JP6139088B2 - Vehicle detection device - Google Patents

Description

  Embodiments described herein relate generally to a vehicle detection device that detects a vehicle entering and exiting a specific area on a road.

  In a vehicle detection device used in a toll collection system, generally, the passage of a vehicle is detected by a transmission-type pole sensor using infrared rays. In other words, a pair of pole sensors are installed on both sides of the driving lane, detect each other by irradiating each other with an infrared beam, and detect the moment when the vehicle crosses the infrared beam. can do. However, since the pole sensor requires excavation work at the time of installation, and a measuring instrument and work for adjusting the installation position are also required separately, construction and adjustment costs are required.

Japanese Patent No. 4123138 JP-A-11-120480

  As described above, there is a demand for the vehicle detection device to suppress construction / adjustment costs as much as possible.

  As an alternative, a camera that has become available at a relatively low cost in recent years has attracted attention. A method of detecting the presence of a vehicle using a camera has been put into practical use in a traffic condition grasping device or the like. In this traffic condition grasping device, vehicle detection is performed by moving image processing.

  However, since a method using a camera is greatly affected by changes in optical conditions, it is difficult to ensure robustness of a highly accurate detection function. For this reason, it is necessary to develop a technology that ensures the same accuracy as that of a conventional transmission pole sensor after adopting a stereo method in which measures for reducing the influence of optical conditions are adopted. Also, due to the ease of installation and equipment cost reduction requirements, especially when the distance between cameras is short, it is difficult to stably secure parallax at a position away from the camera without being affected by fluctuations in optical conditions. It becomes.

  Accordingly, an object of the present invention is to provide a vehicle detection device that can accurately detect the entry / exit of a vehicle from a photographed image without being affected by changes in optical conditions when a stereo camera is employed.

According to this embodiment, the vehicle detection device, a vehicle detection device which detects the vehicle into and out of the inside of the field of view of the stereo camera based on each of images captured by the stereo camera, and edge enhancement means, Disparity data measurement means, disparity data rejection means, distance data measurement means, distance image creation means, state determination means, and entry / exit detection means are provided. The edge emphasizing unit performs edge emphasis on each of the captured images with respect to an edge existing on a line extending in the vertical direction of the vehicle. The parallax data measuring means measures parallax data for each pixel from each of the captured images with the edge enhanced. The parallax data rejection means obtains the variance of the luminance of the pixels in each captured image, and selectively rejects the parallax data for each pixel from the result. The distance data measuring unit measures distance data indicating a distance between the stereo camera and the imaging target for each pixel from the parallax data. The distance image creating means creates a distance image having the distance for each pixel. The state determination unit divides the distance image into a plurality of regions in the traveling direction of the vehicle to form a plurality of reference sections, and each of the plurality of reference sections includes background distance data measured in advance without a vehicle, said distance data measurement means obtains a difference between the distance data measured for each specified time when entering and leaving the vehicle, the determined every time the variation of the difference between every said reference zone, the variation of the difference for each of said reference sections Is compared with a threshold value to determine a state corresponding to the presence or absence of an object for each time. The entry / exit detection means holds state determination for each time, determines transition of the vehicle, and detects entry / exit of the vehicle.

1 is a schematic diagram illustrating a configuration of a vehicle information processing system to which a vehicle detection device according to an embodiment is applied. The block diagram which shows the structure of the vehicle detection apparatus which concerns on this embodiment. The flowchart which shows the flow of a process of the vehicle detection apparatus which concerns on this embodiment. The relationship diagram which shows the relationship between the distance from a stereo camera, and parallax. The figure which divides a captured image in a vehicle advancing direction and the divided | segmented reference periods a, b, and c. The figure which shows the state number allocated based on the determination result. The figure which shows state number transition when a vehicle passes ahead. The related figure which shows the relationship between the presence position of a vehicle when a vehicle passes forward, and a state number. The flowchart which shows the flow of the corresponding point search matching and matching rejection process shown in FIG. The figure which shows the corresponding point search matching by a reference | standard image and a reference image. The figure which shows the calculation process of the parallax by the periphery matching of an attention pixel and its attention pixel. 11 is a flowchart showing a corresponding point search matching process having a function of verifying the luminance ranges of the standard image and the reference image shown in FIG. The figure for demonstrating the process of setting a dead area in a reference | standard area. The flowchart which has the area and shape determination of the connection component in reference | standard area variation | change_quantity determination. The flowchart which has the 2 step | paragraph switching function of reference | standard height. The flowchart which has a two-step switching function about the background distance of a depth direction.

