JPH0652311A - Image processing method - Google Patents

Abstract

PURPOSE:To surely extract a moving object even at any place where the environment such as lightness is extremely changed. CONSTITUTION:This method is composed of a first step to recognize sampling image data in an area, where absolute value difference data are formed by picture elements less than a prescribed threshold value, as still picture data by leading out the absolute value difference data of a luminance value for each picture element between n-th ((n) is a positive integer) sampling image data and (n+1)th sampling image data, second step to recognize the sampling image data in the still picture area and an area formed by the recognized picture elements as background image data continuously corresponding to each picture element of a motion detection picture with the repetition of the first step, and third step to identify an area formed by picture elements more than the prescribed threshold value as the moving body by leading out the absolute value difference data of a luminance value for each picture element between the background image data and the sampling image data after the number of picture elements, which are not recognized as the background image data of the objective area, is less than the prescribed number of picture elements.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method for sampling image data of a target area at a predetermined cycle and identifying a moving object existing in the target area based on the image data.

[0002]

2. Description of the Related Art Generally, moving image processing is performed by using difference processing between sampled image data, but typically, there are the following two methods. In the first method, as shown in FIG. 6, a reference background image is captured in advance in the absence of a moving object, and the brightness value of each pixel of the background image data and the current input image data is acquired. Of the absolute value difference data is identified, and a region formed by pixels in which the obtained absolute value difference data is larger than a predetermined threshold value is identified as a moving body. The second method is, as shown in FIG. 7, deriving absolute value difference data of the brightness value for each pixel between the current input image data and the input image data one frame before, and obtaining the obtained absolute value difference. This is a method of identifying a region formed by pixels whose data is larger than a predetermined threshold value as a moving body.

[0003]

The first method described above is
It has the advantage that it is easy to process and can be identified even when the moving object is stationary in the middle, but it has the disadvantage that if the background brightness changes to a certain extent, accurate identification is difficult unless the background image is switched each time. In addition, there is a drawback that the background image data cannot be input if any moving object is always present in the target area. It could only be used in places where it could be secured, and it was almost impossible to apply it in the open air, where the environment changed significantly. Further, the second method is not affected by environmental changes such as brightness, but it cannot be accurately detected when the moving body moves at high speed or extremely low speed within the sampling period of the image data. There was a drawback. For example, when the moving distance of the moving body within the sampling period is extremely short, the moving bodies overlap between images to be compared and the contour cannot be accurately detected, and when the moving distance is extremely large, a single movement is performed. Even in the case of a body, a plurality of parts recognized as a moving body will be detected. An object of the present invention is to eliminate the above-mentioned conventional drawbacks.

[0004]

To achieve this object, the image processing method according to the present invention is characterized in that each pixel between the nth sampling image data (n is an integer) and the (n + 1) th sampling image data is A first step of deriving absolute value difference data of luminance values and recognizing sampling image data of a region formed by pixels in which the absolute value difference data is smaller than a predetermined threshold value as still image data; With the repetition of one step, the second step of recognizing the sampling image data of the area formed by the pixels recognized as the still image data continuously m times as the background image data and the background image data of the target area After the number of unrecognized pixels becomes smaller than a predetermined number of pixels, the brightness of each pixel between the background image data and the sampled image data is increased. Derive the absolute difference value data, there is to the said absolute value difference data is composed of the third step of identifying a region formed by the pixel to be larger than a predetermined threshold as the moving body. In the above-mentioned configuration, it is preferable that the background image data recognized as the background image data in the second step is the latest sampled image data.

[0005]

In the first step, an area formed by pixels having a value larger than a predetermined threshold value with respect to the absolute value difference data of the brightness value of each pixel between the image data sampled close to each other in time is set. While recognizing as a change area due to the movement of the moving body, an area formed of pixels smaller than a predetermined threshold is recognized as a candidate area that can be a background image, and the sampling image data of the candidate area is recognized as still image data. . In the second step, as the first step is repeated, a region formed by the pixels recognized as the still image data, that is, a candidate region that can be a background image is sampled when the region is continuously recognized m times. By recognizing the image data as the background image data, the noise accompanying the movement of the moving body is canceled. In the third step, by repeating the first and second steps, after the number of pixels not recognized as the background image data of the target area becomes smaller than a predetermined number of pixels, the background image data and the sampling image The absolute value difference data of the brightness value for each pixel between the data is derived, and the area formed by the pixels in which the absolute value difference data is larger than a predetermined threshold value is identified as a moving body. At this time, if the latest sampled image data is adopted as the background image data recognized as the background image data in the second step, even if the background brightness gradually changes, the background image data will surely move. Can be distinguished from the body.

