JPH05312819A - Speed detecting device for moving object - Google Patents

Abstract

PURPOSE:To detect the speed of a moving object such as a vehicle with high precision from an arbitrary point by determining the transfer vector of the moving object after the shiftable article existence region and each optical flow are statistically treated. CONSTITUTION:The focus distance, direction angle, and the turning angle of a TV camera are inputted 402, and at the same time, a digital image of a traveling vehicle is inputted 403. The digital image is treated by arbitrary threshold values, and the vehicle region is separated from other regions and extracted 404. While, the input image is smoothed, differentiation-treated, and the optical flow at each image element is extracted 405 through the least square calculation. The vehicle region and the optical flow are statistically treated 406 on the basis of the direction angle and dimension, and the imcorrect optical flow is discharged. The direction and dimension of the transfer vector are calculated from the left optical flow, and further, a transfer vector is determined 407 in each region. The determined transfer vector is projected 408 on the traveling surface or the vehicle by using the focus distance, direction angle and the turning angle of the camera, and the image element is converted to the actual distance, and the actual speed of the vehicle is calculated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving object speed detecting apparatus suitable for use in a traffic control system for detecting the speed of a traveling vehicle.

[0002]

2. Description of the Related Art Vehicle speed information used in a traffic control system includes a point speed which is an instantaneous speed of the vehicle at a certain point, a traveling speed obtained by dividing a distance traveled by the vehicle by a traveling time, and a vehicle traveling. There is a section speed obtained by dividing the distance by the total time required for it. The speed targeted by the present invention is a point speed of the vehicle.

In a conventional traffic control system, as a device mainly used for measuring a point speed of a vehicle (hereinafter, abbreviated as speed), a device using a vehicle sensor such as a loop coil type and an ultrasonic wave are used. Some use the Doppler effect of.

A speed measuring device using a vehicle detector is a method of calculating a speed by measuring a difference between vehicle passing times of vehicle detectors installed at intervals of several meters. The location of the detector is about 5 m above the road using ultrasonic waves or about 2 m diagonally above the road, and below the road surface using metal sensing.

A measuring device using the Doppler effect is a method of radiating an ultrasonic wave toward a traveling vehicle and measuring a frequency shift amount of a reflected wave due to the Doppler effect to calculate a speed. The installation location of the device is on the roadside of the road where measurement is performed.

[0006]

However, the above-mentioned vehicle speed detecting device, which has been mainly used conventionally, can be applied only in a limited and very narrow range at the installation point because of the structure of the device. It was In addition, when changing the position of the device to measure another point, it was mainly to newly install it by construction or re-install it manually. For this reason, there is a problem that the observer cannot arbitrarily detect the vehicle speed by installing a detector at a position where speed detection is required within a certain range.

There is also a problem that the detection result greatly varies depending on the installation state.

The present invention has been made to solve the above-mentioned problems, and an object thereof is to accurately detect the speed of a moving moving object from an arbitrary point facing the moving surface of the moving object such as a road. The object of the present invention is to provide a moving object velocity detection device that can do the above.

[0009]

In order to achieve the above object, in the present invention, camera information input means for inputting a focal length of a camera for photographing a moving object and a direction angle and a depression angle of an optical axis are set. Region extraction means for extracting a moving object existing region in images captured at different time intervals, and optical flow extraction for extracting optical flows of respective pixels in a plurality of images captured at different times in time. Means, statistical processing means for statistically processing the extracted optical flow with respect to the moving object existing area extracted by the area extracting means and eliminating inaccurate ones, and moving object existing area from the remaining optical flow Moving vector determining means for calculating one of the vectors, and the determined moving vector as the focal length of the camera and the direction angle of the optical axis. And projected onto the moving surface using the angular converts pixel to the actual distance, it was provided with a moving vector projection means for calculating the actual speed of the moving object.

[0010]

The motion vector is an apparent velocity vector on the image generated when the motion of the target moving object is projected on the image, and the optical flow is the same between images that are temporally continuous. Is a vector indicating the movement of the point.

According to the above means, the focal length of the camera, the direction angle and the depression angle of the optical axis are input by the camera information input means, and the presence of a moving object in an image including a moving object such as a traveling vehicle photographed by the camera. The region is extracted by the region extraction means,
Also, the optical flow of each pixel in a plurality of temporally continuous images is extracted. Then, the optical flow is statistically processed with respect to the area extracted by the area extracting means, and inaccurate ones are excluded. Furthermore, determine one vector from the remaining optical flow,
The determined moving vector is projected on the moving surface using the focal length of the camera, the direction angle of the optical axis, and the depression angle, and the actual velocity of the moving object is calculated by converting the pixels into the actual distance.