  Hereinafter, embodiments will be described with reference to the drawings. In addition, the same code | symbol is attached | subjected to the same part in subsequent figures, the detailed description is abbreviate | omitted, and a different part is mainly described.

  FIG. 1 is a schematic diagram illustrating a configuration of a vehicle information processing system to which a vehicle detection device according to the present embodiment is applied. The vehicle information processing system includes a laser sensor 10, a stereo camera 20, an ETC (Electronic Toll Collection) system 30, and a vehicle detection device 100.

  The laser sensor 10 is a sensor that is disposed in the vicinity of the stereo camera 20 and obtains a distance to the traveling lane that is the background of the captured image. The obtained background distance data is sent to the vehicle detection device 100. Here, the laser sensor 10 scans the laser beam vertically from the position close to the stereo camera 20 toward the traveling lane and along the direction crossing the lane, and measures the distance to the point where the laser beam is reflected. This is a sensor for obtaining background distance data.

  The stereo camera 20 is an imaging device in which a plurality of cameras are arranged so as to obtain a predetermined parallax. In the present embodiment, a plurality of cameras are arranged side by side, the lane in which the vehicle travels is captured by each camera, and each captured image is sent to the vehicle detection device 100. Here, as an example, using two cameras, an image captured by one camera is used as a standard image, and an image captured by the other camera is used as a reference image.

  Each camera in the stereo camera 20 is a digital camera that captures a moving image at a preset frame rate. This stereo camera 20 is installed on a pole standing on the island portion of the toll booth, and each camera arranged in the vertical direction with respect to the traveling direction of the vehicle captures an image from an oblique direction so that the vehicle passes laterally ( (Imaging area) is set. That is, the stereo camera 20 is installed at a position where the road surface can be imaged from obliquely above. In addition, the stereo camera 20 is adjusted so that a grounded portion between a tire (axle) and the ground is reflected in a vehicle traveling in the camera field of view. In addition, the stereo camera 20 is installed at a position where imaging can be performed from a direction perpendicular to the traveling direction of the vehicle so that a blind spot due to the vehicle does not occur when imaging a towing structure such as a tow bar. In the toll gate, the stereo camera 20 and the vehicle detection device 100 may be installed at 3 to 4 locations per lane in order to appropriately grasp the position of the vehicle in the lane.

  Note that the stereo camera 20 may be an imaging device that optically obtains upper and lower captured images of the vehicle with a single lens, instead of the upper and lower cameras. For example, the stereo camera 20 may be configured with a single camera if a single lens can branch images from both directions at the same time using a special lens. For example, a lens using a wedge prism polarization system in which two lenses are provided at the light entrance and a prism that bends the optical path is disposed inside can be used as appropriate.

  The captured image obtained by the stereo camera 20 includes a time code indicating the imaging time. The stereo camera 20, the ETC system 30, and the vehicle detection device 100 have a clock device (not shown) that generates time information synchronized with each other. If the stereo camera 20, the ETC system 30, and the vehicle detection device 100 operate in synchronism with other methods, that is, the captured time of the image data of the stereo camera 20 is determined by the ETC system 30 and the vehicle detection device. If 100 can be recognized, the time code may not necessarily be included.

  At this time, in the stereo camera 20, even from a height (within 4.0 m) of a large vehicle, it is assumed that the lower end (tire contact portion) to the upper end (roof) of the vehicle are reflected. Further, in the stereo camera 20, the video is captured as a moving image, and when traveling through the vehicle, assuming that the vehicle travels up to 80 km / h, the travel history into the field of view of the stereo camera 20 is at least per vehicle. It is assumed that several frames (entrance and exit) are shown.

  The ETC system 30 is a toll collection device that automatically collects tolls imposed on vehicles traveling on toll roads such as expressways, and wirelessly communicates with ETC on-board equipment mounted on the vehicles, Get information to identify. This ETC on-board device is generally installed at a position where at least an antenna for performing wireless communication can be visually recognized through a windshield.

  FIG. 2 is a block diagram illustrating a configuration of the vehicle detection device 100 according to the present embodiment.

  As illustrated in FIG. 2, the vehicle detection device 100 includes a display unit 110, a user interface 120, a storage unit 130, a network interface 140, and a control unit 150.

  The display unit 110 is a display device using an LCD (Liquid Crystal Display) or the like, and displays various information including the operation status of the vehicle detection device 100.

  The user interface 120 is an interface that receives instructions from a user such as a keyboard, a mouse, and a touch panel.

