JPH0344589A - Method and device for separating multiple target of image - Google Patents

Abstract

PURPOSE:To enable realization of inexpensive hardware by providing a detecting element which compares the contents of first and second registers and judges the need to rewrite a region number on the basis of comparison. CONSTITUTION:A detecting element 17 receives the content of a window C from a register array 15 through a second register 16, compares it with the content of a window A in a second register 16 and judges the need to rewrite a region number on the basis of comparison. Receiving an instruction on the need to rewrite and data from the detecting element 17, a detecting element 18 checks all pixels stored in the register array 15 and replaces the contents of a pixel which needs to be rewritten. The data on the region number which is turned to be unused by the replacement are sent from the detecting element 17 to a new region number generating element 13 and thereby the region number is made reusable.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a device for tracking multiple targets by processing images obtained by capturing images of targets, and particularly relates to a device for tracking multiple targets by separating and numbering multiple targets. The present invention relates to a method and an apparatus for performing the same.

[Conventional technology]

FIG. 5 shows a conventional image multi-target separation method disclosed in, for example, Japanese Patent Application Laid-Open No. 59-135379. FIG. 5(a) is an example of a binarized image, and FIG. 5(b) is a numbered image. Examples of images that have been processed, Figure 5(C) is the target detection window, Figure 5(d) is the area number detection window, (1) is the processing area of the two-dimensional image obtained by Lasque scan, and (2) is the processing area. Assuming that there are two targets in region (1), the first binarization target is (3
) is the second binarization target, and (4) is the first separation target after assigning the number l to the first binarization target (5) is the second binarization target with number 2. This is the second separation target after numbering.

Multiple targets (including clutter, etc.) existing within processing area (1) are numbered, and targets that are not connected are assigned different numbers to separate the multiple targets. , Here, for convenience of explanation, a method of numbering 1 will be explained for the case where there are two targets.

The data in the processing area (1) is a binary video signal that is manually input sequentially from the upper left to the lower right by raster scanning.

The first binarization target (
2) and the second binarization target (3) are numbered and identified as separate groups. When the data is manually input sequentially from the upper left using Lascou scan, first the first binarization step v
Since A(2) appears, we assign the number l to this mass and set it as the first separation target (4).As the binarization target (2) appears first, we assign the number 2 and set it as the second separation target. (5).

Multiplying the first to fifth windows in FIG. 5(C) (indicated by E from beginning to end in the figure and E from beginning to end) in the binarized image of FIG. 5(a), Controls the numbering of pixels covered by the window. Then, from the first window to the fifth window (indicated by a to e in the figure, hereinafter a to e
Expressed as ) is multiplied by the numbered image in Figure 5(b), and the number of the pixel covered by the first window is changed from the second window to the fifth window.
Select and set from the following windows.

For example, if only A is 1 and B, C, D, and E are O, a new number that has not been used before is taken as the number of the pixel a in the window of FIG. 5(d).

If the pixel connected to A is 1 and there is a pixel that takes l among B, C, D, and E, there is a pixel connected to A, and that means that it has already been numbered.

The windows c, d, and e of FIG. 5(C) corresponding to that pixel
The number stored in is used as the pixel number of a. The window of FIG. 5(C) is applied to the binarized image of FIG. 5(a), and the window of FIG. 5(d) is simultaneously applied to the numbered image of FIG. 5(d). .

In this way, the two targets are numbered as the first separation target (4) and the second separation target (5).

[Problem to be solved by the invention]

Since the conventional image multi-target separation method consists of five windows (one or more), some targets cannot be separated depending on their shape and are given the same number. This will be explained using a diagram.

FIG. 4 is a diagram explaining the problem of numbering using the conventional window, FIG. 4(a) is a diagram showing the result of numbering according to the conventional invention, and FIG. FIG. 3 is a diagram showing the result of desired numbering.

