JP3912869B2 - Traffic flow measuring device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a traffic flow measuring apparatus that measures traffic flow on a highway using image processing technology.
[0002]
[Prior art]
In the conventional measurement of traffic flow at urban high speeds, there is a system in which ultrasonic sensors are installed at predetermined intervals for each lane, and the traffic congestion state of the road is grasped according to the time required for the vehicle to move between the two points.
[0003]
There is a system using a plurality of cameras in traffic flow measurement using conventional image processing. FIG. 10 shows an example of the configuration.
This system installs photographing devices 2a and 2b between two points along the road 1, processes video signals photographed by the respective photographing devices 2a and 2b with the image processing device 3, selects a sample vehicle, The vehicle type and color information is extracted, and the traffic flow state is grasped from the time taken to travel between the two points.
[0004]
[Problems to be solved by the invention]
In the conventional system described above, the traffic flow state can be grasped based on the video signals photographed by the photographing devices 2a and 2b, but a measurement photographing device must be newly installed for each lane, There is a problem that the required accuracy cannot be obtained.
[0005]
The present invention has been made in order to solve the above-described problems. By using an ITV camera for monitoring a road condition that is usually installed on a highway, traffic can be obtained with high accuracy without installing a new photographing apparatus. It aims at providing the traffic flow measuring device which can measure a flow.
[0006]
[Means for Solving the Problems]
The traffic flow measuring apparatus according to claim 1 of the present invention detects a blur of an image from a continuous image input device that reads a continuous image from a camera for monitoring a road condition, and a state of the continuous image input from the continuous image input device. And a vibration detection / correction device that corrects the shake of the detected image, a vehicle detection device that detects a vehicle in a specified region from the continuous image corrected by the vibration detection / correction device, and the vibration The vehicle tracking device that tracks the vehicle in the specified area from the continuous image corrected by the shake detection / correction device, and the specified image passes through the specified region from the continuous image corrected by the vibration shake detection / correction device and its background image. The movement trajectory of the vehicle to be measured is measured, the logical product of the binarized image and the area set for each lane is taken from the continuous image corrected by the vibration shake detection / correction device, and Y-axis projection is applied to the time axis. Shi The trajectory of one vehicle is obtained as a single band from a binary image in which the horizontal axis is time and the vertical axis is the Y direction of the input image, and the ratio of the total area of the band portion to the entire image area is occupied. and the vehicle trajectory measuring apparatus, the vehicle detection device by a combination of the vehicle base number measured from the vehicle trajectory measured by measured vehicle stand number及 beauty the vehicle trajectory measuring apparatus, a vehicle that appears in only one by not including the passage number, and the number of vehicles measuring device for measuring the passing number of vehicles, vehicle speed, which is measured from the vehicle trajectory measured by measured vehicle speed Do及 beauty the vehicle trajectory measuring apparatus by the vehicle tracking system using both the degrees, vehicle appearing only one is characterized by comprising a vehicle speed measuring device for measuring a passing vehicle average speed by not included in the speed measurement.
[0007]
According to a second aspect of the present invention, a traffic flow measuring apparatus according to the first aspect further includes a background image updating apparatus that updates a background image from a continuous image corrected by the vibration blur detection / correction apparatus, and the vehicle trajectory measurement. The apparatus is characterized in that it determines daytime and nighttime based on the average brightness of a still region, and outputs a vehicle-extracted binary image that is shaped according to the determination. Traffic flow measuring device according to claim 3 of the present invention, the in claim 1 or 2, the vehicle running region set for each lane and a measurement target, the vehicle passes the number of the plurality vehicles lines, vehicle speed, vehicle occupancy, characterized in that the measurement by the processing of the image taken by one camera to monitor the entire road.
[0008]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is an overall configuration diagram of a traffic flow measuring apparatus according to the present invention.
In FIG. 1, reference numeral 11 denotes a continuous image input device, which reads a continuous image output from one existing road condition monitoring ITV camera and inputs it to the vibration blur detection / correction device 12. The vibration blur detection / correction device 12 detects image blur from the state of the continuous image and corrects the blur. The corrected image is transferred to the background image update device 13, the vehicle detection device 14, and the vehicle tracking device 15. Output.