  The storage unit 130 is a storage device that stores the control program and control data of the control unit 150, and uses one or more storage units such as an HDD, a RAM, a ROM, and a flash memory. The control data includes, for example, a plurality of parts data representing a three-dimensional shape (eg, tow bar shape pattern, shape data of a circular portion of an axle, shape data of a part of a person), and the like. That is, the storage unit 130 stores a plurality of component data in advance. The shape pattern of the tow bar is information indicating a typical shape pattern of the tow bar, such as a shape pattern indicating a rectangle having a long side along the traveling direction of the vehicle. The shape pattern of the tow bar is stored when distinguishing between two adjacent vehicles and one vehicle that pulls the towed vehicle with the tow bar. Therefore, the shape pattern of the tow bar can be omitted if they are not distinguished from each other.

  The network interface 140 is an interface that communicates with the stereo camera 20 and the ETC system 30 through a network such as a LAN.

  The control unit 150 includes a microprocessor having a memory, operates according to a control program and control data stored in the storage unit 130, and controls each unit of the vehicle detection device 100 in an integrated manner.

  The control unit 150 obtains parallax from each captured image obtained by the stereo camera 20 and calculates distance data with an object in the visual field (in the imaging region) of the stereo camera 20. In addition, the control unit 150 confirms the presence of an object in the field of view based on the difference between the background distance data acquired by the laser sensor 10 and the calculated distance data, records the positional relationship for each time, The moving direction is determined to detect entry / exit of the vehicle. Note that the control unit 150 may have a function of predicting a passing time in the real space (passing time in the communication area of the ETC system 30) in addition to the detection of the entering / leaving vehicle.

  The control unit 150 includes, for example, the following functions (f1) to (f6).

    (F1) An edge enhancement function that performs edge enhancement in the vertical direction of the vehicle for each captured image.

    (F2) A parallax data measurement function for measuring the parallax data between the captured images with edge enhancement for each pixel.

    (F3) A parallax data rejection function that obtains the estimated parallax and dispersion of pixel values in a small area including the target pixel in the standard image and the reference image, and rejects unnecessary parallax data based on the result.

    (F4) A distance data measurement / distance image creation function for obtaining distance data indicating the distance between the stereo camera and the imaging target for each pixel from the parallax data and creating a distance image having this distance data for each pixel.

    (F5) The distance image is divided in the traveling direction of the vehicle to form a plurality of reference sections, and each of the plurality of reference sections is measured in advance with the background distance data and the distance data measuring function measured in the absence of the vehicle. By calculating the difference from the distance data measured at each time when entering and leaving, obtaining the change amount of the difference for each reference section for each time, and comparing the change amount of the difference for each reference section with a threshold value, A state determination function that determines the state at each time.

    (F6) The state determination for each time is held, and the history of the vehicle determines the transition direction of the vehicle by monitoring the flow of the front, side, and rear end, or the flow of the rear end, side, and front, and from the determination result A vehicle detection function that detects vehicle entry and exit.

  Next, the operation of the vehicle information processing system configured as described above will be described.

  FIG. 3 is a flowchart showing a process flow of the vehicle detection device 100 according to the present embodiment.

  First, the vehicle detection apparatus 100 inputs a captured image (one is a reference image and the other is a reference image) captured by the stereo camera 20 by inputting a captured image (step ST1). The vehicle detection apparatus 100 performs parallel equalization and distortion correction on the input captured image in the parallel equalization process (step ST2), and the brightness of the captured image that has been parallel-equalized in the brightness correction process. Correction (step ST3) is performed. The vehicle detection device 100 performs edge enhancement in the vertical direction of the vehicle by the function (f1) of the control unit 150 for each captured image in the image feature extraction processing by the function (f1) of the control unit 150 (step ST4). ).

  In the corresponding point search matching and matching rejection processing, the vehicle detection device 100 obtains a corresponding point for each pixel between each captured image subjected to edge enhancement by a block matching method or the like, and uses the function (f2) of the control unit 150. Parallax data between each captured image is obtained for each pixel (step ST5, step ST6). This corresponding point search matching and matching rejection processing will be described in an embodiment described later. In the distance measurement, the vehicle detection device 100 obtains distance data for each pixel from the parallax data by the function (f4) of the control unit 150, and creates a distance image having this distance data for each pixel (step ST7).

  FIG. 4 is a relationship diagram illustrating the relationship between the distance from the stereo camera 20 and the parallax.

  In practice, as shown in FIG. 4, when an object (car, background) is imaged with the stereo camera 20, the same point on the object appears at different positions in each captured image. This positional shift is called parallax, and the distance from the stereo camera 20 to the object is calculated based on the parallax obtained for each pixel.

