JP5128339B2 - Traffic flow measurement system - Google Patents

Description

  The present invention relates to a traffic flow measurement system of image type traffic flow measurement technology.

  Traffic flow measurement systems are being introduced for the purpose of providing travel time for vehicles (automobiles) traveling on the road. In particular, in order to reduce installation and maintenance costs, an image-type traffic flow measurement system that measures traffic flow by recognizing images taken with a camera from a loop coil sensor type system that has been used in the past ( In the following, replacement with “Image Tracan” is progressing.

  In traffic flow measurement, a measuring section of several tens of meters is provided on the road, the number of vehicles passing through the measuring section, the average speed of the measuring section, the total number of vehicles passing per unit time (for example, 5 minutes or 1 minute), It is common to obtain statistical values such as average speed, but in the image trakan, in the image taken at the time of congestion / congestion, the vehicle in front of the lane overlaps with the following vehicle, or the vehicle between adjacent lanes The problem of overlapping “so-called“ shielding ”is large, and in particular, there has been a problem that the number measurement accuracy deteriorates.

  For such problems, [Patent Document 1] and [Patent Document 2] increase the vehicle detection accuracy by combining the background difference method and the interframe difference method for each of the grayscale image and the differential image, A method has been devised for detecting the occurrence of traffic jams without distinction.

JP 2006-059183 A JP 2004-005234 A

  The background difference method makes it difficult to update the background during traffic jams, and the interframe difference method makes it difficult to detect moving areas at low speeds such as traffic jams, and it is even more difficult to separate individual vehicles from the moving areas There was a problem of being.

  In addition, since the vehicle is detected from the photographed image and then the detected vehicle is tracked, there is a problem that if the vehicle detection fails due to the difficulty of vehicle separation, subsequent tracking becomes impossible. .

  Therefore, an object of the present invention is to correctly measure a traffic flow even in a traffic situation where shielding occurs during congestion or traffic jams.

To achieve the above object, the present invention is traffic flow measurement system, enter the two-dimensional image from the imaging device, and means for storing the two-dimensional image as image data, the vehicle from the image data detected, the tracker Vehicle detection means for storing data, vehicle tracking means for tracking a tracker, which is an index for tracking a vehicle image, stored in the tracker data, and updating the tracker data; and Vehicle verification means for verifying the possibility of tracking and updating tracker data, statistical value measurement means for measuring at least the number or speed as traffic volume statistical values from the tracker data, and storing them in measurement result data, And means for displaying the measurement result data.

  In the traffic flow measurement system according to the present invention, the vehicle verification means calculates a likelihood that is the likelihood of the tracker.

  The traffic flow measurement system of the present invention is characterized in that the statistical value measuring means changes a statistical value calculation method according to the likelihood.

Further, in the traffic flow measurement system of the present invention, the vehicle verification means includes a positional relationship between a single tracker and a road, a positional relationship between a plurality of trackers, a positional relationship between an edge of a moving object image and a road, and an acceleration of a tracker. The likelihood is obtained by using at least one.

  Further, in the traffic flow measurement system of the present invention, the measurement result display means displays a tracking mark at a vehicle image position corresponding to each tracker, and the likelihood of each tracker obtained by the vehicle verification means, It is displayed in the vicinity of the tracking mark.

Furthermore, traffic flow measurement system of the present invention, the measurement result display means, said tracking marks, is characterized in that the display vary depending on the likelihood.

  In the traffic flow measurement system of the present invention, the vehicle verification means changes the likelihood according to the speed obtained by the statistical value measurement means.

Further, in order to achieve the above object, in the traffic flow measurement system according to the present invention, the vehicle verification means uses a likelihood indicating the likelihood of a tracker as a positional relationship between a single tracker and a road, and a position between a plurality of trackers. The at least one of the relationship , the positional relationship between the edge of the moving body image and the road, and the acceleration of the tracker, the likelihood is changed according to the speed obtained by the statistical value measuring means, the statistical value measuring means The statistical value calculation method is changed according to the likelihood, and the measurement result display means displays a tracking mark at a vehicle image position corresponding to each tracker, and the likelihood is calculated in the vicinity of each tracking mark. displayed on, the tracking marks, is characterized in that the display vary depending on the likelihood.

  Furthermore, in order to achieve the above-mentioned problems, the present invention provides a traffic flow measurement system in which a storage device for executing means for inputting an image and storing it in image data is detected, and a vehicle is detected from the image data and stored in tracker data. A vehicle detection device for executing vehicle detection means, a vehicle tracking processing device for executing vehicle tracking means for tracking the tracker stored in the tracker data and updating the tracker data, and the stored tracker correctly tracking the vehicle A vehicle verification apparatus that executes vehicle verification means for verifying the possibility of updating and updating tracker data, and a statistical value that measures at least the number or speed as a traffic volume statistical value from the tracker data and stores it in the measurement result data A measuring device for executing the measuring means and a display device for executing the means for displaying the measurement result data; It is characterized in that is made.

Further, in the traffic flow measurement system of the present invention, the measurement result display means displays a tracking mark at a vehicle image position corresponding to each tracker, and the likelihood of each tracker obtained by the vehicle verification means, It is displayed in the vicinity of the tracking mark, and at least one of the vehicle speed and the vehicle type is displayed in the vicinity of each tracking mark.

  In the traffic flow measurement system of the present invention, the vehicle verification means detects a speed abnormality of each tracker and corrects the speed abnormality tracker speed.

  Further, in the traffic flow measurement system of the present invention, the vehicle verification means detects a speed abnormality of each tracker, and uses the speed of the tracker and the speed of the tracker excluding the tracker, thereby correcting the speed correction coefficient of the tracker. And the statistical value measuring means corrects the speed measurement result using the speed correction coefficient.

