis indicates the amount of power spectrum (frequency). A frequency band for determining the presence or absence of the predetermined indicator lamp can be determined in advance according to the blinking (flashing) cycle of the predetermined indicator lamp such as a hazard lamp or a rotating lamp. For example, by setting a hazard frequency setting frequency band (first frequency band) on the relatively high frequency side and a hazard frequency setting frequency band (second frequency band) on the relatively low frequency side, the spatio-temporal luminance Depending on how much power spectrum amount appears in which frequency band of the power spectrum of the waveform, it can be seen that there is a high-frequency component due to blinking of the indicator lamp, and it becomes possible to detect a stopped vehicle.

  The frequency band mask unit 20 masks (excludes) a power spectrum in a predetermined frequency band (mask setting frequency band) from the power spectrum shown in FIG. The predetermined frequency band is a frequency band caused by unnecessary blinking (flickering) objects such as a line-of-sight guide mark or a road fence installed on the center line of a road or on a pedestrian crossing. As a result, it is possible to accurately detect a stopped vehicle while eliminating the influence of unnecessary blinking (blinking) objects.

  The stopped vehicle determination unit 21 determines the presence or absence of a stopped vehicle based on the power spectrum (frequency component) converted by the Fourier transform unit 19. Specifically, the maximum value of the power spectrum in the hazard target setting frequency band is obtained, and the average value of the power spectrum in the hazard target setting frequency band is calculated. The stopped vehicle determination unit 21 calculates a ratio between the maximum value of the power spectrum and the average value of the power spectrum (maximum value of the power spectrum / average value of the power spectrum), and the calculated ratio is a predetermined ratio threshold (for example, 2 If it is larger, it is determined that there is a stopped vehicle because there is a high frequency component due to blinking of the hazard lamp. In addition, when the calculated ratio is smaller than a predetermined ratio threshold (for example, 2), it is determined that there is no stopped vehicle. The condition for determination based on the power spectrum ratio is, for example, a frequency determination condition.

  By using the ratio, it is possible to detect a stopped vehicle with high accuracy without being affected by the change in the amount of power spectrum that fluctuates according to the contrast of the captured image, and it is not necessary to adjust the imaging conditions. Cost can be reduced.

  The stopped vehicle determination unit 21 scans the spatiotemporal image from which unnecessary noise has been removed in the time axis direction, calculates the number of fluctuations in which the fluctuation range of the luminance value of the spatiotemporal image exceeds a predetermined luminance threshold, and calculates the calculated fluctuation. By determining whether or not the number of times is greater than a predetermined number of times threshold, for example, a traveling vehicle that is traveling by blinking a hazard lamp is excluded, and a stopped vehicle that is stopped by blinking an indicator light or at a low speed It is determined whether or not there is a low-speed vehicle that is traveling (for example, a vehicle that is traveling slowly, a vehicle that moves and repeatedly stops). The condition determined by the number of fluctuations is, for example, a condition for determining the number of fluctuations.

  FIG. 10 is an explanatory diagram showing an example of the variation count determination condition. As shown in FIG. 10, when there is a vehicle blinking an indicator lamp such as a hazard lamp, a bright part and a dark part, which are characteristic parts (change points) of the brightness value, appear repeatedly. By using a luminance threshold smaller than the difference (brightness value fluctuation range) with the luminance value of the bright part, the number of repetitions (fluctuation number) of the bright part and the dark part can be obtained. Since the vehicle (moving blinking vehicle) that is traveling while blinking the indicator light is moving in the traveling direction over time, the number of fluctuations of the traveling vehicle is the number of fluctuations of the stopped vehicle (stop hazard vehicle) or the low-speed vehicle. Less.

  Therefore, it is determined whether or not the calculated number of fluctuations is larger than a predetermined number threshold (for example, 15) or the like, and the traveling vehicle that is traveling by blinking the hazard lamp is excluded based on the determination result, and the indicator lamp is turned off. It is determined whether or not there is a stopped vehicle or a low-speed vehicle that is blinking and stopped. For example, when the number of fluctuations is larger than the number threshold, it is determined that there is a vehicle that is stopped by blinking the indicator light. It can be determined that the vehicle is a low-speed vehicle or a low-speed vehicle.

