KR20080017521A - Method for multiple movement body tracing movement using of difference image - Google Patents

Abstract

A method for tracing a multiple object using difference images is provided to trace multiple objects rapidly and exactly by confirming the existence of a multiple mobile object on the basis of difference images, separating the objects of the multiple mobile object using projection vectors, obtaining the direction of the multiple mobile object through the comparison between the objects obtained at the previous frame and the objects obtained at the present frame, setting a reference block for a big edge portion in the obtained direction, and carrying out block matching. In tracing a multiple mobile object, a control part obtains a difference image from the image signals inputted through a CCD camera. If two difference images are obtained, the control part separates the objects of the multiple mobile object respectively using projection vectors from the two difference images. Then the control part establishes a reference block for a sharply changed edge portion of each of the objects. By carrying out block matching between the established reference block and the data interval of the present difference image, the control part traces the multiple mobile object.

Description

Multi-object tracking method using difference image {METHOD FOR MULTIPLE MOVEMENT BODY TRACING MOVEMENT USING OF DIFFERENCE IMAGE}

1 is a flowchart of a multi-object tracking method using a difference image according to the present invention.

2 to 9 are detailed explanatory diagrams of a multi-object tracking method using a difference image according to the present invention.

The present invention relates to a multi-object tracking method using a car image, and more particularly, to a multi-object using a car image capable of simultaneously tracking a plurality of moving objects using a car image extracted from a sequentially input image through a camera. It is about a tracking method.

In general, the multi-object tracking method is applied to various applications such as object recognition, robots, surveillance systems, unmanned vehicle systems, as well as military purposes such as aircraft and guided weapons.

Multi-object tracking is divided into two methods: separating moving objects and tracking moving objects.

In the first object separation, each object is separated by a projection vector, and the objects are separated by performing projection vectors on the X and Y axes. The second moving object tracking was tracked using difference image based method, block matching.

However, the car image is a method of detecting and tracking a moving object from a still image, but simple object tracking is difficult for complex objects. In addition, block matching can perform accurate object tracking, but it takes too much time because of large amount of computation when block matching is performed for the entire block.

The present invention has been made to solve the above-described problems, and its object is to identify the presence or absence of a moving object based on the difference image and to distinguish each object of the moving object with respect to the difference image using the projection vector, -1) to find the direction of movement of the moving object through comparison between the objects obtained in the current frame (t) and to obtain the correct block matching The present invention provides a multi-object tracking method using a difference image that enables fast and accurate multi-object tracking by setting blocks and performing block matching.

Hereinafter, a preferred embodiment of a multi-object tracking method using a difference image according to the present invention will be described in detail.

First, FIG. 1 is a flowchart of a tracking method of multiple objects according to the present invention.

The NTSC analog signal input from the CCD camera is used to obtain RGB data using the grabber board, and is converted into a frame of the Y signal.

At this time, to find the difference image, it checks whether there is a previous (t-1) Y frame, and if not, it waits until the next image comes in.

When two images are acquired, a difference image is obtained.

The image is acquired again and the objects of the multiple moving objects are separated using the projection vector.

Using the separated data, the corresponding objects of the two frames are distinguished, and the moving direction is detected by comparing the area data of the separated objects.

For accurate block matching, reference blocks are set for the areas where the edge change of each object is severe. Multiple objects are tracked by block matching the set reference block with the data section of the current image.

Next, the difference image detection process will be described with reference to FIGS. 2 to 4.

The difference image according to the movement is obtained by using the similarity of the background data photographed from the CCD camera.

The two original images as shown in FIG. 2 are detected using the projection vector as shown in FIGS. 3 and 4 to detect the most similar parts, and the moving distance is examined to obtain a difference image.

3 is a projection of the original image by the X-axis, respectively, to examine the vertical movement distance. FIG. 3 (a) shows a projected image for each original image, and FIG. 3 (b) compares two projection vectors to detect moving distances for the most similar parts.

4 is a projection along the Y-axis to check the left and right movement distance of the camera. FIG. 4A illustrates a Y-axis projection of each original image, and FIG. 4B compares two projection vectors to detect moving distances for the most similar portions.