This concerns whether targets that touch each other diagonally at one point as shown in Figure 4 should be numbered as the same object, or different objects should be numbered differently. Although it makes sense, in the conventional method, Fig. 4 (a
), it was possible to number only the same number, and it was difficult to number different targets as shown in FIG. 4(b).

This invention was made in order to solve the above-mentioned problems, and is a method that makes it possible to number targets that touch each other at one point in a diagonal direction as other targets, as well as simplifying the logic and hardware for realizing it. and its equipment.

[Means to solve the problem]

The image multi-target separation method and device according to the present invention include:
Binarize raster scan video signals, assign area numbers, and rewrite the assigned numbers.

A second window was placed on the left side of the first window, and a third window was placed directly above it. Using the first to third windows, it is possible to assign different numbers to targets that are in contact with each other at one point in the diagonal direction.

[Effect]

Due to the arrangement of the first to third windows in this invention,
By eliminating the windows diagonally above the left and above the first window, it becomes possible to assign different numbers to the targets that are in contact with the diagonal direction, and the number of windows is reduced from the previous 5 to 3. To,
The circuit implementing the method of this invention is simplified.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings. 1st
In the figure, Fig. 1(a) is the target detection window, Fig. 1(6)
is the area number detection window, from the first window to the third window for target detection.
window (indicated by A to C in the figure, hereinafter referred to as A to C) and the first to third windows for area number detection (indicated by a to C in the figure, hereinafter denoted by a to C). The arrangement relationship of each is the first
As shown in the figure. The first window and the second window perform the same function as in the conventional method.

The third window serves as the fourth window and the fifth window in the conventional method. The initial numbering method for targets is the same as the conventional method, but in reality, targets with many irregularities will be numbered differently, but the details to solve this problem are as follows. This method will be explained using FIG.

Figure 2 is a diagram for explaining the method of assigning one number to one target, and Figure 2 (a) is a diagram for explaining a state in which a target detection window is applied to a binarized image. , Figure 2(b)
is a diagram to explain the state in which the area number detection window is applied to the numbered image, (6) is the number 1 start point where area number 1 is named first, and (7) is the number 1 start point where area number 2 is named first. Number 2 starting point to name.

(8) is the content of the area number control register that stores information for 1947 horizontal scanning lines before the pixel indicated by the white letter A in the target detection window, and (9) is indicated by the white letter a in the area number detection window. Area number and line register contents that store information for 1947 horizontal scanning lines before the pixel to be displayed,
(10) notices that the same target was given numbers l and 2, and it becomes necessary to replace one number with the other.Number rewriting required point (11) completes numbering one target. This is the area end detection point that detects what has happened.

The point represented by 8 in the target detection window is the current pixel, B is the pixel one pixel past, the pixel CJ, is the pixel l horizontal line past the pixel tA, and the pixels A to C are each pixel. While the relative position is fixed, it sequentially moves from the upper left to the lower right of the screen, and FIG. 2(a) shows the movement from left to right during the l horizontal line period. Furthermore, the area detection window also moves in the same way as the target detection window. The target detection window and the area detection window simultaneously move corresponding pixels on the screen.

2 controls writing and rewriting of area numbers.
It is determined by looking at the value of the digitized image, and is controlled by the values of the three windows A to C for target detection.

The value of the window a is determined based on the values of the two windows (su and C) of the area number detection window.

The logic of numbering is expressed by the following equations (1) to (5).

F (A) = 0 → f (a) = O-(1)F(
A) AF (B) 8F (C) -1 → f (a) = n -
=(2)F (A)A F (B)A F (C)=
1 → “(a)-f(b)-(3)F (A)A F
(B)′″A F(C)=1−+r(a)−f(C)
-(4)F(8)A F (B)8F (C)-1→
``(a) = r(b) - (5) When formula (5) holds, f (x) - f (c) → f (x) = f (a)
...(6) Here, F(A) is the content (8) of the area number control register in target detection window A and is 0.
Or take the value of l. (The same applies to F (B) and F (C)) n represents an area number and is a positive integer.