  Thereby, it is possible not to detect the vehicle that is running by blinking the indicator light, and it is possible to detect the stopped vehicle or the low-speed vehicle with high accuracy. In addition, for example, when the night illumination in the tunnel is reflected by the traveling vehicle and the light and darkness is seen repeatedly, or in the semi-underground tunnel where part of the tunnel ceiling is open, Even when a vehicle running in a place where the shade continues appears to blink, it can be distinguished from a stopped vehicle or a low-speed vehicle, and erroneous detection of the stopped vehicle or the low-speed vehicle can be prevented.

  When both the frequency determination condition and the variation count determination condition described above are satisfied, the stopped vehicle determination unit 21 stops blinking a display lamp such as a hazard lamp when both conditions continue for 10 seconds, for example. It is determined that there is a stopped vehicle or a low-speed vehicle, and a stopped vehicle or a low-speed vehicle is detected. Note that the duration for determination is not limited to 10 seconds. In addition, when determining and detecting a stopped vehicle or a low-speed vehicle, it is not necessary to add a variation count determination condition.

  The output unit 22 outputs information related to the stopped vehicle or the low-speed vehicle determined and detected by the stopped vehicle determination unit 21 to an external device such as a traffic control device or a traffic control device. As a result, the person in charge of operation of the external device can acquire information on the stopped vehicle or the low-speed vehicle stopped by blinking the hazard lamp or the revolving light in real time, and an accident or failure that occurs suddenly. It is possible to respond quickly to these events.

  Next, the operation of the stopped vehicle detection device 100 will be described. 11 and 12 are flowcharts showing the processing procedure of the stopped vehicle detection device 100. FIG. The control unit 10 acquires image data (S11), and performs pixel mapping on a pixel block having a predetermined size (S12). The control unit 10 classifies the captured image that has been subjected to pixel mapping and sets a detection region (S13).

  The control unit 10 selects one detection area from the set detection areas to generate a time difference image (S14), and extracts a line image having a luminance value feature from the generated time difference image (S15). The control unit 10 stores the extracted line image in the image memory 15 in association with the imaging time.

  The control unit 10 generates a spatiotemporal image by two-dimensionally arranging a predetermined number (for example, 64) of line images stored in the image memory 15 in time series (S16), and generates a spatiotemporal image from the generated spatiotemporal image. Image noise having a necessary luminance value is removed (S17).

  The control unit 10 calculates the number of luminance fluctuations in the temporal direction (time axis direction) of the spatiotemporal image from which image noise has been removed (S18), and determines whether the number of fluctuations is greater than the number threshold (S19). When the number of changes is greater than the number of times threshold (YES in S19), the control unit 10 satisfies the number of changes determination condition (S20), and when the number of changes is not greater than the number of times threshold (NO in S19), the number of changes is not satisfied. (S21).

  The control unit 10 generates a spatiotemporal luminance waveform by adding the luminance values at the same imaging time of the spatiotemporal image (S22), and performs Fourier transform on the generated spatiotemporal luminance waveform (S23). The control unit 10 calculates the maximum value of the power spectrum in the hazard target setting frequency band based on the power spectrum (frequency component) after Fourier transform (S24), and calculates the average value of the power spectrum in the non-hazard target setting frequency band. Calculate (S25).

  The control unit 10 determines whether the ratio between the maximum value of the power spectrum and the average value of the power spectrum is larger than the ratio threshold (S26). If the ratio is larger than the ratio threshold (YES in S26), the frequency determination condition If it is satisfied (S27) and the ratio is not larger than the ratio threshold (NO in S26), the frequency determination condition is not satisfied (S28).

  The control unit 10 determines whether or not the processing for all the detection areas has been completed (S29). If the processing has not been completed (NO in S29), the processing after step S14 is continued. When the processing of all the detection areas is completed (YES in S29), the control unit 10 determines whether or not both conditions (that is, the variation count determination condition and the frequency determination condition) are satisfied in any of the detection areas. (S30).