  In the traffic flow measurement system of the present invention, the vehicle verification means detects an abnormal speed of each tracker, obtains a likelihood using the speed, and determines the speed of the tracker and the speed of the tracker excluding the tracker. The speed correction coefficient of the tracker is calculated using the likelihood, and the statistical value measuring means corrects the speed measurement result using the speed correction coefficient.

  According to the present invention, it is possible to provide an image-type traffic flow measurement technique with high measurement accuracy even when a plurality of vehicles overlap each other in an image taken at the time of congestion / congestion and separation is difficult. Furthermore, by correcting without removing the abnormal speed value, it is possible to improve the speed measurement accuracy and at the same time achieve both the unit measurement accuracy and the speed measurement accuracy.

  The traffic flow measurement program and system of the present invention includes a means 1 for inputting an image and storing it in image data, a vehicle detection means 2 for detecting a vehicle from the image data and storing it in the tracker data, and a tracker stored in the tracker data. Vehicle tracking means 3 for tracking and updating tracker data, vehicle verification means 4 for verifying the possibility that the stored tracker is correctly tracking the vehicle and updating tracker data, and traffic volume statistics from the tracker data As described above, it is characterized by comprising at least a statistical value measuring means 5 for measuring the number or speed and storing it in the measurement result data, and a means 6 for displaying the measurement result data.

  The vehicle verification means 4 is characterized in that the likelihood is calculated as the likelihood that the tracker is a vehicle, and the statistical value measurement means 5 further changes the statistical value calculation method according to the likelihood. The likelihood is updated for each frame without the detection means 2 strictly specifying individual vehicles. Furthermore, the number counting and the speed measurement can be performed only when the likelihood is high (= the possibility of being a vehicle is high), and in particular, the measurement accuracy when the shielding occurs can be improved.

  Further, the vehicle verification means 4 uses the position relationship between the single tracker and the road, the position relationship between the plurality of trackers, the position relationship between the edge of the moving object image and the road, and / or the acceleration of the tracker. Calculate the degree. Specifically, depending on the positional relationship and acceleration between the tracker group and the road, the vehicle continues to cross the lane, continues to collide with other vehicles, continues to run in places where it cannot travel, or is physically impossible Judgment of abnormal situation such as movement has occurred, and by combining with the image edge, it can judge abnormal situation such as overhang of lanes of large vehicles, and as a result, calculate likelihood appropriately Can do.

  The measurement result display means 6 displays the tracking mark at the vehicle image position corresponding to each tracker, and displays the likelihood of each tracker obtained by the vehicle verification means in the vicinity of each tracking mark, thereby A user such as a monitor can check whether the traffic flow is correctly measured as a result of the image processing. In particular, when the measured statistical values are abnormal, it is possible to determine whether the traffic flow is really abnormal due to a traffic accident, or whether the traffic flow itself is normal but the image processing result is abnormal (that is, measurement error) it can.

  In addition to the likelihood of each tracker obtained by the vehicle verification means, it is preferable to display at least one of the vehicle speed and the vehicle type in the vicinity of each tracking mark.

  Furthermore, the measurement result display means 6 displays the tracking mark by changing the display method according to the likelihood, so that the user can distinguish between the traffic flow abnormality and the image processing abnormality without reading the likelihood value itself. can do. Since it is difficult to accurately read the likelihood value that continuously changes, this problem solving means can also determine a measurement error when monitoring traffic flow measurement status in real time (not recording). .

  The vehicle verification means 4 can change the likelihood according to the speed obtained by the statistical value measurement means 5, and intends the likelihood in light traffic (when the speed is high) with less influence of shielding. Therefore, measurement errors due to lack of likelihood values can be reduced.

  Furthermore, when a speed abnormality occurs, the vehicle verification means 4 can improve the speed measurement accuracy by correcting the abnormal speed using the speed and likelihood of the surrounding vehicle.

  The effects of the present invention can be obtained by executing the above means by a computer including an information processing device, a display device, an input device, a storage device, and an image input device.

Example 1
An embodiment of a traffic flow measurement program and system according to the present invention will be described with reference to FIGS.

  First, an outline of processing according to this embodiment will be described with reference to FIG.

  The traffic flow measurement program according to the present embodiment includes an initialization process S100, an image input process S120, a vehicle detection process S130, a vehicle tracking process S140, a vehicle verification process S150, a statistical value measurement process S160, a measurement result display process S170, and S120 to S170. The repetitive processing S110 and S180, and image data D110, tracker data D120, and measurement result data D130.

  Details of each data will be described later.

  Next, an outline of the processing procedure will be described.

  Hereinafter, the coordinates in the real space are referred to as world coordinates, and the coordinates in the image space taken by the camera are referred to as camera coordinates.

  First, initialization processing S100 is performed as preprocessing. In this process, each data of D110 to D130 is initialized. In addition, information about the coordinate system such as the position of the lane (on the camera coordinates) is also initialized.

  Next, the process proceeds to repetitive processing for each processing cycle (means S110 to S180).

  Here, the processing cycle is a series of cycles from image input to measurement and result display, and in this embodiment, indicates a cycle according to a video standard such as 100 ms.

  In this iterative process, first, an image is input and stored in the image data D110 (means S120). In this embodiment, the image data is stored on the semiconductor memory, but may be stored as a still image file or a moving image file.

  Next, vehicle detection processing is performed, and the detected vehicle candidate is registered in the tracker data D120 as a tracker (Tracker = tracker) (means S130).