  Next, the vehicle detection device 100 measures the number of vehicles from the created distance image.

  As shown in FIG. 3, the vehicle detection device 100 determines, in the reference section change amount determination, the reference distance a that is created on the entrance side in the traveling direction of the vehicle and the reference section that is located on the entry side of the vehicle. b and a reference section c located between the reference section a and the reference section b. FIG. 5 is a diagram illustrating the division of the captured image in the vehicle traveling direction and the divided reference periods a, b, and c.

The vehicle detection device 100 uses the function (f5) of the control unit 150 to perform distance data for each pixel in the reference section a in the distance image and distance data for each pixel in the reference section a in the background distance image, that is, the laser sensor 10. The difference between the acquired background distance data is measured. The vehicle detection device 100 determines that the reference interval “a” of the distance image is in the ON state when the sum total of the differences for each pixel within the reference interval “a” exceeds the threshold value, and if the sum of the difference is not exceeded, The reference section a is determined to be in an off state. Hereinafter, in the reference interval b and the reference interval c, ON / OFF is determined by the same steps. (Step ST8)
The vehicle detection device 100 determines the state number based on the change (on / off) determination results of the reference sections a, b, and c determined by the function (f5) of the control unit 150 in the state determination. . (Step ST9)
FIG. 6 is a diagram illustrating a state number assigned based on the determination result.

  As shown in FIG. 6, in the present embodiment, for example, when all of the reference sections a, b, and c are off, S0 (no vehicle), and when only the reference section a is on, S1 (vehicle head). When only the reference section b is on, S2 (vehicle rear end), when only the reference section c is off, S4 (leading vehicle leading end and leading vehicle rear end), and other states are S3 (vehicle side). To do.

  Based on the determination result of the state number, the vehicle detection device 100 determines the transition state by comparing with the determination history of the past state number by the function (f6) of the control unit 150. FIG. 7 is a diagram illustrating state number transition when the vehicle passes forward. FIG. 8 is a relationship diagram showing the relationship between the vehicle position and the state number when the vehicle passes forward.

  As illustrated in FIGS. 7 and 8, the vehicle detection device 100 determines, for example, that the state number transition “S0⇒S1⇒S3⇒S2⇒S0” is a forward passage of the vehicle. As the determination history for each image frame, the state number overlaps with S0, S1, S1, S3, S3, S3, S3, S2, S2, S0... Sometimes register as history. The vehicle detection device 100 compares these state number transitions with a transition model registered in advance, and determines vehicle traffic. The vehicle detection device 100 can also determine that the vehicle has passed in the reverse direction with the state number transition “S0⇒S2⇒S3⇒S1⇒S0” in the reverse direction.

  Through the above process, the vehicle detection device 100 detects a passing vehicle in the number counting process and measures the number of passing vehicles (step ST10). The vehicle detection device 100 outputs a detection result of the passing vehicle (step ST11).

  According to the above configuration, the vehicle detection device 100 according to the embodiment obtains the parallax data from each captured image obtained by capturing the lane through which the vehicle passes by the stereo cameras 20 installed above and below, and the object is within the field of view. Calculate distance data. Further, the vehicle detection device 100 confirms the presence of an object in the field of view based on the difference between the background distance data and the parallax data, records the positional relationship for each time, determines the vehicle transition, and enters and exits the vehicle. Can be detected.

  The basic configuration and operation of the vehicle detection device 100 have been described above. Here, the embodiment described below mainly relates to distance image creation and reference section change amount determination, and does not impair subsequent vehicle number measurement accuracy by accurately acquiring the distance image.

  In the following embodiments, the case where the vehicle passes forward in the lane will be described.

Example 1
In the first embodiment, a case where a pattern with extremely poor texture change is obtained as an input image is considered.

  FIG. 9 is a flowchart showing the flow of matching point search matching and matching rejection processing shown in FIG. FIG. 10 is a diagram illustrating matching point search matching using a standard image and a reference image. FIG. 11 is a diagram illustrating a parallax calculation process based on a target pixel and peripheral matching of the target pixel.

  As shown in FIG. 9, in the corresponding point search matching (step ST5) of the vehicle detection device 100, as shown in FIG. 10, pixels whose feature points appear from within each captured image with edge enhancement, that is, FIG. (Step ST51), and a rectangular region of interest ROI_L is set at the target pixel position on the reference image (step ST52). Moreover, the vehicle detection apparatus 100 sets the area | region ROI_U of the same size in the same pixel position as ROI_L also on a reference image (step ST53). The vehicle detection device 100 shifts the position of ROI_U set on the reference image and searches for a position that best matches the brightness of ROI_L. The vehicle detection apparatus 100 calculates the difference between ROI_U and ROI_L by, for example, SAD (Sum of Absolute Difference) or the like in the corresponding point matching score measurement, and obtains the position where the calculated difference becomes the minimum value (step ST54). . Thereby, the vehicle detection apparatus 100 measures parallax from the position of ROI_L and the position of ROI_U by parallax measurement (step ST55).