  When both conditions are satisfied (YES in S30), the control unit 10 determines whether the satisfaction of both conditions continues for a predetermined time (for example, 10 seconds) (S31), and continues for the predetermined time. If it is present (YES in S31), a notification that there is a hazard-stopped vehicle is issued to an external device (S32). In this case, low speed vehicles can also be included.

  The control unit 10 determines whether or not there is an instruction to end the process (S33), and if there is no instruction to end the process (NO in S33), the process after step S11 is continued. Further, when both conditions are not satisfied (NO in S30), or when the predetermined time has not been continued (NO in S31), the control unit 10 continues the processing from step S11. When there is an instruction to end the process (YES in S33), the control unit 10 ends the process.

  As described above, in the present invention, it is possible to detect a stopped vehicle or a low-speed vehicle that is stopped by blinking (flashing) a display lamp such as a hazard lamp or a rotating lamp. In addition, there is no need to adjust the imaging conditions such as fine adjustment of the road surface distance and fine adjustment of the threshold value of the brightness value for detecting the speed of the vehicle or specifying the size of the vehicle. Adjustment costs can be reduced. In addition, because it is configured to detect blinking of the indicator lamp, it does not depend on characteristics such as video cameras or road conditions, and it is not necessary to limit the imaging area according to the characteristics, and it stops in a wider area than before It becomes possible to detect a vehicle or a low-speed vehicle.

  Further, in a sudden event detection system that detects a sudden event such as an accident or a vehicle failure, a stopped vehicle may be classified as a sudden event. However, in city traffic, vehicles stop due to traffic jams or signal stops on a daily basis. For such a stopped vehicle, the person in charge of system operation does not need to take any special measures. However, in the case of a stopped vehicle such as an accident, a vehicle failure, or a shoulder stop, it is not an ordinary event but an event that requires some measures to be taken. In such a case, it is assumed that the stopped vehicle blinks (flashes) a display lamp such as a hazard lamp or a rotating lamp. Therefore, by applying the present invention to such a case, the person in charge of the sudden event detection system can easily distinguish and judge a daily event from an event that requires some countermeasures. It is possible to greatly reduce the labor of the operation staff.

  Also, technical development relating to in-vehicle devices has been actively conducted, and a variety of in-vehicle device functions with high functionality are being proposed. For example, an in-vehicle system for preventing a vehicle collision, a system for detecting a driver's sleep or awake state, and the like are being put into practical use. In such a system, there is a form in which a hazard lamp or the like is used as a notification means from the vehicle to the external device, and by applying the present invention to such a system, an abnormal situation can be notified more easily. Become.

  In the above-described embodiment, the stop vehicle is not detected by obtaining the temporal change of the vehicle in the imaging region, but only the detection of blinking (flashing) of a display lamp such as a hazard lamp and the exclusion of the traveling vehicle. There is one feature that is not found in the past.

  In the above-described embodiment, the imaging region is configured to be divided into four detection regions. However, the number of divisions is not limited to four, and may be other numbers, for example, 2, 3 or the like. Moreover, the structure which does not divide a detection area | region may be sufficient. In this case, the load of the subsequent Fourier transform process can be reduced. Whether or not to divide the imaging region can be configured so as to be appropriately set according to the road condition.

  In the above-described embodiment, when two or more line images at the time of imaging are arranged two-dimensionally in time series, line images at successive imaging times may be arranged, or every other imaging time is thinned out. Alternatively, every two line images may be arranged.

  In the above-described embodiment, the frequency determination condition is an example, and the present invention is not limited to this. For example, by performing a pattern matching process on the waveform of the power spectrum (frequency distribution shape) obtained by Fourier transform of the spatiotemporal luminance waveform, it is possible to detect a stopped vehicle whose indicator lamp is blinking. . Further, instead of using the average value of the power spectrum of the set frequency band not subject to the hazard, it is possible to detect a stopped vehicle whose indicator lamp is blinking by normalizing the frequency component and using the normalized spectrum value.