](a) is the content (9) of the area number line register in area detection window a, taking values from 0 to n, n = 0.
represents no area.

Also, f(x) is the content of the area number line register <9>
It is.

In the target detection window and area detection window (hereinafter referred to as window) where there is no target on the screen, an area number O is written as expressed by equation (1). When the window reaches the starting point (6) with number l, l is added as the area number, as expressed by equation (2). Next, at the point where the window advances by one pixel, F (A) and F (
Since only the value of B) is l and F (C) is O, (3)
Assign the value l of f(b) to f(a) as expressed by the equation. When the window comes to the starting point of number 2 (7), F(A
) Since the chisel is l TeF (B) and F (C) are o, we assign a number to f (a) using formula (2), but since 2 number l is already in use, number 2 is newly adopted. . In order to control the area number, as shown in the shaded area in Fig. 2(a), as shown in the area number control register contents (8), 1 is stored in the area where the target exists and O is stored in other areas. put. Similarly, regarding the area numbered image, the area number of each pixel on one horizontal line is stored as the content (9>) of the area number line register, as shown by the hatched area in FIG. 2(b).

When the window reaches the number rewriting required point (10), the window 3% (
Area number 1 and area number 2 are mixed in the area.

In this case, F (A) - F (C) = F (B
) - 1, write the value f(b) = 2 to f (a) according to equation (5), and write the value f (x) (area number line register) equal to the value 1 of f (C) according to equation (6). Replace all of the contents (9)) with the value 2 of f (a). After this, the number I is stored as an unused number in the vacant number register, and when assigning a new number, the contents of the vacant number register are checked, and if a vacant number exists, it is used with priority over the new number.

At the region end detection point (11), it is determined that the region has ended.This method sets a flag for the region number used in the previous line, and then sets a flag for the region number used in the current line. Judgment is made by comparing.

In the case of Figure 2 (b), the area number used in the previous line is only 2, and the area number used in the current line is 0.
Therefore, the numbering of the area numbered with area number 2 has been completed.

Next, one embodiment of the present invention will be described.

In FIG. 3, (+2) is an area number control unit that receives and receives a multilevel video signal and controls area numbering depending on the presence or absence of a target using a binary image after binarization; (
13) is a new area number generator that is used when a target exists in a window hole and there are no similar marks around it and a new numbering is required. (+4) stores the area number of the pixel in window a. The first register, (15) is a register string that stores the area number of one horizontal scanning line past window a, (16)
is the second register used to determine the replacement of the area number, and (+7) needs to be replaced with the IM number when two or more types of area numbers are assigned to one target, but the key points are A rewriting necessity detection unit for determining whether rewriting is necessary or not, (+
8) is an area number rewrite detection unit that searches for an area number that is the same as the area number that needs to be rewritten in the contents of the register column (15) when it is necessary to rewrite an area number, and 09) is an area number rewrite detection unit that searches for an area number that is the same as the area number that needs to be rewritten. This is an area end detection unit for detecting that attachment has been completed.

The area number control unit (
+2), the video signal binarized by the threshold is stored as the content (8) of the area number control register), and is multiplied by the target detection windows A to E to obtain (1).
The area numbering is controlled according to the logic of formula (5). Since it is necessary to assign a new number to a pixel that satisfies formula (2), a command is issued from the area number control unit (12) to the new area number generation unit (13) to generate a new area number. At the same time, it is transferred to the register (14) of
9) and sets a flag to detect the end of the area.

The first register (14) receives the contents of windows b-e from the register array (15>), selects one necessary piece of data, and stores it as the contents of window a.

The rewriting necessity detection unit (17) receives the contents of the window (C) from the register row (+5) via the second register (16).

By comparing the contents of window a from the first register, it is determined whether the area number needs to be rewritten.

The area number rewriting detecting section (18) receives a rewriting necessity instruction and data from the rewriting necessity detecting section, looks at all pixels stored in the register string, and replaces the contents of the pixels that need to be rewritten. Area number rewriting detection unit (+8) is register string (15)
The contents of the area numbers stored in are called out one pixel at a time at high speed, and the area number equal to the area number of window (C) is replaced with the area number of search window a. The time required for replacement is 1
This is the time when the entire line corresponds to one pixel of image quality.

The data of the area number that has become unused due to the area number replacement is transferred from the rewriting necessity detection unit (17) to the new area number generation unit (
13) to make the area number reusable. At the same time, the data of the unused area number is sent to the area end detection unit (19).
, and change the area number flag. The area end detection unit (+9) compares the area number flag used in the current line with the area number flag used in the previous line at the end of the horizontal line, that is, at the right edge of the screen.
A region end detection point (11) detects whether or not there is an unused region number.It is determined at the region end detection point (11) that the target numbering of the unused region number has been completed, and the information is sent to the outside.

The numbering of regions for region division has been explained above, but when incorporating this into a multi-target tracking device etc.
If the number 1 is replaced at the same time as the maximum video level, the coordinates of the maximum video level (coordinates of the center of gravity, vertical and horizontal length, etc.), the feature parameter calculation section also sends the data to the calculation section.
Controls to replace data for each area number.

Then, when the end of the region is detected, the characteristic parameters of the target to which the end number has been assigned become the final values, and are used as information on the selection of the tracking target and the tracking point.

〔Effect of the invention〕

As described above, by using the image multi-target separation circuit of the present invention, it is possible to assign different numbers to targets that are in contact with each other in diagonal directions, and since the number of window configurations is small, it is possible to realize inexpensive hardware. It has become possible to create a fire control system and a missile guidance system that can track multiple targets at the same time, and it can also be applied to devices that measure and understand multiple shapes in real time, which would have great effects.

[Brief explanation of drawings]

Figure 1 is a diagram explaining the relationship between the windows of this invention, and Figure 2 is a diagram illustrating the relationship between the windows of this invention.
A diagram for explaining a method of assigning one number to one target; FIG. 3 is a diagram for explaining an embodiment of the present invention;
FIG. 4 is a diagram for explaining the problem of numbering using windows in the conventional invention, and FIG. 5 is a diagram for explaining the conventional method.
(1) is the processing area (2) is the first binarization target, (3) is the second binarization target, (4) is the first separation target, (5) is the second separation target, ( 6) is the start point of number l, (7> is the start point of number 2, (8) is the content of the area number control register, (9) is the content of the area number line register,
(10) is the number rewriting required point, (11) is the area end detection point, (12) is the area number control section, (+3) is the new area number generation section, (14) is the first register, (1
5) is the register string, (16) is the second register, (17
) is a rewriting necessity detection section, (1g) is an area number rewriting detection section, and (19) is an area end detection section. In the drawings, the same or corresponding parts are designated by the same reference numerals.

Claims (2)

[Claims] (1) A raster scanning video signal obtained from a camera, video playback device, etc. is binarized and stored in a register to control the area number, and a second window is opened on the left side with the first window in the center. In real time, multiple targets in the screen are separated from each other by numbering using the first to third windows arranged in the positional relationship of the third window directly above. An image multi-object separation method characterized by: (2) Extract the target using the third window from the first window, which is arranged in a positional relationship with the second window on the left side and the third window directly above the first window. an area number control section that controls numbering; a first register that selects and stores the contents of an area number to be written in the first window according to a command from the area number control section;
a register column for storing the area number from the register for one horizontal scanning line; a second register for storing the contents of the area number of the third window from the register column; A rewriting necessity detection unit compares the contents from the registers and determines whether or not the area number needs to be rewritten, and upon receiving a command from the rewriting necessity detection unit, the contents of the register row are transferred to the third window. An image multi-target separation device comprising: an area number rewriting detection unit that replaces all pixels having the same value; and an area end detection unit that detects completion of target numbering.