  In principle, the vehicle detection device 100 searches for the similarity of the texture pattern obtained from the captured image of the stereo camera 20 and obtains the coincidence point, thereby obtaining the parallax at an appropriate position.

  By the way, when the illuminance on the road is extremely bright or dark, the vehicle detection device 100 makes it difficult to obtain a texture pattern particularly on the road surface for the captured image of the stereo camera 20 when obtaining the parallax, and a matching error occurs. To do. This is more likely to occur in regions with a higher resolution that are farther from the camera. Further, these areas have large distance measurement errors per pixel unit. For this reason, the matching error is considered as a risk that leads to erroneous detection.

  In order to prevent this erroneous detection, in the matching rejection process (step ST6) of the vehicle detection device 100 shown in FIG. 9, the function (f3) of the control unit 150 measures the variance from the luminance in ROI_L (step ST61). When the measured variance is equal to or less than the threshold value, the vehicle detection device 100 rejects the parallax measurement target, assuming that there is a risk of erroneous matching (step S63). Further, the vehicle detection device 100 similarly measures not only the variance measured from the luminance in the ROI_L but also the variance due to the luminance in the ROI_U (step S62). More rejected (step S63).

  With the above processing, the vehicle detection device 100 can measure an accurate distance by preventing erroneous matching when searching for corresponding points even when a pattern with extremely poor texture change is obtained as an input image.

(Example 2)
In the second embodiment, a case where a low quality image having many pixels exceeding the upper limit and the lower limit of the luminance range of the stereo camera is used is considered.

  FIG. 12 is a flowchart showing a corresponding point search matching process having a function of verifying the luminance ranges of the standard image and the reference image shown in FIG.

  As illustrated in FIG. 12, the vehicle detection device 100 sets a rectangular region of interest ROI_L at the pixel position of interest on the reference image, similarly to the first embodiment (step ST52). Moreover, the vehicle detection apparatus 100 sets the area | region ROI_U of the same size in the same pixel position as ROI_L also on a reference image (step ST53). The vehicle detection device 100 shifts the position of ROI_U set on the reference image and searches for a position that best matches the brightness of ROI_L. The vehicle detection apparatus 100 calculates the difference between ROI_U and ROI_L by, for example, SAD (Sum of Absolute Difference) or the like in the corresponding point matching score measurement, and obtains the position where the calculated difference becomes the minimum value (step ST54). . Thereby, the vehicle detection apparatus 100 measures parallax from the position of ROI_L and the position of ROI_U by parallax measurement (step ST55). Here, the vehicle detection device 100 determines whether or not the luminance of the pixels in the region ROI_U corresponding to the measured parallax is within the upper and lower limits of the preset luminance range by the corresponding point search matching (step) ST56). At this time, if the luminance of the pixel exceeds the upper limit, there is a possibility that an overexposure risk has occurred. Further, when the luminance of the pixel is lower than the lower limit, there is a possibility that a blackout risk has occurred. The vehicle detection apparatus 100 rejects the parallax of the pixels that are out of the preset luminance range from the measurement target.

  Here, if the mechanism for controlling the amount of light of the cameras is not synchronized between the cameras, the vehicle detection device 100 will be overexposed (a large number of luminances exceeding the upper limit) at only one image at the same point in the video. The situation is also conceivable. For this reason, in the vehicle detection device 100 according to the second embodiment, the determination is simply performed for each pixel of interest.

  With the above processing, the vehicle detection device 100 can prevent an erroneous matching of corresponding point search caused by a low-quality image having many pixels exceeding the upper limit and lower limit of the luminance range of the stereo camera, and can measure an accurate distance. it can.

(Example 3)
In the third embodiment, a case where a reflection portion and a shadow portion with strong sunlight are mixed on the road surface is considered.

  If there is a strong reflection of sunlight and shadows on the road surface, edge changes at the borders will become noticeable, and areas with high brightness and dark shadows may exist in the captured image. . In the corresponding point search matching result in this case, an error tends to occur near the boundary portion where the edge change is remarkable.

