JPH05112191A - Running vehicle detection device - Google Patents

Abstract

PURPOSE:To stabilize separation/detection by detecting changes due to the elapse of time concerning picture element data only in the detection slit part along the previously set direction of running to accumulate them, performing the line segment detection of the changing parts to judge to which background and to which vehicle the travel amount belongs, superimposing separate picture images per vehicle, and separating them per vehicle. CONSTITUTION:For example, the position close to a bumper of ordinary vehicle is set in a detection slit, and slit image is recorded in an image memory 1 by a TV camera according to the elapse of time. Picture elements in the detection slit part only which are set previously by a picture element read circuit 2 are read and accumulated in a storage memory 3 according to the elapse of time. Timely changes of data which is read by a concentration detection circuit 4 are detected, and image on the background only is updated by a storage/update circuit 5. Also, the vehicle region and travel amount in the detected and changed part are calculated by a line segment detection circuit 6. A differential image in which the background image to be input from a subtraction circuit 8 based on the result is subtracted is shifted in a shift circuit 10 and superimposed in a vehicle storage circuit per vehicle to store it.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a traveling vehicle detection apparatus for automatically extracting and detecting only a portion corresponding to a vehicle from a traveling image of each vehicle.

[0002]

2. Description of the Related Art When processing image data obtained by observing a road from the side or diagonally above the road, the image of the vehicle to be detected and the background image and the vehicle to be detected are the same or reverse. The vehicle traveling in the traveling direction is stored as an image at the same time while being visible and hidden. In general, background image data changes due to factors such as weather over time, so in the past, the background image was automatically updated by checking the detection of changes over time in the image data of the same area for the entire image area. Was going on.

In this case, since the whole area is processed, a large amount of calculation is required for change detection. Further, when a plurality of vehicles are present in the image at the same time, it is not possible to detect where the image area is the background and which vehicle is each vehicle. Alternatively, vehicle detection is performed by a method such as taking the logical sum of images in which the passage of time continues only for the (positive) portion.

With such a conventional method, it is difficult to determine / separate / detect a plurality of visible / hidden vehicles from different vehicle areas without detecting changes in the entire screen area. Met.

[0005]

In the case of processing background brightness changes due to factors such as weather changes with the passage of time and image data in which a plurality of vehicles appear and disappear in an image at the same time, the conventional image area is used. Large-scale calculations were required because the changes in the entire area were examined over time, and it was difficult to automatically and stably determine the background and individual vehicle areas for separation / detection. ..

It is an object of the present invention to stably separate / detect a background or individual vehicles even when the background brightness changes due to weather changes or a plurality of vehicles appear and disappear in an image. I am trying.

[0007]

According to the present invention, in order to solve the problems of the prior art, preset in the image area,
The change over time is detected only for the pixel data of only the detection slit portion along the vehicle traveling direction. Furthermore, only the data of the detection slits are accumulated over time, line segments are detected for the changed parts, and which part of the image area is the background or which vehicle and the amount of movement on the image are determined. Each vehicle is superposed as a separate image, and each vehicle can be separated and extracted.

That is, a means for sequentially accumulating pixel data of a preset slit portion of an image frame of a photographed image on a memory for a slit image, a means for detecting a line segment for the slit image, and a means for detecting the line segment are obtained. A circuit for obtaining a difference image with respect to the background image recorded in advance for the obtained vehicle existing area and a difference image for the amount of vehicle movement obtained by line segment detection are arranged to be superimposed.

[0009]

In the conventional case, the vehicle which is a moving object is detected by taking the logical sum of the difference image with the background for every pixel value for all the pixel values of the photographed image, but the calculation amount is reduced in the present invention. In order to do so, paying attention to the fact that the moving direction of the vehicle is almost constant, the vehicle detection processing is performed by calculating only the portion (slit portion) where the vehicle passes in the captured image (image frame). I have to.

According to the present invention, the detection slit is preset in the digital image of the traveling vehicle along the traveling direction of the vehicle. When shooting from diagonally above, multiple lines are set so that the vehicle travels within the slit. Although various patterns can be considered as the setting of the detection slit, as a simple example, a TV camera is installed on the side of the vehicle traveling direction to take an image, and the scanning line corresponding to the position of a bumper of a normal car is set. The detection slit may be used as it is.

[0011]

1 is an embodiment of the present invention, in which 1 is an image memory for storing a digital image input from a television camera, and 2 is a preset one or a plurality of detection slit portions. A pixel readout circuit, 3 is a storage memory for accumulating the detection slit data read out by the readout circuit 2 with the passage of time, 4 is a change detection circuit of data read out from the same image area over time, and 5 is a background A storage / update circuit for storing / updating a slit image consisting of only a line segment detection for a changed portion detected by the change detection circuit 4 to detect a vehicle region and a line segment detection for calculating the movement amount of the vehicle. Circuit, 7
A background image memory that stores / updates each image area determined to be a background, 8 a subtraction circuit that subtracts the background image from the input image, 9 a difference image memory that stores the subtraction result, 1
Reference numeral 0 is a shift circuit for taking out individual vehicle areas determined by the change detection and line segment detection circuit 6 and shifting them by the amount of vehicle movement, and 11 is a vehicle memory for superimposing and storing these vehicle area images for each vehicle. It represents a circuit.

As a simple example, a case where a television camera is installed on the side of the vehicle traveling direction and the vehicle is observed so that its scanning direction is parallel to the vehicle traveling direction will be described. With the passage of time, the vehicle starts moving from the right or left of the image to the image area according to the traveling direction, passes through the center of the image, and disappears in the opposite direction.

The scanning line at the position of the bumper of an ordinary car is used as a detection slit as it is, and this one-dimensional data is accumulated with the passage of time, and the scanning line direction in the original image area is taken as the sub-scanning direction to determine the passage of time. The data is stored as two-dimensional data in the main scanning direction, and this image will be referred to as a spatiotemporal image hereinafter. Considering the change over time in the same position of the original image, that is, the same position of the scanning line set in the slit, that is, when the change of the data in the direction of the main scan in the spatiotemporal image is considered, Although there is almost no change in the background, the change occurs in the vehicle part as the corresponding point moves in the detection slit direction as the vehicle travels. Therefore, it is possible to detect, as a traveling vehicle, a pixel having a large change with time at the same position of the scanning line set in the slit.

When the thus-detected change pixels are viewed in the spatiotemporal image, a trajectory on a line corresponding to the running form such as speed, acceleration, and moving direction is drawn. FIG. 2 shows the information that appears in the spatiotemporal image.

Reference numerals 20-1, 20-2, 20- in the figure
3, ... are multiple images taken at the same position by the camera,
Reference numerals 21-1, 21-2, ... Represent images on the scanning lines set in the slits, and 22 is a spatiotemporal image. The spatiotemporal image 22 is the image 21-1 on the image 20-1 and the image 20-
2-2 on image 2 and 21-3 on image 20-3
, And ... are accumulated in chronological order. As a result, on the spatiotemporal image 22, a right-down trajectory appears for a vehicle traveling from left to right, a left-down trajectory appears for a vehicle traveling from right to left, and a vehicle traveling from left to right appears. When accelerating, the slope of the trail to the right gradually changes.

Therefore, if line segment detection is performed on the trajectory in the spatiotemporal image, it is possible to determine which area of the image at each time is the background and which area the vehicle is in, and the pixel movement amount in the traveling direction. Even if a plurality of vehicles are present in the image area at the same time, it is possible to detect the number of vehicles by dividing the slit direction and performing line segment extraction by a known processing technique.

The background area is B, the area including the background of each of i vehicles appearing in the image at the same time and the background is F ₀ ,
Assuming F ₁ , ..., F _n , regions of the same kind should line up in the traveling direction, that is, in the direction orthogonal to the detection slit. In the case where vehicles traveling in opposite directions pass each other in the image area being photographed, an example immediately before they overlap each other is shown in FIG. Therefore, the background area in the original image can be known as it is from the area regarded as the background in the detection slit. Therefore,
It is possible to automatically update the background image stored so far in the area regarded as the background.

This background image can be subtracted from the original image to create a so-called difference image. Image regions where only individual vehicles exist can also be determined by the above spatiotemporal image processing, and individual vehicles can be separated by superimposing only the same vehicle parts by shifting in the traveling direction by the amount of movement with time. / It can be detected.

Brightness or color information of the image data is stored as data represented by two-dimensional coordinates having a vehicle running direction (that is, a scanning line direction of the image data) x and y orthogonal to the x direction and the observation time t. Then, change this value to f (x,
y, t), and they are accumulated sequentially as time passes. Here, the image data area to be stored at each time is defined as 0 ≦ x ≦ (X−1) and 0 ≦ y ≦ (Y−1).

When the height near the bumper of an ordinary vehicle is y = y _slit in the image, the _slit is set to y = y _slit . In this case, f (0, y _slit , t) and f ((X-1), as slit images at each time,
One scan line data having y _slit , t) at both ends is read.
In the case of this example, the readout circuit 2 determines that the y coordinate of the pixel is y at the time when the image data is stored in the image memory 1 at each time.
_It is determined whether or not it matches the _slit, and only when it matches, the data is output to the storage memory 3. The data read from the slit is stored in the storage memory 3 as data expressed by the order u according to the order of reading along the slit and the time t, and this value is represented by s (u, t), Sequentially stored as time passes.

In this example, the _slit is set to y = y _slit , and the pixel f (0, y _slit , t) to the pixel f ((X-1),
Since y _slit , t) are sequentially read, u≡x. The data read from the slit will be referred to as a slit image below, considering it as two-dimensional data having u and t as two axes.

Regarding the t-axis direction of the slit image, pixels having the same coordinates are arranged in order in the original image area as time passes. Therefore, if the background in which the pixel does not move is memorized, the change in the t direction is small because there is only a gradual change due to a weather change or the like. This value is sb (u). If the vehicle crosses the image area over time,
Vehicle parts appeared in the pixel at a certain time t = t _in, when again moves the different background to the pixel at t = t _out, the value of t _in ≦ t ≦ t _out is large and sb (u) Differences occur. FIG. 5 shows an example of changes in the slit image.