  In the above embodiment, the maximum value of the power spectrum is obtained in the hazard target set frequency band, and the average value of the power spectrum is calculated in the non-hazard target set frequency band. However, the present invention is not limited to this. . For example, other statistical values such as the average value of the power spectrum in the hazard target frequency band or an evaluation value indicating the frequency may be used, and other statistical values or evaluation values indicating the frequency in the non-hazard target frequency band Can also be used.

  In the above-described embodiment, the stop vehicle or the low-speed vehicle is detected by performing the processing shown in the example of FIG. 10 without performing frequency analysis on the spatio-temporal image generated by the spatio-temporal image generation unit 16. You can also.

  In the above-described embodiment, the hazard lamp is described as an example of the predetermined indicator lamp. However, the predetermined indicator lamp is not limited to this, and an emergency situation such as a rotating lamp or a warning lamp or Any device may be used as long as it blinks (blinks) when the vehicle stops in an emergency or the like to notify such a situation.

  In the above-described embodiment, the vehicle includes not only a four-wheeled vehicle such as a normal passenger car, a bus, and a truck, but also a two-wheeled vehicle such as a motorcycle.

  The disclosed embodiments are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

It is a block diagram which shows an example of a structure of the vehicle detection system which concerns on this invention. It is explanatory drawing which shows an example of the imaging region of a video camera. It is explanatory drawing which shows an example of the detection area | region of a captured image. It is explanatory drawing which shows an example of a time difference image. It is explanatory drawing which shows the other example of a time difference image. It is explanatory drawing which shows the example of extraction of a line image. It is explanatory drawing which shows an example of the production | generation method of a spatiotemporal image. It is explanatory drawing which shows an example of the production | generation method of a spatiotemporal luminance waveform. It is explanatory drawing which shows an example of the power spectrum of a spatiotemporal luminance waveform. It is explanatory drawing which shows an example of a fluctuation frequency determination condition. It is a flowchart which shows the process sequence of a stop vehicle detection apparatus. It is a flowchart which shows the process sequence of a stop vehicle detection apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Stop vehicle detection apparatus 10 Control part 11 Interface part 12 Detection area setting part 13 Time difference image generation part 14 Line image extraction part 15 Image memory 16 Spatio-temporal image generation part 17 Image noise removal part 18 Spatio-temporal luminance waveform generation part 19 Fourier Conversion unit 20 Frequency band mask unit 21 Stopped vehicle determination unit 22 Output unit 200 Video camera

Claims (12)