  In order to prevent this error, as shown in FIG. 9, the vehicle detection device 100 measures the variance from the luminance in the attention area ROI_L on the reference image in the matching rejection process as in the first embodiment (step ST61). ). In the vehicle detection device 100, when the edge change is significant, in the determination based on the parallax / dispersion of the matching rejection process (step ST63), since the luminance in the ROI_L is increased, it is determined whether the variance exceeds the threshold value. . Further, in the vehicle detection device 100, when the resolution of the captured image is low, mismatching is likely to occur. Therefore, the parallax between the attention area ROI_L and the area ROI_U is determined, and it is determined whether it is equal to or less than the threshold value.

  The parallax of the pixel that satisfies the above two determination conditions is rejected from the measurement target.

  With the above processing, the vehicle detection device 100 prevents in advance false matching due to the presence of extreme contrast such as a shadow boundary portion that hinders matching with detailed texture information in the captured image, and the accurate distance Can be measured.

Example 4
In Example 4, a case where texture information is scarce, such as a hood part of an automobile, is considered.

  The vehicle detection device 100 sets a region of interest ROI_L on the reference image by corresponding point search matching. The vehicle detection device 100 arranges the pixels in the ROI_L in ascending order centering on the target pixel in the ROI_L, and when there is a close value that is equal to or less than the threshold, the parallax between the adjacent pixels is considered in consideration of the risk of erroneous matching. Is rejected from the measurement target.

  With the above processing, the vehicle detection device 100 can prevent erroneous matching in advance and measure an accurate distance based on the parallax when the texture of the vehicle hood or the like is poor.

(Example 5)
In Example 5, consider the case where a failure occurs such that an obstacle is continuously observed in a part of the visual field to be detected,
FIG. 13 is a diagram for explaining a process of setting a dead area in the reference section.

  As shown in FIG. 13, the vehicle detection device 100 has a case in which some trouble occurs in the sensor such as mud or raindrops adhering to the camera lens, and an obstacle is continuously observed in a part of the field of view. Consider.

  The vehicle detection apparatus 100 measures the change amount of the reference section from the measured difference in the reference section change amount determination. The vehicle detection device 100 determines the amount of change (on / off) for each reference section. When determining that the vehicle is ON, the vehicle detection device 100 stores the number of continuous ON determinations at that position in the history counter. The vehicle detection device 100 determines whether or not the number of continuous ON determinations stored in the history counter has reached a set value, and when it reaches, sets a dead area flag. Next, the vehicle detection apparatus 100 refers to the insensitive area flag and replaces the change amount determination result from on to off for a section that is continuously on. Replacement takes place in three areas: (1) entry side, (2) central area, and (3) exit side.

  Note that the vehicle detection device 100 initializes the history counter and the insensitive area when the amount of change in the reference section changes from on to off, and restores the original state. At this time, the vehicle detection device 100 may give a condition such as executing it when it is continuously off for n frames or more, instead of clearing zero at a time.

  With the above processing, the vehicle detection device 100 sets the insensitive area when a failure occurs such that an obstacle is continuously observed in a part of the field of view of the detection target. By turning off the determination result of the change amount, it is possible to provide a highly accurate vehicle number measuring function.

(Example 6)
In the sixth embodiment, a case where a relatively small flying object such as rain or snow is present in the field of view will be considered.

  FIG. 14 is a flowchart including the area / shape determination of the connected component in the reference interval change amount determination.

  When the outer contour of an obstacle is observed beyond the reference height due to snow, large rain, or the like, the vehicle detection device 100 is captured as many point-like objects in the field of view, and depending on the quantity, erroneous detection may occur. There is a risk.

  In order to prevent this erroneous detection, as shown in FIG. 14, in the vehicle detection device 100, the distance data included in the distance image acquired in the distance measurement (step ST7) and the background distance data acquired from the laser sensor 10 are used. The difference between them is measured (step ST81). The vehicle detection device 100 compares whether the measured difference is equal to or higher than the reference height with respect to the road surface (step ST82), and does not measure the difference equal to or lower than the reference height. The vehicle detection apparatus 100 measures the number of pixels whose measured difference is equal to or greater than the reference height (step ST83). The vehicle detection apparatus 100 performs labeling (connected component extraction) in consideration of the connectivity between the measured pixels (step ST84), and determines the area of the labeled pixels (corresponding to the number of pixels) (step ST85). A vehicle having a certain area or more is determined as a vehicle candidate, and a vehicle having a certain area or less is determined as noise. The vehicle detection device 100 replaces a distance that is greater than or equal to the reference height with a distance that is less than or equal to the reference height for pixels determined to be noise (step ST86). The vehicle detection apparatus 100 again measures the number of pixels exceeding the reference height (step ST87), and performs state determination based on an on / off combination for each reference section (step ST9). At this time, referring to the shape of the label, as other information of the area, a condition for performing noise determination is provided by shape classification in the vertical direction, horizontal direction, and diagonal direction for each label.