As a result, it is possible to accurately detect the speed of a moving object such as a vehicle running from an arbitrary point facing the road.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A moving object velocity detecting apparatus according to the present invention will be described in detail below with reference to the drawings.

FIG. 1 is an overall configuration diagram showing an embodiment of a moving object speed detecting device according to the present invention. A television camera 1 for picking up an image of a moving object such as a vehicle traveling on a road, and a direction angle of the television camera 1. A sensor 2 for detecting an image, an analog / digital conversion circuit (A / D) 3 for converting an image captured by the television camera 1 into a digital image, an information processing device 4 including a computer, a display for displaying image data, information data, etc. A device (CRT) 5, an auxiliary storage device 6 including a magnetic disk, an optical disk, etc. is provided. in this case,
The information processing device 4 includes a CPU 40 and a memory 41.

When the information processing apparatus 4 composed of the CPU 40 and the memory 41 is functionally disassembled, as shown in FIG. 2, a camera information input section 402, an image input section 403,
Region extraction unit 404, optical flow extraction unit 405,
It includes a statistical processing unit 406, a movement vector determination unit 407, a movement vector projection unit 408, and a data bus 409.

Next, the operation of the moving object speed detecting apparatus of the present embodiment will be described by taking the case of detecting the speed of a vehicle traveling on a road as an example.

In FIGS. 1 and 2, a television camera 1 is installed at an arbitrary point facing a vehicle traveling on a road, and the traveling vehicle is imaged by the delevi camera 1. The captured image is converted into a digital image by the analog / digital conversion circuit (A / D) 3 and input to the image input unit 403. At the same time, the sensor 2 inputs the focal length, the direction angle, and the depression angle of the delevi camera 1 to the camera information input unit 402.

The digital image of the traveling vehicle input to the image input unit 403 is processed by the region extraction unit 404 with an arbitrary threshold value and divided into a vehicle region and other regions.

On the other hand, in the optical flow extraction unit 405, as shown in the flow chart of FIG. 3, smoothing processing is performed on the input image by an arbitrary window function (step 7).
Perform 1). Next, a differential process (step 72) is performed on the smoothed image to obtain horizontal and vertical density differential values (I _x , I _y ) in each pixel. Further, the difference value (I _t ) of the densities is obtained from the images of the two frames which are temporally continuous. Constructed using these values the equation _{I x · u + I y ·} v + I t = 0 u: horizontal component of the movement vector v: by a least squares calculation from the vertical component of the motion vector (step 73), an optical in each pixel Find the flow (u, v). This method is called the gradient method.

Further, in the statistical processing unit 406, from the results obtained from the area extracting unit 404 and the optical flow extracting unit 405, as shown in the flowchart of FIG. By the two statistical processing called the statistical processing (step 82) based on
Eliminate optical flows that are assumed to be erroneous within a region. Below, these two statistical processes are shown in FIG.
This will be described with reference to FIGS. 7 and 8.

(X, y) in FIG. 6 represents the coordinates of the image plane. The region R shown here is the region extraction unit 40.
It is assumed that the area is one target moving object extracted in 4. W _i (u _i , v _i ) is a vector indicating an optical flow and indicates one existing in the region R. The subscript i used here indicates the i-th optical flow in the area.

First, in the statistical processing based on the direction angle of the optical flow (step 81), the direction angle θ _i of each W _i is obtained from the input as shown in FIG.

[0023] ^{_{θ i = tan ~ 1 (v}} i / u i) (0 ≦ θ i <360 °) using the value of the theta _i, the direction angle on the horizontal axis as shown in FIG. 7,
Create a histogram with the vertical axis representing the number of optical flows. Optical flows not included in an arbitrary range D1 centered on the peak position P1 of this histogram are excluded.

Next, in the statistical processing based on the magnitude of the optical flow (step 82), each W remaining in the statistical processing based on the direction angle (step 81) is further added.
_i of size | W _i | a seek.

| W _i | = √ (u _i ² + v _i ² ) Using the value of | W _i |, the horizontal axis represents the size and the vertical axis represents the number of optical flows as shown in FIG. Create a histogram. Peak position P2 of this histogram
The optical flow not included in the arbitrary range D2 centered on is further excluded.