A vehicle detection device that detects a vehicle blinking a predetermined indicator light based on a captured image obtained by imaging a road,
A time difference image generating means for generating a time difference image obtained by a difference between pixel values of two captured images having different imaging time points;
Line image extraction means for extracting a line image composed of one or a plurality of pixel lines along a predetermined direction from the time difference image generated by the time difference image generation means;
Storage means for storing the line image extracted by the line image extraction means in association with an imaging time point;
Time-series image generating means for generating a time-series image in which line images at a plurality of imaging points stored in the storage means are two-dimensionally arranged in time series;
Frequency component calculating means for calculating a frequency component corresponding to a temporal change related to a characteristic pixel that is a characteristic pixel of the time series image generated by the time series image generating means;
A vehicle detection device comprising: detection means for detecting a vehicle based on the frequency component calculated by the frequency component calculation means. Scanning means for scanning the time-series image generated by the time-series image generation means in the direction in which the image was captured;
A fluctuation number calculating means for calculating the number of fluctuations when the fluctuation range of the pixel value of the time-series image exceeds a predetermined first threshold when scanned by the scanning means;
Determining means for determining whether or not the number of fluctuations calculated by the number of fluctuations calculating means is greater than a predetermined second threshold;
The detection means includes
The vehicle detection device according to claim 1, wherein the vehicle detection device is configured to detect a stopped vehicle or a low-speed vehicle based on a determination result of the determination unit. Based on the frequency component calculated by the frequency component calculating means, the first frequency in the first frequency band and the second frequency in the second frequency band lower than the first frequency band are calculated. A frequency calculation means;
A ratio calculation means for calculating a ratio of the first frequency and the second frequency calculated by the frequency calculation means;
The detection means includes
The vehicle detection device according to claim 1, wherein the vehicle detection device is configured to detect a stopped vehicle or a low-speed vehicle based on the ratio calculated by the ratio calculation means. A frequency band setting means for setting a predetermined frequency band in advance,
The frequency calculation means includes
The vehicle detection device according to claim 3, wherein the frequency is calculated by excluding the frequency band set by the frequency band setting means. A sorting unit that divides the captured image into a plurality of regions,
The time difference image generation means includes:
The vehicle detection device according to any one of claims 1 to 4, wherein the vehicle detection device is configured to generate a time difference image of the captured image divided by the classification unit. Noise removing means for removing noise of the time series image generated by the time series image generating means,
The frequency component calculating means includes
The vehicle detection according to any one of claims 1 to 5, wherein a frequency component of a pixel value of a time-series image from which noise has been removed by the noise removing unit is calculated. apparatus. Color component extraction means for extracting a color component of a specific color from the captured image,
The time difference image generation means includes:
The vehicle detection device according to any one of claims 1 to 6, wherein a time difference image of the color component extracted by the color component extraction unit is generated. Adding means for adding pixel values at the same imaging time point of the time series image generated by the time series image generation means;
The frequency component calculating means includes
The frequency component corresponding to the temporal change of the pixel value obtained by the addition by the adding unit is configured to be calculated. 8. Vehicle detection device. A vehicle detection device that detects a vehicle blinking a predetermined indicator light based on a captured image obtained by imaging a road,
A time difference image generating means for generating a time difference image obtained by a difference between pixel values of two captured images having different imaging time points;
Line image extraction means for extracting a line image composed of one or a plurality of pixel lines along a predetermined direction from the time difference image generated by the time difference image generation means;
Storage means for storing the line image extracted by the line image extraction means in association with an imaging time point;
Time-series image generation means for generating a time-series image in which line images at a plurality of imaging points stored in the storage means are two-dimensionally arranged in time series;
Scanning means for scanning the time-series image generated by the time-series image generation means in the direction in which the image was captured;
A fluctuation number calculating means for calculating the number of fluctuations when the fluctuation range of the pixel value of the time-series image exceeds a predetermined first threshold when scanned by the scanning means;
Determining means for determining whether or not the number of changes calculated by the number of changes calculation means is greater than a predetermined second threshold;
A vehicle detection apparatus comprising: a detection unit that detects a stopped vehicle or a low-speed vehicle based on a determination result of the determination unit. A vehicle detection system comprising one or a plurality of imaging devices and the vehicle detection device according to any one of claims 1 to 9,
The vehicle detection device includes:
Acquisition means for acquiring a captured image obtained by imaging a road with the imaging device;
Output means for outputting information relating to the vehicle detected by the detection means to the outside. A vehicle detection system comprising: A vehicle detection method for detecting a vehicle blinking a predetermined indicator light based on a captured image obtained by imaging a road,
Generating a time difference image obtained by the difference between the pixel values of two captured images having different imaging time points;
Extracting a line image composed of one or a plurality of pixel lines along a predetermined direction from the generated time difference image;
Store the extracted line image in association with the time of imaging,
Generate a time-series image in which the stored line images at a plurality of imaging points are arranged two-dimensionally in time series,
Calculating a frequency component corresponding to a temporal change with respect to a characteristic pixel that is a characteristic pixel of the generated time-series image;
A vehicle detection method for detecting a vehicle based on a calculated frequency component. Scan the generated time-series image in the direction that time passes,
Calculating the number of fluctuations in which the fluctuation range of the pixel value of the scanned time-series image exceeds a predetermined first threshold;
Determine whether the calculated number of changes is greater than a predetermined second threshold;
The vehicle detection method according to claim 11, wherein a stopped vehicle or a low-speed vehicle is detected based on the determination result.

Images