  With the above processing, when a relatively small flying object such as rain or snow is present in the field of view, the vehicle detection device 100 acquires information on the outer contour of the flying object, determines its occupied area, and rejects noise. By doing so, it is possible to accurately measure an obstacle without depending on disturbance of weather or illuminance.

(Example 7)
In the seventh embodiment, a case where the reference height based on the road surface is increased due to snow accumulation or the like will be considered.

  FIG. 15 is a flowchart having a two-step switching function of the reference height.

  As shown in FIG. 15, the vehicle detection device 100 changes the reference height between the standard specification and the standard specification when measuring pixels whose distance per pixel exceeds the reference height in the reference section change amount determination. Two types having higher specifications and added set values are stored in advance. The vehicle detection apparatus 100 refers to the state number from the state determination record (step ST89) after measuring the difference in the reference section change amount determination and before measuring the change amount of the reference section (step ST88). When there is no obstacle detection state, the vehicle detection device 100 applies the specification with a higher reference height. When the number of pixels exceeding the higher specification standard increases and the ON determination is made for the reference section, the state number where an obstacle is detected changes. Thereafter, the vehicle detection apparatus 100 refers to the state number, uses the standard height of the standard specification in the determination of the change amount of the standard section, and similarly measures pixels exceeding the standard height of the standard specification, Measure the amount of change. If the number of pixels exceeding the standard specification is reduced and the reference section is determined to be off, there is no obstacle and the state number changes. If the status number changes and there is no obstacle detection status, it will be raised again and the specified reference height will be applied.

  In addition, the vehicle detection apparatus 100 prepares a plurality of reference heights, and does not depend on the state number, and increases the set value at the reference height (standard + set value) as the distance from the camera increases. Measure the amount of change.

  By the above processing, even when the vehicle detection device 100 detects an obstacle suddenly on the road surface, the tolerance is improved by applying the threshold set with a higher standard considering the vehicle size. Therefore, the vehicle detection device 100 is not affected by a falling object and can accurately measure an obstacle when the vehicle enters.

(Example 8)
In Example 8,
FIG. 16 is a flowchart having a two-stage switching function for the background distance in the depth direction.

  As shown in FIG. 16, the vehicle detection device 100 measures the difference between the distance data included in the acquired distance image and the background distance data acquired from the laser sensor 10 in the reference section change amount determination.

  The vehicle detection device 100 sets an imaging region with two axes, that is, a traveling direction of the vehicle and a distance (depth) direction from the camera. That is, the imaging region is subdivided into two regions on the near side and the far side of the boundary line when a boundary line corresponding to the threshold is set in the distance (depth) direction from the camera. On the near side, the road is usually reflected as the background, and the height from the road surface is measured from the difference between the distance from the vehicle traveling on the near side and the background distance of the road portion. On the other hand, the road is not limited to the background, and it is uncertain that the road is a wall that partitions lanes as seen at a toll booth, or that the background is an adjacent passage. Therefore, the vehicle detection device 100 sets the back side with a virtual road (wall) as a background, and determines the obstacle from the difference between the distance from the vehicle traveling on the back side and the virtual road (wall) distance. Measure the height.

  For each calculated height, the vehicle detection device 100 measures pixels that exceed the specification standard of a reference height prepared in advance, and performs on / off determination of a reference section (entrance / stop / exit part) ( Step ST810).

  With the above processing, the vehicle detection device 100 can stably determine the on / off of the reference section regardless of the position of the passage of the vehicle, and can accurately measure the passage of the vehicle.

  As described above, according to the embodiment, in the matching rejection process of the vehicle detection device 100 illustrated in FIG. 9, the variance is measured from the luminance in the ROI_L. The vehicle detection device 100 rejects the parallax measurement target if a risk of mismatching occurs when the measured variance falls below the threshold. Further, the vehicle detection device 100 similarly measures not only the variance measured from the luminance in the ROI_L but also the variance due to the luminance in the ROI_U, and similarly rejects it from the parallax measurement target when it is equal to or less than the threshold. As a result, the vehicle detection device 100 can prevent an erroneous matching at the time of searching for a corresponding point and measure an accurate distance even when a pattern with extremely poor texture change is obtained as an input image.

  Therefore, the adoption of the stereo camera allows the vehicle detection device of the present embodiment to detect the entry / exit of the vehicle from the captured image with high accuracy without being affected by the change in the optical conditions.