  The detection method used in the vehicle detection process may be a known method. For example, in the present embodiment, first, the image is differentiated, and then the horizontal edge on the camera coordinates (strictly speaking, the edge perpendicular to the lane on the world coordinates when the side-fire angle of the camera with respect to the road traveling direction is large) = A part with a slanted edge on camera coordinates) is searched for, and its neighboring coordinates are registered as a tracker. At this point, a plurality of trackers may be assigned to one vehicle without separating / specifying individual vehicles.

  Details of the tracker data D120 will be described later.

  Next, in the vehicle tracking process, each registered tracker is tracked and the tracker data is updated (means S140). Specifically, with respect to a tracker that has been successfully tracked, the likelihood described later in the tracker data is increased, the camera coordinate value one cycle before is saved, and the current camera coordinate value is updated with the post-tracking coordinate. Further, the lane number of the tracker is updated using the current camera coordinate value.

  The tracking method used in the vehicle tracking process may be a known method. For example, using the template matching method, first, an image of a detected vehicle is registered as an initial template. When tracking, the template is searched in the direction of vehicle travel, and if found, the template is updated as tracking success.

  Next, in the vehicle verification process, it is verified whether each tracker being tracked correctly tracks the vehicle, and the verification result is registered in the tracker data (means S150). Details of the vehicle verification process S150 will be described later.

  Next, in the statistical value measurement process, the number of vehicles and the speed are measured and registered in the measurement result data D130 (means S160). Details of the statistical value measurement process S160 and the measurement result data D130 will be described later.

  Finally, the measurement result registered in the measurement result data is displayed in the measurement result display process (means S170). Details of the measurement result display processing S170 will be described later.

  Next, FIG. 2 shows an apparatus configuration of the traffic flow measurement system.

  The traffic flow measurement system 200 according to the present invention includes an information processing device 210, an input device 220, a display device 230, a storage device 240, a storage medium 250, and an image input device 260.

  The storage medium 250 stores the traffic flow measurement program 100.

  The traffic flow measurement program 100 records the traffic flow measurement program according to the present invention in a format that can be executed by a computer, and is read and executed by the traffic flow measurement system 200.

  The input device 220 executes processing related to input from the user 201 among the processing performed by the traffic flow measurement program 100. Specifically, an input to a measurement result screen 1700 described later is processed.

  The display device 230 executes processing related to display for the user 201 among the processing performed by the traffic flow measurement program 100. Specifically, the display of a measurement result screen 1700 described later is processed.

  Image data D110, tracker data D120, and measurement result data D130 are stored in the storage device 240, and processing, input, and processing are performed on each data in the information processing device 210, the input device 220, and the display device 230, respectively. Display is executed.

  The image input device 260 captures an image used for measurement in real time, or inputs an image from an already recorded moving image or the like, and stores it in the storage device 240.

  In the present invention, the storage device 240 and the storage medium 250 are handled as separate storage devices, but they may be the same device or different devices. For example, the traffic flow measurement program 100 may be stored in the storage device 240 or may be stored on a computer different from the present system that can be accessed through a communication line or the like.

In addition, the storage device and the storage medium in the present invention do not limit the implementation method.
For example, a hard disk drive or a semiconductor memory may be used.

  Furthermore, although the embodiment shows a traffic flow measurement system using a single computer, in the present invention, the above-described apparatus configuration does not necessarily have to be realized by a single computer. For example, when two computers have a communication function with each other, input / output processing is performed using the input device 220 and the output device 230 on one computer, and the information processing device 210 and the storage device 240 are used on another computer. Then, data processing and storage processing may be performed. That is, it may be a stand-alone system or a multi-client system such as a Web system.

  Next, FIG. 3 shows an example of an angle of view of an image handled in this embodiment.

  In the present embodiment, it is assumed that the camera 300 with a built-in image processing function is attached on the pole on the shoulder side and is installed at an angle of view overlooking the measurement area 310. In addition, since the vehicle is photographed from behind, on the camera coordinates, the vehicle enters from the lower side of the screen and moves toward the upper left side of the screen. Further, in FIG. 3, there is a median strip in three lanes on one side, and the road shoulder is omitted. Hereinafter, the lane numbers are referred to as 1, 2, 3 from the left.

  In this angle of view, the lanes 1 to 2 and the lanes 2 to 3 protrude due to the side-fire angle, and the vehicle in the back lane is blocked by the vehicle in the front lane. In addition, when the distance between the vehicles is clogged due to a traffic jam or the like, shielding occurs in the front-rear direction even in the same lane.

  In the present embodiment, an example of rear photographing is given as an example of the angle of view, but the target angle of view in the present invention is not limited. For example, the vehicle may be photographed from the front. In this embodiment, the description will be made with three lanes without a shoulder, but the number of lanes to which the present invention can be applied is not limited.

  Next, FIG. 4 shows the data format of the tracker data D120.

  The tracker data D120 includes, for each tracker, ID, likelihood, instantaneous speed, existing lane, tracker camera coordinates P (t) at the current time t = (Xt, Yt), and at a time t−1 one cycle before. The tracker camera coordinates P (t−1) = (Xt−1, Yt−1) are included.

  The ID is an identifier that identifies the tracker.

  The likelihood is an index indicating the probability that the tracker is correctly tracking the vehicle. Details will be described later.

  The instantaneous speed is the instantaneous speed of the tracker, and is obtained from the processing time of one cycle and the real space moving distance of the tracker.

  The coordinates P (t) and P (t−1) of the tracker are updated by the vehicle tracking process S140.

  Next, FIG. 5 shows a process flow of the vehicle verification process S150.

  In the vehicle verification process S150, the following process is repeated for all trackers (means S500, S560). Hereinafter, since there is a comparison process between trackers (a tracker and another tracker), the tracker referred to in this repetition process is referred to as “own tracker”.

  In the repetitive processing, first, the coordinates of the own tracker are acquired (means S510), and the following processing is performed for the own tracker.

  In the lane crossing determination process, it is determined whether or not the own tracker crosses the lane (means S520).

  In the inter-tracker distance determination process, it is determined whether or not the distance between the own tracker and another tracker is not abnormal (means S530).

  In the travel impossibility determination process, it is determined whether or not the tracker is not traveling on an area where travel is not possible (means S540).

  In the occlusion determination process, it is determined whether another tracker in the vicinity of the own tracker has been illegally generated due to an overhang between the lanes of the vehicle being tracked by the own tracker (means S550).

  Details of each determination process from means S520 to S550 will be described later.

  Next, FIG. 6 shows a process flow of the lane crossing determination process S520.

  In the lane crossing determination process S520, first, it is determined whether or not the coordinates of the tracker are on the lane position (means S600).

  If the tracker is on the lane position, it is considered that the possibility that the tracker has correctly captured the vehicle position has decreased, and the likelihood of the tracker is lowered (means S610).

  If the tracker is not on the lane position, it is assumed that the tracker correctly captures the vehicle position, and the process returns as it is (means S620).

  Next, FIG. 7 shows a processing example of the lane crossing determination processing S520.

  In FIG. 7, a truck is running on the lane 2, and two trackers 710 and 720 are assigned to the truck.

  When the lane crossing determination process is executed, the likelihood does not change because the tracker 720 does not straddle any lane, but the likelihood decreases because the tracker 710 straddles the lane marker between the lanes 2 and 3. Therefore, if the tracking is continued, the likelihood of the tracker 710 continues to decrease, and the possibility of being rejected increases.

  Next, FIG. 8 shows a processing flow of the inter-tracker distance determination processing S530.

  In the inter-tracker distance determination processing S530, the positional relationship between the own tracker and other trackers is checked (means S800, S850).

  First, it is confirmed whether or not the selected other tracker has been compared (checked) with the own tracker (means S810).

  If already checked, the process proceeds to the next other tracker process (means S850).

  If not checked, next, the coordinates of another tracker to be compared with the own tracker are acquired (means S820).

  Next, it is determined whether or not the distance between the own tracker and the other tracker is short (means S830).

  When the distance between the own tracker and the other tracker is short, it is considered that two or more trackers are assigned to the same vehicle, and the likelihood of the other tracker is lowered (means S840).

  If the distance is not close, it is considered that the two trackers are assigned to different vehicles, and the comparison with the next other tracker is performed (means S850).

  A method for determining whether or not the distance between the own tracker and the other tracker is short is not particularly limited. For example, it may be determined that the number of pixels between the trackers on the camera coordinates is less than or equal to a predetermined threshold, or the camera coordinates of each tracker are converted to world coordinates, and the actual distance between the trackers is the actual distance. It may be determined that the distance is less than a distance that cannot be traveled (for example, 0.3 m between the front and rear vehicles or 0.2 m to a parallel vehicle).

  Next, FIG. 9 shows a processing example of the inter-tracker distance determination processing S530.

  In FIG. 9, a passenger car is running on the lane 2, and the own tracker 900 and another tracker 910 are assigned to the passenger car.

  Here, when the inter-tracker distance determination process is executed, the distance of the other tracker 910 is too close to the own tracker 900, and thus the likelihood of the other tracker 910 decreases. Therefore, if the tracking is continued, the likelihood of the tracker 910 continues to decrease, and the possibility of being rejected increases.

  Next, FIG. 10 shows a processing flow of the travel impossibility determination processing S540.

  In the travel impossibility determination processing S540, it is first determined whether or not the own tracker coordinates are on the travel impossibility position (means S1000).

  If the tracker is on the untravelable position, it is considered that the possibility that the tracker has correctly captured the vehicle position has decreased, and the likelihood of the tracker is lowered (means S1010).

  If the tracker is not on the untravelable position, it is assumed that the tracker correctly captures the vehicle position, and the process returns as it is (means S1020).

  Here, the untravelable position is a position where the vehicle cannot physically travel (such as a median strip) and a position where the vehicle can physically travel but is not logically traveling or is highly likely (when there is no traffic jam) These positions are calculated from the lane position set when the initialization process S100 is executed.

  Next, FIG. 11 shows a processing example of the travel impossibility determination processing S540.

  In FIG. 11, a truck is running on the lane 3, and two trackers 1110 and 1120 are assigned to the truck.

  Here, when the travel impossible determination process is executed, the likelihood does not change because the tracker 1120 is on the lane 3, but the likelihood is high because the tracker 1110 is in the central separation zone where the tree 1100 is located, that is, the travel impossible region. Go down. Therefore, if the tracking is continued, the likelihood of the tracker 1110 continues to decrease, and the possibility of being rejected increases.

  Next, FIG. 12 shows a process flow of the shielding determination process S550.

  In the shielding determination process S550, the shielding relationship between the own tracker and another tracker is checked (means S1200, S1270).

  First, it is confirmed whether or not comparison (check) between the selected other tracker and the own tracker has been completed (means S1210).

  If already checked, the process proceeds to comparison with the next other tracker (means S1270).

  If not checked, next, the coordinates of another tracker to be compared with the own tracker are acquired (means S1220).

  Next, it is determined whether or not the coordinates of the own tracker and the coordinates of the other trackers are side by side (strictly speaking, they are aligned perpendicular to the lane in the real space) (means S1230).