[0009]
The background image update device 13 updates the background image from the continuous image in which the vibration shake is corrected, and outputs the background image to the vehicle trajectory measurement device 16. The vehicle trajectory measurement device 16 measures the movement trajectory of the vehicle passing through the designated region from the continuous image corrected by the vibration blur detection / correction device 12 and the background image updated by the background image update device 13, The measurement result is output to the vehicle number measuring device 17 and the vehicle speed measuring device 18. The vehicle trajectory measuring device 16 further measures the occupation rate.
[0010]
Further, the vehicle detection device 14 detects a vehicle in the designated area from the corrected continuous image by the vibration shake detection / correction device 12 and outputs it to the vehicle number measurement device 17. The vehicle number measurement device 17 measures the number of passing vehicles from the vehicle detection signal output from the vehicle detection device 14 and the vehicle locus measured by the vehicle locus measurement device 16.
[0011]
On the other hand, the vehicle tracking device 15 tracks the vehicle in the designated area from the corrected continuous image by the vibration blur detection / correction device 12 and outputs the tracking signal to the vehicle speed measuring device 18. The vehicle speed measuring device 18 measures the passing vehicle average speed from the vehicle tracking signal output from the vehicle tracking device 15 and the vehicle locus output from the vehicle locus measuring device 16.
[0012]
Next, the operation of the above embodiment will be described in the case of measuring the number of vehicles, the speed, and the occupation rate within a certain time on the highway.
First, a continuous image is taken from the ITV camera for road condition monitoring by the continuous image input device 11 and input to the vibration blur detection / correction device 12. This vibration blur detection / correction device 12 performs an optimum overlay on an image (reference image) serving as a standard of continuous images by using a normalization cross-correlation method or a SSDA (Sequential Similarity Detection Algorithm) method. Find the state. The image is cut out by correcting a preset region using the shift amount (vertical direction ΔY, horizontal direction ΔX of the image) as an image blur amount from the optimum overlay state.
[0013]
As shown in the flowchart of FIG. 2, the superposition in the vibration blur detection / correction device 12 may be obtained by correlating only the still area of the road surface portion where the vehicle is not present, not the entire image, and obtaining the deviation amount. That is, the vibration blur detection / correction device 12 first sets a stillness correlation template as a blur detection template (step A1) and sets a correlation calculation region for blur detection (step A2). Next, a shake amount (ΔX, ΔY) is obtained by template matching using normalized cross-correlation (step A3). Then, the blur correction is performed from the blur amount (step A4), and the image is cut out in the blur-corrected area from the original image (step A5). The cut image is output as a processed image to the background image update device 13, the vehicle detection device 14, and the vehicle tracking device 15 (step A6).
[0014]
The background image update device 13 updates the background image B (T) according to the change amount of each point of the previous frame background image B (T-1) and the input image I (T). The point where the amount of change is large is not updated the value of the background image of the previous frame. Specifically, the luminance value B (X, Y, T) of each point (x, y) of the background image is obtained using the following equation.
[0015]
B (X, Y, T) = αI (X, Y, T) + (1-α) B (X, Y, T −1 )
α = 1 / | I (X, Y, T) −B (X, Y, T−1) |
However, when | I (X, Y, T) −B (X, Y, T−1) | <ε, α = 0ε means that the parameter vehicle detection device 14 sets the region ROI (Region Of) set for each lane. The vehicle in the (Interest) is detected according to the processing procedure shown in FIGS. FIG. 3 shows the processing procedure of the vehicle detection device 14, and FIG. 4 shows the details of the vehicle detection image processing in FIG.
[0016]
As shown in FIG. 8, the vehicle detection device 14 first has a binary value in which ROI-1, which is a vehicle detection area set for each lane, is set to "1" and the other areas in the image are set to "0". An image is generated (step B1). FIG. 8 shows an ROI for measuring the number of vehicles. ROI-1 is an area for vehicle detection, and ROI- 3 is an area for vehicle trajectory measurement. Then, image processing is performed to detect the vehicle from the binary image (step B2).
[0017]
In the vehicle detection image processing, smoothing processing (median filter) is performed on the processing region image as shown in FIG. 4 (step C1). Next, using the fact that the front surface of the vehicle has many horizontal edges, a vehicle part is extracted by spatial differential filter processing from the image subjected to shake correction by the vibration shake detection / correction device 12 (step C2). (Step C3). The binarized threshold value is automatically calculated by a known method such as discriminant analysis or is set as a fixed threshold value. Since the differential image is susceptible to noise, the binarized differential image is shaped by the maximum value filter (step C4) and the minimum value filter (step C5) to obtain an edge extracted binary image (step C6).