  The laser sensor 10 may be a sensor that scans a laser beam and sends received data obtained by receiving the reflected laser beam to the vehicle detection device 100. In this case, the vehicle detection device 100 needs to have a function of calculating distance data from the received wave data.

  As mentioned above, although embodiment of this invention was described, this embodiment is shown as an example and is not intending limiting the range of invention. This embodiment can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. This embodiment and its modifications are included in the scope of the present invention and the gist thereof, and are also included in the invention described in the claims and the equivalent scope thereof.

  DESCRIPTION OF SYMBOLS 10 ... Laser sensor, 20 ... Stereo camera, 30 ... ETC system, 100 ... Vehicle detection apparatus, 110 ... Display part, 120 ... User interface, 130 ... Memory | storage part, 140 ... Network interface, 150 ... Control part.

Claims (9)

A vehicle detecting apparatus for detecting a vehicle into and out of the inside of the field of view of the stereo camera based on each of images captured by the stereo camera,
The captured images of the respective, relative to edge existing on a line extending in the vertical direction of the vehicle, and edge enhancement means for performing edge enhancement,
Parallax data measuring means for measuring parallax data for each pixel from each of the edge-enhanced captured images;
Dispersion of luminance of the pixels in each captured image, parallax data rejecting means for selectively rejecting the disparity data for each pixel from the result,
Distance data measuring means for measuring, for each pixel, distance data indicating a distance between the stereo camera and the imaging target from the parallax data;
A distance image creating means for creating a distance image having the distance for each pixel;
The distance image is divided into a plurality of regions in the traveling direction of the vehicle to form a plurality of reference sections, and each of the plurality of reference sections is measured in advance in the absence of a vehicle and the distance data measuring means. The difference between the distance data measured at a specified time when the vehicle enters and exits is obtained, the change amount of the difference is obtained for each reference section for each time, and the change amount of the difference is compared with a threshold value for each reference section. Then, state determination means for determining the state corresponding to the presence or absence of the object for each time,
A vehicle detection apparatus comprising: state determination for each time, determination of transition of the vehicle, and detection of entry / exit of the vehicle. The parallax data rejection means refers to a pixel that is a corresponding point of each captured image based on a matching result of corresponding point search, determines whether the pixel exceeds the upper limit or the lower limit of the luminance range, and depending on the determination result The vehicle detection device according to claim 1, wherein the parallax data of the pixel is rejected.   2. The disparity data rejecting unit rejects disparity data of a pixel whose variance in an area referred to at the time of matching in corresponding point search exceeds an upper limit threshold and is equal to or lower than a lower limit threshold. The vehicle detection device described. The parallax data rejecting means arranges pixels of the target region centered on the target pixel in the target region in the captured image that is set by matching corresponding point search in ascending order, below the threshold value set in advance, and close to the threshold The vehicle detection device according to claim 1, wherein the parallax data of the pixel having the obtained value is rejected.   The state determination unit holds a history of changes in each of the plurality of reference sections, measures the number of continuous on determinations for which the determination result of the change amount is determined to be present for each reference section, The vehicle detection device according to claim 1, wherein when the threshold value is exceeded, the state determination result of the corresponding section is replaced with a determination invalid state. The state determination unit acquires connection information between pixels in which a measurement distance of an imaging target for each pixel exceeds a reference distance with respect to a road surface, and an area of each connection region obtained from the connection information is equal to or less than a predetermined value. In this case, the vehicle detection device according to claim 1, wherein the vehicle detection device is regarded as noise and the distance of the corresponding pixel is replaced with a distance equal to or less than the reference distance. The state determination means includes two reference distances having different heights with respect to the road surface, and in a state where there is no object other than the vehicle on the road surface, for comparison of the measurement distance of the imaging target for each pixel. The higher reference distance is applied, and when there is an object other than the vehicle on the road surface, the lower reference distance is applied for comparison of the measurement distance of the imaging target for each pixel. Item 1. The vehicle detection device according to Item 1. The state determination means does not depend on the state determination result, and applies the higher reference distance to the comparison of the measurement distance of the imaging target for each pixel as the vehicle moves away from the stereo camera. The vehicle detection device according to claim 7, wherein a change amount of each reference section is measured.   The state determination means applies a background distance on the road surface for an area close to the stereo camera when determining whether or not the measurement distance for each pixel exceeds a reference distance based on the road surface, and the stereo camera The vehicle detection apparatus according to claim 1, wherein a virtual background distance is applied to a region far from the vehicle.

Images