  If not, the two trackers are considered to be assigned to different vehicles, and the process proceeds to comparison with the next other tracker (means S1270).

  In the case of side-by-side alignment, an edge on the image at the own tracker position is further extracted (means S1240), and it is confirmed whether the edge is horizontally long and straddles the lane (means S1250).

  If the edge is horizontally long and does not cross the lane, the two trackers are considered to be assigned to different vehicles, and the process proceeds to comparison with the next other tracker (means S1270).

  If the edge is horizontally long and crosses the lane, it is assumed that the other tracker is assigned to the part where the vehicle in the tracker's position protrudes into the adjacent lane (that is, the same vehicle), and the likelihood of the other tracker Is lowered (means S1260).

  Note that the method for increasing or decreasing the likelihood described above is not particularly limited. For example, quantitative value addition / subtraction, such as adding 1 to the likelihood when tracking is successful and subtracting 1 to decrease the likelihood, may be performed. Further, an arbitrary decimal value may be used instead of the quantitative value, or the likelihood obtained in the vehicle verification process S150 may be normalized as a probability value of 0.0 to 1.0.

  Furthermore, in this embodiment, the processing configuration is used in which the likelihood is increased in the vehicle tracking process S140 and the likelihood is decreased in the vehicle verification process S150. However, the timing for increasing or decreasing the likelihood in the present invention is not limited. Absent. For example, instead of increasing the likelihood at the time of tracking, a method of increasing the likelihood when it is determined to be normal at the time of vehicle verification (for example, when it is determined not to cross at the time of lane crossing determination) has the same function as this embodiment. realizable.

  Next, FIG. 13 shows a processing example of the shielding determination processing S550.

  In FIG. 13, a truck is running in the lane 2, and the own tracker 1320 and another tracker 1310 are assigned to the truck.

  Here, when the occlusion determination process is executed, since the own tracker 1320 and the other tracker 1310 are arranged side by side, the edge 1300 on the image is extracted, and the edge 1300 straddles the lanes 2 and 3, and thus the likelihood of the other tracker 1310. Go down. Therefore, if the tracking is continued, the likelihood of the tracker 1310 continues to decrease, and the possibility of being rejected increases.

  Next, FIG. 14 shows a process flow of the statistical value measurement process S160.

  In the statistical value measurement processing S160, the following processing is repeated for all trackers (means S1400, S1450).

  In the iterative process, first, it is determined whether or not the likelihood of the own tracker exceeds a threshold (means S1410).

  If it exceeds the threshold, it is considered that the tracker is correctly assigned to the vehicle, the number is added (means S1420), and the speed is calculated (means S1430).

  When the threshold value is not exceeded, it is considered that there is a high possibility that the tracker is not correctly assigned to the vehicle, and the number is subtracted (means S1440).

  The threshold is determined according to the likelihood calculation method set in the vehicle tracking process S140 and the vehicle verification process S150. For example, when adding or subtracting the likelihood by one, if threshold = 5, it is regarded as a vehicle if tracking is successful for five cycles or more. When the threshold value is normalized to 0.0 to 1.0, if the threshold value is 0.8, the vehicle is regarded as a vehicle when the probability that the vehicle seems to exceed 0.8.

  Next, FIG. 15 shows the data format of the measurement result data D130.

  The measurement result data D130 includes, for each tracker, ID, likelihood, average speed, existing lane, camera coordinates P (t) = (Xt, Yt) of the tracker at the current time t, and the tracker at the time ts at the start of tracking. Camera coordinates P (ts) = (Xts, Yts).

  ID, likelihood, lane, and P (t) hold the same data as the same name data of the tracker data D120.

  The average speed is an average speed of the tracker, and is obtained from the time difference between the tracking start time and the current time and the real space moving distance of the tracker.

  The tracker coordinates P (ts) are registered by the vehicle detection process S130.

  Next, FIG. 16 shows a process flow of the measurement result display process S170.

  In the measurement result display process S170, first, the following process is repeated for all trackers (means S1600, S1660).

  In the iterative process, first, the likelihood value of the tracker is displayed (means S1610).

  Next, it is determined whether or not the likelihood of the tracker exceeds a threshold value (means S1620).

  If the threshold value is exceeded, the tracker is considered to be correctly assigned to the vehicle, and a high likelihood tracking mark is displayed (means S1630).

  If the threshold is not exceeded, it is considered that there is a high possibility that the tracker is not correctly assigned to the vehicle, and a low likelihood tracking mark is displayed (means S1640).

  Next, the instantaneous speed of the tracker is displayed (means S1650).

  When the display process for each tracker is completed, the statistical value is finally displayed (means S1670).

  The display position is illustrated in FIG.

  Next, FIG. 17 shows an example of a measurement result screen.

  On the measurement result screen 1700, the tracking marks 1710 and 1740 are superimposed on the input image, the instantaneous speed 1720 is below the tracking marks, the likelihood values 1730 and 1750 are statistically displayed on the tracking marks, and the statistics are displayed on the road area. A value 1760 is displayed for each.

  In the present embodiment, the tracking mark is switched between a solid line display and a dotted line display according to the likelihood of the tracker.

  In the example of FIG. 17, the likelihood of the tracking mark 1710 is 9, which is determined to be high likelihood and displayed as a solid line, but the likelihood of the tracking mark 1740 is 3, which is low likelihood. It is determined that the dotted line is displayed. In the statistical value display at the upper right, the cumulative number of passing vehicles for each lane from the start of measurement and the average speed of all passing vehicles are displayed.

  This display example does not limit the display method in the present invention. For example, the tracking mark may be displayed by changing the color of the high likelihood in blue, the low likelihood in yellow, or the like, or may be continuously changed from yellow to blue according to the likelihood value. .