[0018]
And the logical product of the said vehicle detection image and a ROI image is taken (step B4), and it projects in a X-axis direction (step B5). The projected value is smoothed and binarized with a fixed threshold value, and the longest projected line segment is extracted (step B6). The binarized X-axis range is Y-axis projected (step B7), smoothed and binarized by the same method as in the X-axis projection, and extracted as a line segment. Next, it is determined whether or not the length of the extracted projected line segment is greater than or equal to a threshold value (step B8). If not, the process returns to step B4 via step B3, and vehicle detection is performed for other lanes. To do. If it is determined in step B8 that the length of the projected line segment is equal to or greater than the threshold value, it is assumed that a vehicle exists in the vehicle detection ROI-1, and the X, Y projected line segment coordinates are set as the vehicle region (step B9) The vehicle detection flag of the lane is turned on. The vehicle detection is performed in the same manner for other lanes.
[0019]
After processing several frames, the number of vehicles is incremented in a frame in which the detection flag string is turned on from off, and the time when the vehicle enters the vehicle detection ROI-1. In addition, the time when the same vehicle exits the vehicle detection ROI-1 in the frame in which the detection flag string is turned off from on.
[0020]
The vehicle tracking device 15 performs a vehicle detection operation in the vehicle detection ROI-1 for each lane in the same manner as the vehicle detection device 14, and when detecting the vehicle, tracks the vehicle by template matching.
[0021]
Next, the processing procedure of the vehicle tracking device 15 will be described with reference to the flowchart shown in FIG. First, as shown in FIG. 9, binary images of a vehicle detection ROI-1 and a vehicle tracking ROI-2 for each lane are generated (step D1).
[0022]
Next, as shown in the image processing procedure of FIG. 4, the image subjected to blur correction is subjected to spatial differential filter processing / binarization processing / shaping processing to extract a vehicle (step D2). Then, the vehicle detection process shown in FIG. 3 is performed for all lanes (steps D3 and D4), and it is determined whether or not a vehicle is detected (step D5). If no vehicle is detected, the process goes to step D4 via step D3. Return and execute the vehicle detection process again.
[0023]
The logical product of the image detected by the vehicle and the image of the vehicle detection ROI-1 is taken, and the line segment is extracted by X-axis projection / Y-axis projection in the same manner as the vehicle detection device 14. If the length of the extracted line segment is equal to or greater than the threshold value, it is assumed that a vehicle exists in the vehicle detection ROI-1, and the detected area is cut out as a template for tracking, and the vehicle tracking flag for the lane is turned on. (Steps D6 and D7), and the frame is set as the time when the vehicle detection ROI-1 is entered. At this time, the number of tracked vehicles is incremented (step D8).
[0024]
Then, every time the frame is updated for all the numbers being tracked, the logical product of the image of the blur corrected frame and the image of the vehicle tracking ROI-2 is taken, and within the range of this vehicle tracking ROI-2, The vehicle is tracked by template matching (steps D9 and D10). For template matching, a known cross-correlation method or SSDA method is used.
[0025]
If the vehicle can be tracked by matching, the position of the vehicle is updated (step D12), and the matched area is used as a template in the next frame, and the vehicle tracking ROI-2 is updated (steps D13 and D14). ). Then, it returns to step D10 and continues vehicle tracking. At this time, it is determined whether or not the vehicle position has exited the tracking area (step D11). If the vehicle position has exited from the tracking area, the tracking is terminated, and the frame is set as the time from the vehicle tracking ROI-2. Further, the travel distance is calculated from the position at the end of the tracking and the speed is calculated (step D15), and the vehicle tracking flag is turned off (step D16). The number of tracked vehicles, the tracking start frame, and the end frame are shown.
[0026]
If there is another tracked vehicle in the same lane, the process returns to step D9 to continue the other vehicle tracking process. When the tracking process for one lane is completed, the process returns to step D3 to perform the process for another lane.
[0027]
Further, the vehicle trajectory measurement device 16 measures the trajectory of the vehicle that has passed through the vehicle trajectory measurement ROI-3 for each lane within a certain time, and the measurement results are used as the vehicle number measurement device 17 and the vehicle speed measurement device 18. Output to. From this trajectory, the number, speed, and occupation ratio of vehicles that have passed through the vehicle trajectory measurement ROI-3 within a predetermined time are obtained.
[0028]
In this processing procedure, as shown in step E1 in FIG. 6, first, a binary image of the vehicle trajectory measurement ROI-3 for each lane is generated. This vehicle trajectory measurement ROI-3 is the same as the vehicle tracking ROI-2 in the vehicle tracking device 15 as shown in FIG. Further, the position where the vehicle exits from the vehicle trajectory measurement ROI-3 is set to a position where the vehicle enters the vehicle detection ROI- 1 in the vehicle detection device 14 shown in FIG.
[0029]
Then, image processing for vehicle trajectory measurement is performed on the binary image (step E2). FIG. 7 shows the image processing procedure for this vehicle trajectory measurement.
In this image processing procedure, a difference image between the image subjected to the shake correction by the vibration shake detection / correction device 12 and the background image updated by the background image update device 13 is generated (step F1). After performing the smoothing process (median filter) (step F2) and the binarization process (step F3) on the difference image, it is determined whether it is daytime processing (step F4). A noise extraction process using a filter (step F5) and a maximum value filter (step F6), and a shaping process using a maximum value filter (step F7) and a minimum value filter (step F8) are performed to extract a vehicle portion and extract a vehicle extracted binary image. Is output (step F9). If it is determined in step F4 that it is not daytime processing, that is, nighttime processing, no noise removal processing is performed. Nighttime judgment is made when the average luminance of a still region including a building and a road wall that is not affected by road lights is equal to or less than a threshold value and the variance thereof is equal to or less than the threshold value.
[0030]
Then, as shown in FIG. 6, the logical product of the binarized image extracted from the vehicle and the ROI image is obtained (step E4), Y-axis projection (step E5), and further binarized and smoothed. This is expanded on the time axis (step E6). A trajectory of one vehicle is obtained as one band from the binary image in which the horizontal axis is time and the vertical axis is the Y direction of the input image (step E7). The number of passing vehicles is the number of passing vehicles, the horizontal coordinate value at the start point of the belt is the time when the vehicle enters the ROI-3 for vehicle trajectory measurement, and the horizontal coordinate value of the end point is output from the ROI-3 for vehicle trajectory measurement. Time. Further, the ratio of the total band area to the entire image area is defined as the occupation ratio (step E8).
[0031]
Next, the vehicle number measuring device 17 measures the number of vehicles passing within a predetermined time from the vehicle detected number obtained by the vehicle detecting device 14 and the vehicle locus obtained by the vehicle locus measuring device 16.
[0032]
Measurement procedure utilizes the fact that sets the vehicle trajectory measurement ROI- 3 and the vehicle detection ROI-1, as shown in FIG. 8, out of the vehicle trajectory measuring ROI- 3 is a vehicle trajectory measurement result The vehicle detection result is compared with the time when the vehicle detection result is ROI-1 for vehicle detection. If the frame difference is equal to or less than the threshold value, the vehicle is regarded as the same vehicle, and the number of passing vehicles is incremented. Vehicles that appear only on one side are not included in the number of passing vehicles.
[0033]
Further, the vehicle speed measurement device 18 measures the vehicle speed distribution within a predetermined time from the vehicle tracking result obtained by the vehicle tracking device 15 and the vehicle locus obtained by the vehicle locus measuring device 16.
[0034]
As shown in FIG. 9, the processing procedure uses the fact that the vehicle trajectory measurement ROI-3 and the vehicle tracking ROI-2 are set in the same region, so that the vehicle trajectory measurement ROI-3 is a vehicle trajectory measurement. The time of entering and the time of exiting from ROI-3 are compared with the time of entering the vehicle tracking ROI-2 and the time of exiting from ROI-2, which are vehicle tracking results. Are regarded as the same vehicle, and this is a speed measurement vehicle. Vehicles that appear only on one side are not included in the speed measurement vehicle.
[0035]
【The invention's effect】
As described above in detail, according to the present invention, there is provided a device that processes an image captured by a single camera that monitors the entire road and measures the number of vehicles passing through a plurality of lanes, vehicle speed, and vehicle occupancy. In the number measurement, the vehicle detection ROI-1 vehicle set for each lane is detected to measure the number of vehicles, and the vehicle track measurement ROI- 3 set for each lane within a certain time Thus, the number of vehicles can be measured with high accuracy by measuring the number of passing vehicles by using together with the means for measuring the number of vehicles and not including the vehicles appearing in only one of them in the number of passing vehicles.
[0036]
Further, in the vehicle speed measurement, a means for measuring the vehicle speed by tracking the vehicle of the vehicle tracking ROI-2 set for each lane and the fixed time of the vehicle trajectory measuring ROI-3 installed for each lane The vehicle speed is measured with high accuracy by measuring the average speed of the passing vehicle by not using the vehicle that appears only on one side in the speed measurement. be able to.
[Brief description of the drawings]
FIG. 1 is an overall system configuration diagram of a traffic flow measuring apparatus according to an embodiment of the present invention.
FIG. 2 is a flowchart showing a processing procedure of image blur detection / correction means in the embodiment;
FIG. 3 is a flowchart showing a processing procedure of vehicle detection means in the embodiment.
FIG. 4 is a flowchart showing an image processing procedure of a vehicle detection unit in the embodiment.
FIG. 5 is a flowchart showing a processing procedure of vehicle tracking means in the embodiment.
FIG. 6 is a flowchart showing a processing procedure of vehicle trajectory measuring means in the embodiment.
FIG. 7 is a flowchart showing an image processing procedure of the vehicle trajectory measuring means in the embodiment.
FIG. 8 is a view showing a region ROI for vehicle number measurement in the embodiment.
FIG. 9 is a diagram showing a vehicle speed measurement region ROI in the same embodiment;
FIG. 10 is a diagram showing a system configuration example of traffic flow measurement using a plurality of conventional cameras.
[Explanation of symbols]
11 Continuous Image Input Device 12 Vibration Shake Detection / Correction Device 13 Background Image Update Device 14 Vehicle Detection Device 15 Vehicle Tracking Device 16 Vehicle Trajectory Measurement Device 17 Number of Vehicles Measurement Device 18 Vehicle Speed Measurement Device

Claims (3)

A continuous image input device for reading a continuous image from a camera for monitoring road conditions;
A vibration blur detection / correction device that detects image blur from the state of the continuous image input from the continuous image input device and corrects the blur of the detected image;
A vehicle detection device that detects a vehicle in a specified region from the continuous image corrected by the vibration blur detection and correction device;
A vehicle tracking device that tracks a vehicle in a specified region from the continuous image corrected by the vibration blur detection and correction device;
The moving trajectory of the vehicle passing through the designated area is measured from the continuous image corrected by the vibration shake detection / correction device and the background image thereof, and binarized from the continuous image corrected by the vibration shake detection / correction device. The logical product of the image and the area set for each lane is taken, Y-axis projection is applied to the time axis, the time is plotted on the horizontal axis, and the Y axis of the input image is taken from the binary image . A vehicle trajectory measuring device that obtains a trajectory as a single band and uses the ratio of the total area of the band portion to the entire image area as an occupation rate;
The vehicle detection apparatus by a combination of the vehicle base number measured from the vehicle trajectory measured by measured vehicle stand number及 beauty the vehicle trajectory measuring apparatus, in the vehicle that appears in only one not to include the number of passing vehicles, A vehicle number measuring device for measuring the number of passing vehicles;
In combination with the vehicle tracking system vehicle speed, which is measured from the vehicle trajectory measured by measured vehicle speed Do及 beauty the vehicle trajectory measuring apparatus by, passing by vehicles appearing on only one is not included in the speed measurement A traffic flow measuring device comprising a vehicle speed measuring device for measuring a vehicle average speed.   In addition to a background image update device that updates a background image from a continuous image corrected by the vibration blur detection / correction device, the vehicle trajectory measurement device determines daytime and nighttime with an average brightness of a still region, and The traffic flow measuring device according to claim 1, wherein the vehicle extracted binary image shaped according to the processing is output. The vehicle traveling region set for each lane and a measurement target, the vehicle passes the number of the plurality vehicles lines, vehicle speed, that the vehicle occupancy measured by the processing of the image taken by one camera to monitor the entire road The traffic flow measuring device according to claim 1 or 2.

Images