Next, the object separation will be described with reference to FIG. 5.

Object separation separates objects to extract the moving distance of each object when there are multiple moving objects.

That is, as shown in Fig. 5, the respective images are separated using the projection vector of the difference images of the two original images. 5A is a difference image of an original image of two frames. In FIG. 5B, the first X-axis projection is performed, and only a portion having data in the X-axis direction is set as an area. (C) of FIG. 5 projects in the Y-axis direction with respect to the area | region divided by FIG. 5 (b), and performs separation again. In FIG. 5B, the Y-axis projection vector is implemented using the regions where the three regions are set, and separated into four objects.

FIG. 5D illustrates an X-axis projection vector for each of the areas set separately from FIG. 5C, to more accurately separate objects. FIG. 5E shows that each object is separated through three projection vectors.

Next, the block matching described above will be described.

Block matching algorithms are frequently used in the detection of motion vectors. By detecting how the object moves by the time difference, matching is performed by dividing n frames and (t-1) frames by block unit. The block matching algorithm is a block in which the sum of squares of the pixel differences symmetrical with MSE and MAE and the smallest value with absolute sum match.

In the present invention, the MAE block matching algorithm is used. FIG. 7 shows a method of performing block matching.

The reference block in the t frame is a block for matching the (t-1) frame to find the most similar block. (t-1) Frame matching section is a section to find if there is a similar block like the reference block. (t-1) Since block matching is performed for the entire frame, a large amount of calculation is required. Set the block.

The target block is a block to be matched with the reference block in the matching section. The result obtained by matching the reference block and the target block is set to the block where the smallest value is matched, and the moving distance is obtained by subtracting the reference block coordinates from the block matching the reference block, as shown in the figure. Δk, m + Δl) and reference coordinates (i, j) minus

As shown in FIG. 7, the block matching sequence moves one pixel to the right from the (n, m) coordinate point, which is the first coordinate of the matching section. When Δk is moved by k-B, block matching is performed for the entire matching section, starting again from the position of (n + 1, m). The MAE block matching algorithm equation is shown in FIG. 8.

Hereinafter, a reference block setting process will be described.

In block matching, the reference block is a standard of similar blocks to be found in the matching section. To increase accuracy, the reference block should be set for the object to be searched and the change part. The reference block setting uses two difference images to detect the moving direction of the moving object using the block corresponding to each object, and sets the reference coordinate before setting the reference block for each moving object. The reference block is set at the largest edge in the block.

FIG. 6 is a diagram for setting reference coordinates for setting a reference block to an object that is currently moving using a moving direction of the moving object. The reference coordinate is set at the boundary of the matching section to play a role in setting the reference block. In four cases as shown in FIG. 6, using the corresponding object blocks of (n-1 Frame)-(n-2 Frame) and the current image of (n Frame)-(n-1 Frame). Set the reference coordinate for.

FIG. 9 illustrates a method of setting reference coordinates for three images and setting a reference block for a portion of which edge change is severe. Comparing the difference blocks (d) of (a) and (b) and (e) of (b) and (c) by comparing the corresponding blocks, the moving direction of the object is extracted as shown in (f), and as shown in (g). Set the reference coordinate. Then, the reference block is set as shown in (h) for the portion having the high edge within a certain range. The matching section sets the difference image data region of (h) as the matching section.

As a result, multiple objects can be tracked by obtaining a moving distance between the reference block and the matching block.

As described above in detail, according to the present invention, the presence or absence of a moving object is determined based on the difference image, the object obtained from the previous frame (t-1) by distinguishing each object of the moving object with respect to the difference image using the projection vector. To find the direction of movement of the moving object through the comparison between the objects obtained from the current frame (t) and the block matching by setting the reference block for the part with the large edge in the direction of movement. By doing this, there is an effect that fast and accurate multi-object tracking is possible.

Claims (1)

In the multi-object tracking method using the difference image, Obtaining a difference image by receiving an image signal from a CCD camera; Separating the objects of the multi moving object using the projection vector from the two difference images, Setting a reference block for the edge change of each separated object; A method of tracking a multi-object using a difference image, comprising: matching a set reference block with data sections of a current difference image to block multiple objects.

Images