[0006]

According to the present invention, the background image data as the reference image used for comparison for identifying the moving object from the input image data is updated each time the input is made based on the past input image data. An image processing method capable of automatically and reliably creating a background image even when a moving body is always present in the input image, and further reliably extracting the contour of the moving body regardless of the moving speed of the moving body. Is now available. Further, if the latest sampled image data is adopted as the background image data recognized as the background image data in the second step, even if the background brightness gradually changes, it is possible to avoid environmental changes. It has become possible to provide an image processing method that can flexibly deal with the background image after the change automatically and surely.

[0007]

EXAMPLES Examples will be described below. As shown in FIG. 1, the image data of the target area photographed by the image pickup means 1 is input to a computer as the image processing means 2,
The image processing system is configured so that the data in which the moving body identified by the image processing on the computer is represented as a binary image is output to the CRT as the display unit 3. Such an image processing system is configured for the purpose of extracting various kinds of information from image data taken by the image pickup means 1 and automatically performing some processing based on the information. Is useful as a wide range of sensing devices for achieving specific purposes.

The image processing means 2 is necessary for processing, a plurality of screens as a storage means 2a composed of a plurality of semiconductor memories described below, a calculation means 2b for calculating each pixel between the screens. Data storage means 2c for storing various data (a plurality of threshold values used for binarization processing, sampling time data, etc.).

The storage means 2a is composed of a plurality of screens as a unit area for storing all the image data of one frame input from the image pickup means 1 for each pixel, and an input screen for storing the current image data, A motion detection screen for storing data representing the behavior of the moving body between frames, a reference screen for storing image data representing the background in the frame, a moving body extraction screen for detecting the moving body, and the calculation means 2b. It is composed of a plurality of work screens used for the calculation, and finally, the absolute value difference data of the brightness value for each pixel of the reference screen and the input screen is derived, and the absolute value difference data has a predetermined value. A region formed by pixels larger than a threshold is identified as a moving body, data binarized by the threshold is stored in the moving body extraction screen, and the data stored in the moving body extraction screen is stored in the CRT. And outputs.

An image processing method by the image processing means 2 will be described below. As shown in FIG. 2 to FIG. 5, a video input screen for inputting sampling image data at every Δt time from time T = 1, a motion detection screen showing the motion of the moving body detected by the image processing, and a reference representing the background image. screen,
The following calculation is performed by the calculating means 2b on the moving body extraction screen representing the moving body. First, all pixels of the motion detection screen are initially set to 00H (T = 1 in FIG. 2), and the image data obtained by the image pickup means 1 is subjected to a predetermined cycle (time Δ).
Sampling at t intervals) and storing in the input screen. (At that time, the previous image data stored in the input screen is saved in the work screen A.)

The first step will be described. The absolute value difference data of the luminance value of the pixel is derived for each pixel from the previous sampled image data and the current sampled image data. Then, the absolute value difference data is binarized with a predetermined threshold value and saved in the work screen B (not shown). That is, if the absolute value difference data is large, it is considered that some change has occurred in the pixel, and the work screen B
An area formed by pixels having a value larger than the predetermined threshold value is set to 00H and recognized as a moving image area. On the other hand, if the absolute value difference data is small, it is considered that the pixel has not changed at all, and the area formed by the pixels on the work screen B smaller than the predetermined threshold value is set to 0.
It is set to FFH and recognized as a still image area, and the sampled image data in that area is recognized as still image data. An 8-bit counter is provided for each pixel of the motion detection screen to increment pixels in a region corresponding to a region recognized as a still image region and pixels in a region corresponding to a region recognized as a motion image region. Is set to 00H. That is, each pixel of the motion detection screen initially set to 00H becomes an 8-bit counter.

