JPH03200067A - Method for separating vehicles in picture processing system vehicle sensor - Google Patents

Abstract

PURPOSE:To accurately measure a traffic volume by recognizing the existence of parallel vehicles surely and detecting the vehicles separately when a picture which indicates the existence of the parallel vehicles in an image picking up picture exists. CONSTITUTION:When a part of a parallel traveling small vehicle 12 which travels the adjacent lane of a discriminating object large vehicle 11 is hidden for instance, a position on the measuring area plane (x) axis of a feature point C of the vehicle 12 observed while overlapping at a time (t) is (a) and on an optical axis 1 which connects the point(a) and a camera center line, actual feature points C and G of the automatic vehicle are present together. On the other hand, in the condition of a time t', the feature points C and G advance up to the position indicated with a C' and a G' respectively. Then, optical axes 2 and 3 which connect a camera center and the points C' and G' respectively become different line segments and also the positions on a measuring area face become different points, points (c) and (d), respectively. Thus, by discriminating whether the feature point is the head part of the parallel traveling vehicle or not, the vehicles observed while overlapping can be detected separately.

Description

Detailed Description of the Invention [Industrial Application Field] The present invention relates to an image processing type vehicle sensor that images a vehicle and recognizes the presence or absence of the vehicle, its shape, moving speed, etc. based on the imaging results. Relating to a vehicle separation method.

[Prior Art] Conventionally, as a method for separating images of vehicles running parallel to adjacent lanes in an image processing type vehicle sensor, for example, there is a method described in Paper: 17th Image Engineering Conference 16-13 r
Attempts at Measuring Vehicle Dynamics As shown in JP 295-298, a method has been proposed in which the width of the vehicle to be determined is measured on the image capture screen and the images of the vehicle are separated.

[Problems to be Solved by the Invention] However, in the conventional method, if the measured vehicle width on the imaging screen exceeds a preset value, or if - vehicles are continuously (1) For a wide variety of vehicles, the vehicle width length must be uniquely determined. Can not.

(2) When a moving vehicle is photographed from an angle, the width of the same vehicle differs as shown in (a), (b), and (c) depending on the viewing position, as shown in FIG.

There is a problem like that, and it is necessary to accurately cut out (calculate) the vehicle width.
I couldn't.

For this reason, in the conventional method, when there are vehicles running parallel to each other, there are often two
There was a problem where multiple vehicles were mistakenly recognized as one vehicle.

Therefore, the purpose of the present invention is to eliminate such problems,
The object of the present invention is to propose a vehicle separation method that can reliably recognize the presence of parallel vehicles and separate the vehicles when there is an image showing the presence of parallel vehicles in the captured image.

[Means for Solving the Problems] In order to achieve such an object, the present invention images a moving vehicle to be detected at a first time using an image processing vehicle sensor, and as a result of the imaging, When a feature point indicating the presence of a parallel vehicle exists in the obtained two-dimensional image, the vehicle to be sensed is subsequently imaged at a second time;
Based on the result of the imaging at the first time, the first position coordinates of the feature point in the three-dimensional space at the second time are determined by a monocular camera, assuming that the feature point is a part of the vehicle to be sensed. Calculated using a stereoscopic viewing method, based on the result of imaging at the first time,
Calculating second position coordinates of the feature point in a three-dimensional space at the second time when the feature point is assumed to be a part of the parallel vehicle using a monocular stereoscopic viewing method; Based on the imaging results, the third positional coordinates of the feature point in the three-dimensional space are calculated using a monocular stereoscopic method, and when the third positional coordinates match the first positional coordinates, the feature point is When the feature point is determined to be part of the vehicle to be sensed and the third position coordinates match the second position coordinates, the feature point is determined to be part of the parallel vehicle, thereby separating the parallel vehicle. shall be.

[Function] The present invention focuses on the fact that the position of the vehicle to be sensed in a three-dimensional space can be calculated from the two-dimensional image obtained as a result of imaging, and based on the imaging result at the first time, Assume the position in the three-dimensional space at time 2. At this time,
The feature points representing the characteristics of the parallel vehicle in the imaging result at the first time are divided into two cases: a case where the feature points are assumed to be part of the vehicle to be sensed, and a case where the feature points are assumed to be part of the parallel vehicle. By comparing the calculated assumed position in each case with the position actually obtained from the imaging result at the second time, it is possible to determine whether the above feature point is part of the vehicle to be sensed or part of the parallel vehicle. be identified. As a result, even if the vehicle to be sensed and the vehicle running parallel to each other are close to each other and it is difficult to separate them, it is possible to reliably separate unspecified vehicle types.

[Example] Hereinafter, examples of the present invention will be described in detail with reference to the drawings.

Before explaining the present invention, a vehicle sensor to which the present invention is applied will be explained using FIG. 2.

The vehicle sensor includes a camera section 1 and a control section 2. The camera unit 1 can be set to measure a road with up to four lanes, and is installed on a support 3 so as to photograph a moving vehicle from diagonally above. The camera unit 1 photoelectrically converts a subject image and outputs the imaging result in the form of an image signal.

The control unit 2 performs image analysis processing, which will be described later, on one screen worth of image data (signal) received from the camera unit 1, and determines whether there is a vehicle running in parallel. Next, a processing procedure for vehicle separation on image data obtained from the camera unit 1 will be described.

In this example, a screen in which the vehicle 11 and the vehicle 12 to be determined are running side by side as shown in FIG. 3 is used as an example.

In the screen of Figure 3, a large bus traveling from the top of the screen (upstream on the road) to the bottom of the screen (downstream) in the center lane is the vehicle 11 to be identified, and a part of the vehicle body such as the roof is located on the adjacent center side. It's stuck in the lane. Further, the parallel vehicle 12 running in the adjacent lane is a small vehicle with a part of its body hidden by the adjacent bus.

(a) Extract the image of the vehicle part from the image data sent from the camera unit 1 using a well-known subject separation method.

Specifically, the image data is converted into a binary value (black/white) by comparing the brightness level of each pixel indicated by the image data with a threshold value, and as shown in FIG.
1", a screen is created in which the background portion 22 is represented by a dot "0".

As a result, an image 21 representing the contours of the vehicle 11 to be determined and the parallel vehicle 12 is extracted.

(b) Each feature point a-'-g of image 21 using the conventional method
Detects the position of on the screen.

(c) By detecting sides a-b and sides c-g that are parallel to the width direction of the road and located at the bottom of the vehicle image, the head portion of the vehicle to be determined is detected.

This vehicle head detection method exists as a conventional method.

(d) Determine whether side c-g is a part of the vehicle 11 to be determined or a part (vehicle head) of the parallel vehicle 12 according to the processing procedure described later, and determine the presence or absence of a parallel vehicle. do.

Next, a specific processing procedure for determining parallel vehicles according to the present invention in Section 4 above will be explained.

FIG. 1 shows the positional relationship of the vehicles 11 and 12 shown in FIG. 3 when viewed from the side of the road.

Note that this figure shows two positional relationships between the vehicles 11 and 12: the positional relationship at the first time t, and the positional relationship at the second time t' after a certain period of time (the state in Figure 3 is the positional relationship at the time t'). ) Even if a three-dimensional object is measured with one camera, the position of the object in three-dimensional space can be determined from the camera installation conditions and the positional conditions of the measurement target plane (same as the measurement area in this case). can be found.

The positions of feature points c and g, which appear to be molded on the screen in FIG. 4, on the X-axis of the measurement area plane are points (A) as shown in FIG.
Actual vehicle feature points C and G, which correspond to the feature points on the screen, both exist on the optical axis 1 connecting point (A) and the camera center.

However, in the state at time t', feature points C and G advance to positions indicated by symbols C' and G'', respectively.

At this time, a line segment connecting the camera center and point C' (optical axis 2) and a line segment connecting the camera center and point G' (optical axis 3)
are different line segments, and their positions on the measurement area surface are also different from point (1) and point (1), respectively.

The present invention pays attention to this point, and determines whether feature point C or g of the vehicle image 21 in FIG. 3 is the head portion (part of the bumper) of a parallel vehicle by the following procedure. That is, Process (1) Based on the image data obtained at time t,
Figure 4: Find point C (or point g). Also, find the distance between (a) and (b) in FIG.

Process (2) A line segment (optical axis 1 shown in Figure 1) connecting the point (1) on the X coordinate obtained in Process (1) and the center of the camera is obtained and detected as the head of the vehicle 11 to be identified. Find the y-axis or a line segment parallel to the y-axis that passes through point A.

These two line segments intersect with the optical axis 1 mentioned above. Find the points P and G that point.

Process (3) Next, time t' (a certain period of time has passed since time t)
The moving distance of the target object is determined by calculating the value (point (wa)) on the X coordinate of the vehicle head point A of the vehicle 11 to be determined that was detected last time on the binarized screen, and the time of the aforementioned point 29 point Q is calculated. Assume the values on the X coordinate at t' (point (11) 8 and point (ki)).

Processing (4) The value on the X coordinate of point 0 (or point g) at time t' (point (1) in Figure 1) is calculated by processing (
Check which of the points (force) and (ri) found in 3) are included.

(i) If the value of the above X coordinate is included in the point (force), a part of the head of the parallel vehicle (ii) If the value of the above X coordinate is included in the point (ki),
It is determined that it is part of the vehicle to be determined.

In addition, in process (3), in order to absorb errors due to variations in the height of a part of the bumper or the shape of the front surface of the vehicle depending on the vehicle type, the point (h) and the point (kin) are assumed to be areas of points with a certain range.

[Effects of the Invention] As explained above, according to the present invention, even when a large vehicle such as a bus appears to overlap on the screen with an adjacent vehicle running in parallel, the vehicle is separated into individual vehicles. This allows for more accurate traffic measurement. Furthermore, since it is no longer necessary to take into account the influence of overlapping adjacent vehicles when setting the camera's viewing angle, it is no longer necessary to extend the arm for installing the camera unit over the lane. Therefore, it is also effective in improving aesthetic appearance.

[Brief explanation of drawings]

FIG. 1 is a side view showing a procedure for determining the presence or absence of a parallel vehicle according to an embodiment of the present invention, FIG. 2 is a perspective view showing the arrangement of a vehicle sensor according to an embodiment of the present invention, and FIG. 4 is an explanatory diagram showing the imaging results of the vehicle in FIG. 3; FIG. 5 is a perspective view showing the driving state of the vehicle according to the embodiment of the present invention; FIG. It is an explanatory diagram. DESCRIPTION OF SYMBOLS 1... Camera part, 2... Control part, 3... Support column, 4... Measurement area. Figure 4

Claims (1)

[Claims] A moving vehicle to be sensed is imaged at a first time by an image processing vehicle sensor, and a feature point indicating the presence of a parallel vehicle is included in a two-dimensional image obtained as a result of the imaging. If there is, the vehicle to be sensed is subsequently imaged at a second time, and based on the result of the imaging at the first time, the feature point is assumed to be a part of the vehicle to be sensed. calculate a first position coordinate in a three-dimensional space at the second time of the vehicle using a monocular stereoscopic method, and based on the result of imaging at the first time, identify the feature point as a part of the parallel vehicle. The second feature point of the feature point in the assumed case
Calculate the second position coordinates in the three-dimensional space at the time of , using a monocular stereoscopic viewing method, and calculate the third position coordinates of the feature point in the three-dimensional space using the monocular stereoscopic method based on the result of the imaging at the second time. Calculated using a stereoscopic viewing method, and when the third position coordinates match the first position coordinates, the feature point is determined to be a part of the vehicle to be sensed, and the third position coordinates are the second position coordinates. A method for separating a vehicle in an image processing type vehicle sensor, characterized in that when the characteristic point matches the characteristic point, the feature point is determined to be a part of a parallel vehicle, thereby separating the parallel vehicle.