The above processing is performed in one area,
Eliminate optical flows that are assumed to be erroneous.

As shown in the flowchart of FIG. 5, the movement vector determination unit 407 averages the optical flows remaining in the statistical processing unit 406 in the vector calculation processing (step 91) to determine the direction and magnitude of the movement vector. Is calculated and the center of gravity of the starting point of the optical flow is calculated in the starting point calculation process (step 92) to determine the starting point of the movement vector. As a result, one movement vector is determined for each area obtained by the area extraction unit.

In the movement vector projection unit 408, first, the ratio of the number of pixels in the horizontal direction of the digital image obtained in the image input unit 403 and the ratio of the actual distance to the pixel is calculated from the effective horizontal image pickup range of the camera 1.

D = R / C D: Real distance value for one pixel R: Actual distance in the horizontal direction of the camera C: Number of pixels in the horizontal direction of the image In the present invention, the road on which the vehicle runs is a plane. Assuming that the motion vector on the image obtained by the motion vector determination unit 407 is projected on the plane and the value of the ratio is used, the motion vector in the real space is obtained.

The notation of equations and variables used will be described below. First, the overall coordinate setting will be described with reference to FIG.

In FIG. 9, (X, Y, Z) are coordinates in a three-dimensional space with the focal point of the camera 1 as the origin O, and the XZ plane is set parallel to the road plane.

Further, (X, Y) is a coordinate of the image plane, and the y axis is set so as to exist on the YZ plane.

H is represented by a distance obtained by subtracting half the average vehicle height of the vehicle from the distance from the origin O to the road plane.

The plane P is a projection plane parallel to the road plane, and is set at a distance from the origin O to H along the Y axis.

F is the focal length of the camera, the direction of which is perpendicular to the image plane and passes through the image origin.

Q ₀ is a point on the image indicating the starting point of the movement vector obtained by the movement vector determination unit 407. Further, q ₁ is a point on the image after moving from q ₀ by the movement vector.

Q ₀ and Q ₁ are points at which q ₀ and q ₁ are projected on the projection plane, and the vector W thereof is a motion vector in the real space.

FIG. 10 shows FIG. 9 from the X-axis direction, and θ shows the depression angle from the Z-axis to the optical axis direction of the camera.

FIG. 11 is an example of the image in FIG.
w represents a movement vector from q ₀ to q ₁ , u is an x-direction component of w, and v is a y-direction component of w.

As described above, under the settings described with reference to FIGS. 9, 10 and 11, the following equation is derived.

Y ₀ = Y ₁ = -H X ₀ = DHx ₀ / (fsin θ-Dy ₀ cos θ) Z ₀ = H (Dy ₀ sin θ + fcos θ) / (fsin θ
−Dy ₀ cos θ) X ₁ = DHx ₁ / (f sin θ −Dy ₁ cos θ) Z ₁ = H (Dy ₁ sin θ + f cos θ) / (f sin θ
−Dy ₁ cos θ) U = X ₁ −X ₀ , V = Z ₁ −Z ₀ (x ₀ , y ₀ ): image coordinate of point q ₀ (x ₁ , y ₁ ): image coordinate of point q ₁ (x ₁ , y ₁ ) = (x ₀ , y ₀ ) + (u, v) (X ₀ , Y ₀ , Z ₀ ): Three-dimensional space coordinates of point Q ₀ (X ₁ , Y ₁ , Z ₁ ): Point Three-dimensional space coordinates (U, V) of Q ₁ : X-direction and Z-direction components of vector W: Value of actual distance with respect to one pixel: Using such an expression, the value of W, which is a motion vector in real space, is real. Calculated by distance.

The _value (X ₀ , Y ₀ , Z ₀ ) obtained here indicates the position with reference to the camera 1 of the vehicle to be detected, and W is the distance between the images of two consecutive frames required for the detection. Is the displacement of the position at. Therefore, if the time interval of the image of one frame is T, the speed S can be obtained by S = ‖W‖ / T.