Therefore, the change detection circuit 4 can check the change in the slit image in the t-axis direction to determine whether each area in the slit image is the background or the vehicle at time t. Each pixel s (u, t) of the slit image determined as the background is stored / updated by the circuit 5 as a background slit image. At the same time, in the original image area, all pixels in the direction orthogonal to the slit judged to be the background, in the example, since the slit is horizontal, are judged to be the background and all the pixels are stored as the background image in the memory. Memory in circuit 7 /
Update.

When the appearance of a difference is recognized by the change detection in the slit image, the vehicle crosses the image area. Therefore, as shown in FIG. 5, a line trajectory corresponding to the vehicle speed is drawn. If the velocity is constant, the trajectory is a straight line. If the vehicle is accelerating, the curve is curved in the positive direction over time, and if the vehicle is decelerated, the curve is curved in the negative direction over time. Therefore, if the line segment is detected, the speed of the vehicle on the image, that is, the amount of moving pixels on the image area at a unit time change can be detected.

A known method may be used to detect the line segment in the slit image. For example, linear regression by the least squares error method is performed on the adjacent slit areas for several frames.

When two or more vehicles appearing and disappearing in the stored image when the vehicle traveling in the opposite direction just intersects in the image area being photographed, the line segment detection circuit 6 determines the number of vehicles. Since only the number of lines can be detected,
The area where only one vehicle appears can be determined, and the background only or the area where two or more vehicles appear can be deleted. FIG. 6 shows an example of determination and deletion. At time t, the background image is subtracted from the image obtained by the television camera by the circuit 8, the difference image is stored in the memory 9, and the circuit 10 shifts and superimposes it in the traveling direction according to the moving amount of each vehicle. The same part can be matched and overlapped. The zone surrounded by the dotted line in the figure is the erased zone. FIG. 7 shows an example of superposition using the movement amount of each vehicle detected from the slit image. By the method described above, the images of the individual vehicles can be obtained separately and stored in the vehicle storage circuit 11.

As a simple example, the case of photographing from the side has been described, but the present invention effectively works even when photographing from an obliquely upper position. For example, in the case of FIG. 8, two or more slits may be set along the traveling direction and the data read from that area may be combined.

That is, as shown in FIG. 9, images 23-1, 23-2 and 23 corresponding to a plurality of scanning lines on the image 20-i.
-3 is extracted and the images 23-1, 23-2, 23-3
Can be arranged on one horizontal scanning line to form the spatiotemporal image 22.

[0029]

As described above, according to the present invention, by detecting the change with time of only the preset detection slit portion in the entire image area and performing the line segment detection for the changed portion, Detecting whether each area of the image data is the background or the vehicle and changing the brightness due to weather changes, or even when multiple vehicles are visible or hidden in the image at the same time, the size is smaller than before. It is possible to provide a traveling vehicle detection device that can operate only by calculation.

[Brief description of drawings]

FIG. 1 shows an embodiment of the present invention.

FIG. 2 is a diagram illustrating information that appears in a spatiotemporal image.

FIG. 3 shows a state before and after a vehicle overlaps.

FIG. 4 shows a state in which vehicles are overlapped.

FIG. 5 is a diagram illustrating detection of a background by detecting change in a slit image.

FIG. 6 is a diagram for explaining determination of a vehicle appearance region and erasing of an overlapping region.

FIG. 7 is a diagram illustrating individual superposition of vehicle images using a vehicle movement amount.

FIG. 8 shows an example in which a plurality of slits are set.

FIG. 9 shows an example of shooting from diagonally above.

[Explanation of symbols]

 1 Image Memory 2 Pixel Readout Circuit 3 Storage Memory 4 Change Detection Circuit 5 Storage / Update Circuit 6 Line Segment Detection Circuit 7 Background Image Memory 8 Subtraction Circuit 9 Difference Image Memory 10 Shift Circuit 11 Vehicle Storage Circuit 20 Image 21 Image on Scan Line 22 Spatiotemporal image

Claims (1)

[Claims] 1. An image memory circuit for storing a digital image inputted from a television camera, and one or a plurality of image memory circuits preset in an image area stored in the image memory along a vehicle traveling direction on the image. A circuit that reads out the pixel values of only the detection slit part of the book, a memory circuit that sequentially accumulates the pixel data at the same time read out by this readout circuit, and the same image area of the data inside the memory circuit A circuit that detects changes in the output data over time, and a memory that stores / updates a slit image composed of only the background based on the output of this change detection circuit.
An update circuit, a circuit for detecting a line segment in the space detected by the change detection circuit in a space having coordinate axes orthogonal to the read image area coordinates and time, and a background by the change detection and line segment detection circuits. The background image memory circuit that stores / updates each image area corresponding to the slit inside area and the circuit that calculates and outputs the difference between the background image memory circuit and the image memory circuit that stores the input image, and the difference image is stored. A circuit for obtaining a differential image based on the vehicle area on the image and the vehicle movement amount on the image corresponding to the slit determined to be the vehicle area by the change detection and line segment detection circuit, and a circuit for superimposing the difference image And a traveling vehicle detection device.