The second step will be described. With the repetition of the first step described above, the counter value of each pixel continuously recognized as the still image area is set to each pixel of the motion detection screen (it becomes 00H each time the motion area is recognized). ), A predetermined count value (here, 03H
The sampling image data of the area corresponding to the larger pixel is recognized as the background image data, and the background image data is saved in the reference screen (in FIG. 2, T = 3 and thereafter, for example, T = 3. , The background image data means the part other than the black part that represents the figure).

The third step will be described. When the number of pixels that are not recognized as the background image data of the target area becomes smaller than a predetermined number of pixels (reference screen in FIG. 3, T = 6), the moving body can be extracted thereafter. That is, the absolute value difference data of the brightness value for each pixel between the background image data and the latest sampling image data is derived, and the area where the absolute value difference data is larger than a predetermined threshold is the area of the moving body. It recognizes and binarizes it with the threshold value, and saves it in the moving object extraction screen. The data saved on the moving object extraction screen is output to the CRT.

After that, when some motion is detected on the motion detection screen and the counter value of the pixel in that area is reset to 00H, from then on, until the value becomes new to 03H among those counters. Motion detection is performed using the conventional background image data, and when it becomes 03H, the latest sampling image data of the area corresponding to the pixel is updated as new background image data.

Another embodiment will be described below. In the previous embodiment, it is explained that the sampling image data of the area corresponding to the pixel in which the counter value for each pixel of the motion detection screen becomes larger than 3 in the second step is recognized as the background image data. Is not limited to any positive integer m
Can be set to. Also, although it has been described that the background image data is saved in the reference screen, the background image data recognized as the background image data is the latest sampled image data so that the brightness of the target area is constant. Can flexibly respond to changes in the environment. That is, T = 6 in FIG.
After the reference screen of (1), the area composed of pixels having a counter value of 03H or more on the motion detection screen is always replaced with the latest sampled image data.

In the above embodiment, the procedure by the computing means 2b has been described with a focus on the basic structure, and therefore detailed description of the computation using the work screen is omitted, but the detailed procedure of the computation is particularly limited. However, the existing procedure (depending on the instruction set and the logic circuit of the microcomputer to be used) can be used. In that case, needless to say, the efficient use of the work screen and the reduction of the processing time are taken into consideration.

It should be noted that reference numerals are given in the claims to make the objects of the drawings convenient, but the present invention is not limited to the configurations of the accompanying drawings by the entry.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of an image processing system.

FIG. 2 is a memory data transition diagram illustrating a procedure for generating / updating a background image.

FIG. 3 is a memory data transition diagram illustrating a procedure for generating / updating a background image.

FIG. 4 is a memory data transition diagram illustrating a procedure for generating / updating a background image.

FIG. 5 is a memory data transition diagram illustrating a procedure for generating / updating a background image.

FIG. 6 is an explanatory diagram of an image processing method showing a conventional example.

FIG. 7 is an explanatory diagram of an image processing method showing a conventional example.

Claims (2)

[Claims] 1. An image processing method for sampling image data of a target area at a predetermined cycle and identifying a moving object existing in the target area based on the image data, the method being the nth time (n is a positive integer). ) Sampling image data and n +
The absolute value difference data of the brightness value for each pixel between the first sampled image data is derived, and the sampled image data in the area formed by the pixels in which the absolute value difference data is smaller than a predetermined threshold is stopped. The first step of recognizing as image data, and the background of the sampled image data of the area formed by the pixels recognized as the still image data continuously m times (m is a positive integer) with the repetition of the first step. A second step of recognizing as image data, and after the number of pixels not recognized as background image data of the target area becomes smaller than a predetermined number of pixels, each pixel between the background image data and the sampling image data The absolute value difference data of the luminance value is derived, and the region formed by the pixels in which the absolute value difference data is larger than a predetermined threshold value is identified as a moving body. An image processing method comprising the third step. 2. The image processing method according to claim 1, wherein the background image data recognized as the background image data in the second step is the latest sampled image data.