  Furthermore, regarding the statistical value display, the cumulative passing number and average speed from the start time of the previous measurement unit time (for example, 5 minutes or 1 minute) may be displayed instead of from the measurement start time.

  In addition, regarding the positional relationship of each display item, the positional relationship may be changed, such as displaying the likelihood value at the lower right of the tracker and displaying the instantaneous speed upward.

(Example 2)
Another embodiment of the traffic flow measurement program and system according to the present invention will be described with reference to FIGS.

  In the present embodiment, in addition to the function of the first embodiment, a determination based on the acceleration of the tracker is added as a likelihood determination criterion.

  First, FIG. 18 shows a process flow of the vehicle verification process S150 in the second embodiment.

  In the vehicle verification process S150 in the second embodiment, the same processes as in the first embodiment are performed for the means S510 to S550, but thereafter, the acceleration determination process S1800 is performed.

  Next, FIG. 19 shows a processing flow of the acceleration determination processing S1800.

  In the acceleration determination processing S1800, first, using the current tracker coordinate P (t) and the previous tracker coordinate P (t-1), the movement amount of the own tracker is calculated (means S1900), and the speed and acceleration are calculated. Is calculated (means S1910).

  Next, it is determined whether or not the acceleration in the traveling direction is abnormal (means S1920).

  If the acceleration in the traveling direction is abnormal, the likelihood of the tracker is lowered (means S1930) and the process proceeds to means 1940.

  If the acceleration in the traveling direction is not abnormal, the process proceeds to the means 1940 as it is.

  Next, it is determined whether or not the lateral acceleration is abnormal (means S1940).

  If the lateral acceleration is abnormal, the likelihood of the tracker is lowered (means S1950) and the process returns (S1960).

  If the lateral acceleration is not abnormal, the process returns as it is (S1960).

  Here, the determination method of whether the acceleration is abnormal is not particularly limited. For example, with respect to the traveling direction, an acceleration that accelerates from 0 km / h to 100 km / h in one cycle is physically impossible, so it is regarded as abnormal. Also, if the own tracker acceleration is significantly larger than the forward tracker acceleration and it is estimated that a collision will occur after one cycle, it is regarded as abnormal. Further, the lateral acceleration is regarded as abnormal because an acceleration that crosses two lanes in one cycle is physically impossible.

  Next, FIG. 20 shows a processing example of the acceleration determination processing S1800.

  In FIG. 20, a passenger car is running on lane 1, a truck is running on lane 2, and a passenger car is running on lane 3. The upper left is the vehicle position at the current time t, and the lower right is the time t- The vehicle position at the time Δt is shown.

  For lane 1, the tracker 2000 is normally assigned to the vehicle one cycle before, but at this time the tracker (2010) is stuck to the road surface. In such a case, the traveling direction acceleration becomes abnormal, and the likelihood of the tracker 2010 decreases. The tracker attached to the road surface hardly moves even if it is tracked. Therefore, if the tracking is continued, the likelihood of the tracker 2010 continues to decrease and the possibility of being rejected increases.

  For lanes 2 and 3, tracker 2030 is assigned to the truck and tracker 2020 is assigned to the passenger car one cycle before, but currently the tracker (2040) assigned to the passenger car is assigned to the lane 2 truck. This is an example. In such a case, the lateral acceleration becomes abnormal, and the likelihood of the tracker 2040 decreases. If the tracking is continued, the tracker 2040 is now in a lane crossing state, so the likelihood of the tracker 2040 continues to decrease and the possibility of being rejected increases.

(Example 3)
Another embodiment of the traffic flow measurement program and system according to the present invention will be described with reference to FIGS.

  In the present embodiment, in addition to the functions of the second embodiment, a function for reducing a speed measurement error, which is a problem in image traffic flow measurement using a monocular camera, is realized.

  First, the principle that a speed measurement error occurs will be described with reference to FIG.

  In FIG. 21, the traveling speeds of the vehicles 2105 and 2110 taken by the camera 2100 are measured, and the vehicle 2105 indicates the vehicle position at time t, and the vehicle 2110 indicates the vehicle position at time t + Δt.

  In image-type traffic flow measurement using a monocular camera, since there is no depth (distance) information of the image captured by the camera, coordinate calculation is performed assuming that each pixel in the image corresponds to a road surface on the camera visual axis. . For example, the coordinate calculation is performed assuming that the bumper point 2115 at the time t is located at the point 2120 on the road surface on the image.

  Here, when the moving distance is compared between the case where the lower part of the vehicle (bumper lower edge in FIG. 21) is captured as the vehicle tracking point and the upper part of the vehicle (roof edge in FIG. 21) is captured, the bumper lower edges 2115 and 2135 are obtained. When captured, the distance between the corresponding road surface points 2120 to 2140 is regarded as the moving distance Dt on the image at the time difference Δt. In this case, there is almost no difference between the true moving distance between the bumper lower edges (2115 to 2135) in the real space and the moving distance Dt on the image. On the other hand, when the roof edges 2125 and 2145 are captured, the distance between the corresponding road surface points 2130 to 2150 is regarded as the moving distance Dm on the image at the time difference Δt. In this case, the moving distance Dm on the image becomes larger than the true moving distance between the roof edges (2125 to 2145) in the real space. In either case, the movement time Δt is the same, but when the roof edge is captured, the movement distance Dm becomes large, so that there is a problem that the measured speed becomes high (over-measurement).

  The present embodiment realizes a function that solves this overspeed measurement by appropriately correcting the speed.

  Next, FIG. 22 shows a process flow of the vehicle verification process S150 in the third embodiment.

  In the vehicle verification process S150 in the third embodiment, the same processes as those in the first and second embodiments are performed for the means S510 to S550. Thereafter, instead of the acceleration determination process S1800 in the second embodiment, a speed / acceleration determination process S2200 is performed. Run.

  Next, FIG. 23 shows a processing flow of the speed / acceleration determination processing S2200.

  In the speed / acceleration determination process S2200, the means S1910 and 1910 perform the same process as in the second embodiment, and then determine whether the speed and acceleration in the traveling direction are abnormal (means S2300). In this process, whether the speed is abnormal or not is added to the traveling direction acceleration abnormality determination S1920 of the second embodiment. As a reference for the speed abnormality, the own tracker speed is compared with other tracker speeds in the same lane. Use significantly higher. For example, if the preceding tracker speed is 20 km / h with respect to the own tracker speed of 80 km / h, it is likely that the own tracker will make a rear-end collision immediately, and therefore it can be determined that the speed is abnormal.

  Next, after performing the process of means S1930 as in the second embodiment, if the speed is abnormal, the speed correction process S2310 is executed. Details of this processing will be described later.

  Note that the processing after the means S1940 is the same as that in the second embodiment.

  Next, FIG. 24 shows a processing flow of the speed correction processing S2310.

  In the speed correction process S2310, first, a preceding tracker running in the same lane as the own tracker is searched (means S2400), and a near speed correction coefficient Cv1 is calculated (means S2405). Cv1 is defined as the preceding tracker instantaneous speed / own tracker instantaneous speed. That is, it is a coefficient for correcting the own tracker speed to the same value as the preceding tracker speed.

  Next, an average value Va of instantaneous speeds in the same lane is calculated (means S2410), and an in-lane speed correction coefficient Cv2 is calculated (means S2415). Cv2 is defined as Va / owner tracker instantaneous speed. That is, it is a coefficient for correcting the own tracker speed to the same value as the average speed in the lane.

  Next, the neighborhood likelihood ratio Rt1 is calculated (means S2420). Rt1 is defined as 1+ (local tracker likelihood / previous tracker likelihood). This is a coefficient that represents how reliable the own tracker speed and the preceding tracker speed can be using the likelihood that represents the likelihood of a vehicle.

  Next, a neighborhood speed correction coefficient Cv3 with likelihood is calculated (means S2425). Cv3 is defined as Cv1 × Rt1. That is, it is a coefficient for correcting in consideration of the reliability based on the likelihood so that the own tracker speed approaches the preceding tracker speed.

  Next, the average lane tracker likelihood value La is calculated (means S2430), and the in-lane likelihood ratio Rt2 is calculated (means S2435). Rt2 is defined as 1+ (owner tracker likelihood / La). This is a coefficient representing how reliable the instantaneous speed average value Va in the same lane as the own tracker speed is using the likelihood.

  Next, a lane speed correction coefficient Cv4 with likelihood is calculated (means S2440). Cv4 is defined as Cv2 × Rt2. That is, it is a coefficient for correcting in consideration of the reliability based on the likelihood so that the own tracker speed approaches the instantaneous speed average value Va in the same lane.

  Finally, Cv1 to Cv4 are clipped with the maximum value 1.0 (means S2445). This is to prevent overcorrection.

  In addition, it correct | amends by speed calculation process S1430 in statistical value measurement process S160 using the correction coefficient Cv1-4 calculated | required by speed correction process S2310. Here, several methods can be used as the correction method. For example, when the measurement speed before correction of the tracker is V and the measurement speed after correction is Vc, simply Vc = V × Cv1 (similarly Vc = V × Cv2, Vc = V × Cv3, Vc = V × Cv4), the average value of the four corrected speeds may be used as the final corrected speed, or the average value Cva of Cv1 to Cv4 may be obtained and Vc = V × Cva. The speed correction method in the present invention is not limited.

  Next, FIG. 25 shows a processing example of the speed correction processing S2310.

  In FIG. 25, there are two trucks in lane 2 and two passenger cars in lane 3, and trackers (2500 and 2520) are assigned to the lower part of the normal bumper on the passenger cars. It is in a situation where it is allocated and protrudes further into lane 3. In such a situation, the tracker 2510 is considered to be traveling in the lane 3, but the speed before correction shows a high value due to the problem of overspeed measurement described above.

  Here, it is assumed that the likelihood L1 of the tracker 2500 is 10 and the speed V1 is 20 km / h. Similarly, the tracker 2510 has L2 = 2, V2 = 40 km / h, the tracker 2520 has L3 = 8, and V3 = 30 km / h. Assuming that the own tracker is 2510 and the other trackers are 2500 and 2520, Cv1 to Cv4 are calculated by the speed correction processing S2310. For example, Cv1 = (20/40) = 0.5, Cv2 = ((20 + 30) / 2) / 40 = 0.625, and similarly Cv3 = 0.6, Cv4≈0.764.

  Finally, when the speed of V2 is corrected using the obtained Cv1 to Cv4, the speed 40 km / h before the correction is corrected to 20 km / h to slightly over 30 km / h.

  In the present invention, in a computer environment for executing the processing according to the present invention, even if any one processing means in the above embodiment is divided into two or more processing means, two or more arbitrary processing means can be realized. The processing means may be integrated and realized as one processing means, and the implementation form is not limited as long as the function provided by the present invention is not impaired.

  The traffic flow measurement program and system of the present invention can be widely applied to image-type traffic flow measurement technology.

The figure which shows the traffic flow measurement program by this invention, and the whole process outline | summary of a system. The figure which shows the apparatus structure of the traffic flow measurement system by this invention. The figure which shows the view angle example of a process target image. The figure which shows the data format and example of tracker data. The flowchart of a vehicle verification process. The flowchart of a lane crossing determination process. The figure which shows the process example of a lane crossing determination process. The flowchart of the distance determination process between trackers. The figure which shows the process example of the distance determination process between trackers. The flowchart of a travel impossibility determination process. The figure which shows the process example of a driving impossibility determination process. The flowchart of a shielding determination process. The figure which shows the process example of a shielding determination process. The flowchart of a statistical value measurement process. The figure which shows the data format and example of measurement result data. The flowchart of a measurement result display process. The figure which shows the example of a measurement result screen. 10 is a flowchart of a vehicle verification process in the second embodiment. The flowchart of an acceleration determination process. The figure which shows the process example of an acceleration determination process. Explanatory drawing of the generation principle of speed measurement error. 10 is a flowchart of vehicle verification processing in the third embodiment. Flow chart of speed / acceleration judgment processing Flow rate correction process flowchart The figure which shows the process example of a speed correction process.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Traffic flow measurement program S100 Initialization process S110 Repeat process S120 Image input process S130 Vehicle detection process S140 Vehicle tracking process S150 Vehicle verification process S160 Statistical value measurement process S170 Measurement result display process S180 Repeat process D110 Image data D120 Tracker data D130 Measurement result data

Claims (13)

Enter the two-dimensional image from the imaging device, and means for storing the two-dimensional image as image data,
Vehicle detection means for detecting a vehicle from the image data and storing it in tracker data;
Vehicle tracking means for tracking a tracker, which is an index for tracking a vehicle image , stored in the tracker data, and updating the tracker data;
Vehicle verification means for verifying the possibility that the stored tracker is correctly tracking the vehicle and updating tracker data;
Statistical value measuring means for measuring at least the number or speed as traffic volume statistics from the tracker data, and storing it in measurement result data;
Means for displaying the measurement result data;
A traffic flow measurement system characterized by comprising. The vehicle verification means calculates the likelihood that is the likelihood of the tracker.
The traffic flow measurement system according to claim 1. The statistical value measuring means changes a statistical value calculation method according to the likelihood.
The traffic flow measurement system according to claim 2. The vehicle verification means obtains likelihood using at least one of a positional relationship between a single tracker and a road, a positional relationship between a plurality of trackers, a positional relationship between an edge of a moving object image and a road, and an acceleration of the tracker.
The traffic flow measurement system according to any one of claims 1 to 3, wherein The measurement result display means includes
A tracking mark is displayed at the vehicle image position corresponding to each tracker,
In addition, the likelihood of each tracker obtained by the vehicle verification means is displayed in the vicinity of each tracking mark.
The traffic flow measurement system according to claim 4, wherein: The measurement result display means, the tracking mark, and displays the changed depending on the likelihood
The traffic flow measurement system according to claim 5, wherein: The vehicle verification means includes
The likelihood is changed according to the speed obtained by the statistical value measuring means.
The traffic flow measuring system according to claim 1 or 2, wherein The vehicle verification means uses a likelihood indicating the likelihood of the tracker as a positional relationship between a single tracker and a road, a positional relationship between a plurality of trackers, a positional relationship between an edge of a moving object image and a road, and an acceleration of the tracker. Using one,
Change the likelihood according to the speed obtained by the statistical value measuring means,
The statistical value measuring means changes a statistical value calculation method according to the likelihood,
The measurement result display means displays a tracking mark at a vehicle image position corresponding to each tracker, and displays the likelihood in the vicinity of each tracking mark,
The tracking mark, and displays the changed depending on the likelihood
The traffic flow measurement system according to claim 1. The measurement result display means includes
A tracking mark is displayed at the vehicle image position corresponding to each tracker,
In addition, the likelihood of each tracker obtained by the vehicle verification means is displayed in the vicinity of each tracking mark , and at least one of the vehicle speed and the vehicle type is displayed in the vicinity of each tracking mark.
The traffic flow measurement system according to claim 4 or 5, wherein The vehicle verification means includes
Detects abnormal speed of each tracker and corrects the abnormal speed tracker speed.
The traffic flow measurement system according to any one of claims 1 to 3, wherein The vehicle verification means includes
While detecting the speed abnormality of each tracker, the speed correction coefficient of the tracker is calculated using the speed of the tracker and the speed of the tracker excluding the tracker,
The statistical value measuring means corrects the speed measurement result using the speed correction coefficient.
Traffic flow measurement system of claim 1 0, characterized in that. The vehicle verification means includes
While detecting the speed abnormality of each tracker, the likelihood is obtained using the speed, the speed of the tracker, the speed of the tracker excluding the tracker, and the likelihood are used to calculate the speed correction coefficient of the tracker,
The statistical value measuring means corrects the speed measurement result using the speed correction coefficient.
Traffic flow measurement system of claim 1 0, characterized in that. A storage device for inputting a two-dimensional image from the imaging device and executing means for storing the two-dimensional image as image data;
A vehicle detection device for executing vehicle detection means for detecting a vehicle from the image data and storing the vehicle in tracker data;
A vehicle tracking processing device for executing a vehicle tracking means for tracking a tracker, which is an index for tracking a vehicle image , stored in the tracker data, and updating the tracker data;
A vehicle verification device that verifies the possibility that the stored tracker is correctly tracking the vehicle and executes vehicle verification means for updating the tracker data;
A measuring device for measuring at least the number or speed of traffic statistics from the tracker data and executing a statistical value measuring means for storing in the measurement result data;
A display device for executing means for displaying the measurement result data;
A traffic flow measurement system characterized by comprising.

Images