As can be seen from the above description, according to the present embodiment, the traveling vehicle is subjected to the statistical processing as shown in FIG. 4 using the area obtained by the processing of the area extracting unit 404 shown in FIG. Since the movement vector of the vehicle is determined, it is not necessary to extract the area with high accuracy, and it is sufficient to extract the range including the area to be detected on the image (area of a single vehicle or area of multiple vehicle groups). .. Therefore, the processing load due to the region extraction is small, and the result is not greatly affected by the movement vector determination. Furthermore, the movement vector determination unit 4
In 07, since the statistical processing is performed using the direction angle of each optical flow, it is less affected by the flow including the calculation error, and more stable detection can be performed.

Further, since the video camera 1 is used for detection, it is possible to detect the speed in a wider range as compared with the conventional apparatus, and it is also possible to simultaneously detect a plurality of vehicles existing in the image. It is possible.

Further, since the values of the depression angle of the camera are sequentially input from the camera information input unit 402 and the velocity is calculated by the movement vector projection unit 408 accordingly, it is extremely easy to change the velocity detection point. Become.

Although the present invention has been specifically described based on the embodiments, it is needless to say that the present invention is not limited to the above embodiments and can be variously modified without departing from the scope of the invention. .. For example, although a vehicle traveling on a road is taken as an example of a moving object, a human being or other objects can be detected in the same manner.

[0047]

As described above, according to the present invention,
Camera information input means for inputting a focal length of a camera for shooting a moving object, a direction angle and a depression angle of an optical axis, and an area extracting means for extracting a moving object existing area in an image taken at a set time interval, Optical flow extraction means for extracting the optical flow of each pixel in a plurality of images captured at different times and temporally continuous, and the extracted optical flow for the moving object existing area extracted by the area extraction means Statistical processing means for statistically processing and eliminating inaccurate ones, movement vector determination means for calculating one vector for the moving object existing region from the remaining optical flow, and the determined movement vector for the focus of the camera The actual velocity of the moving object is calculated by projecting it onto the moving surface using the distance and the direction angle of the optical axis and the depression angle to convert the pixels into the actual distance. Movement so was and a vector projection means for the speed of a moving object such as a vehicle in motion from any point facing the moving surface such as a road can be detected accurately.

Further, there is an effect that the speeds of a plurality of moving objects can be detected at the same time.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram showing an embodiment of a moving object velocity detection apparatus according to the present invention.

FIG. 2 is a functional block diagram of the information processing device in FIG.

FIG. 3 is a flowchart showing a processing procedure in an optical flow extraction unit.

FIG. 4 is a flowchart showing a processing procedure in a statistical processing unit.

FIG. 5 is a flowchart showing a procedure of processing in a movement vector determination unit.

FIG. 6 is an explanatory diagram for explaining statistical processing.

FIG. 7 is an explanatory diagram for explaining statistical processing based on direction angles.

FIG. 8 is an explanatory diagram illustrating a statistical process based on a size.

FIG. 9 is an explanatory diagram for explaining coordinate setting and variable notation.

FIG. 10 is an explanatory diagram for explaining coordinate setting and variable notation.

FIG. 11 is an explanatory diagram for explaining coordinate setting and variable notation.

[Explanation of symbols]

1 ... Television camera, 2 ... Sensor, 3 ... Analog / digital conversion circuit, 4 ... Information processing device, 5 ... Display device (CR
T), 6 ... Auxiliary storage device, 401 ... CPU, 402 ... Camera information input unit, 403 ... Image input unit, 404 ... Region extraction unit, 405 ... Optical flow extraction unit, 406 ... Statistical processing unit, 407 ... Movement vector determination Part, 408 ... Movement vector projection part, 409 ... Data bus.

Claims (1)

[Claims] 1. A moving object speed detecting device for detecting a moving speed of a moving object from an image obtained by shooting the moving object by image processing, wherein a focal length of a camera for shooting the moving object and a direction angle and a depression angle of an optical axis are input. Camera information inputting means, area extracting means for extracting a moving object existing area in images taken at set time intervals, and pixels in a plurality of images taken at different times consecutively in time An optical flow extracting means for extracting the optical flow of, and a statistical processing means for statistically processing the extracted optical flow with respect to the moving object existing area extracted by the area extracting means and eliminating inaccurate ones, and Vector determining means for calculating one vector for the moving object existing region from the optical flow, and the determined moving vector. A moving vector projecting means for projecting the tor on the moving surface of the moving object by using the focal length of the camera, the direction angle of the optical axis, and the depression angle to convert the pixels into the actual distance and calculate the actual velocity of the moving object. A moving object velocity detecting device characterized by: