JP2000341704A - Device for detecting target position of color image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a target image position detection means capable of accurately detecting a target independently of the tone of the target. SOLUTION: In the target position detection device, a stimulus value conversion part 1 converts image signals of red(R), green(G), and blue(B) outputted from a color image sensor 101 for photographing a target color image and outputting three color signals of R, G and B into three stimulus value signals X, Y, Z corresponding to the inputted color signals at real time in each pixel outputted from the sensor 101 by executing prescribed operation and a chromaticity coordinate operation means 2 calculates the signals x, y of chromaticity coordinate points including no information about the size of stimulus values corresponding to the tone of the image from the three stimulus value signals X, Y, Z, finds out chromaticity difference signal ΔC from the signals x, y and detects a target by using the signal ΔC.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a color target position detecting device for tracking a target using a visible image captured by a television camera.

[0002]

2. Description of the Related Art Conventionally, this type of target tracking apparatus has realized automatic tracking of the imaging direction by detecting the movement of a target from an image taken by a television camera. As a method of detecting this movement, a method of detecting a position of a block having a maximum correlation with a reference image block defined by a peripheral image of a target or an image of a characteristic portion from a captured image is generally used. I have been.

In this method, when the size of the target object changes or deforms, there is a disadvantage that tracking becomes difficult, and in addition, a circuit scale for obtaining a correlation becomes extremely large. Further, there is a problem that an attempt to reduce the circuit scale in the CPU processing increases the processing load and loses real-time performance.

Therefore, there have been proposed several methods for using color information of a target based on a color image signal of the target. As one of them, for example, R of a color image signal,
Consider G and B as vector components on rectangular coordinates, and use the unit vector of the color to be tracked as a reference vector and the inner product with the R, G, and B vectors of the image signal.

[0005]

(Equation 5) To obtain the difference in the vector direction as cos θ (where θ is the angle between the vector of the corresponding pixel of the captured image and the reference vector).

Japanese Unexamined Patent Publication No. 4-184283 discloses a method of improving the target discrimination ability by using R, G, and B of a color image signal.

[0007]

(Equation 6) Or

[0008]

(Equation 7) In the method of converting into r, g, and b color system chromaticity values and comparing the values, a determination range for a set of two reference values is set from the target r or b and g, and the correspondence of the captured image is set. The comparison circuit determines the r or b and g of the pixel to be processed, and if both values fall within the determination range, the pixel is binarized to obtain a target area and then correspond to the luminance value (R + G + B). The method of detecting the position at the maximum luminance point in the target area of (1) has been proposed.

[0009]

In the above-mentioned conventional example, the information on the brightness of the image is excluded, and the influence of the brightness is suppressed. In addition, there is a disadvantage that the values are almost the same, and it is extremely difficult to identify the target point.

Furthermore, in order to obtain the value of cos θ, it is necessary to divide by the absolute values of the R, G, and B vectors of the image signal. A method of multiplying this is used,
This method also has the disadvantage that as the values of R, G, and B increase, the reciprocal decreases and the ratio of the quantization error increases, which is increased by the multiplier, and the target detection error increases.

According to the method disclosed in Japanese Patent Application Laid-Open No. 4-184283, the target point can be easily identified and the detection ability is improved as compared with the former method. , In the case of digital operation, direct division is performed to obtain r, g, and b.
In the image region where the value of (+ G + B) is large and the image is bright, the quantization noise increases.

Therefore, the disadvantage that the target determination due to the error causes an error has not been solved, and the maximum luminance point is set as the target detection position in the final stage, so that the target area determined by the chromaticity determination is within the target area. There is a drawback that the detection position jumps due to a change in the luminance distribution of.

On the other hand, without using the above-described division operation, a target is obtained by using two stimulus value signals of RY and BY having information relating to hue and saturation with respect to the Y stimulus value. A method of determining the position of an object and outputting a tracking signal is disclosed in
No. 5,515,564.

However, in this method, in order to determine the target without being affected by the luminance of the target, the relationship between the two stimulus values RY and BY of the target and the Y stimulus value (luminance) is determined. It is necessary to set an allowable error range for determining a modeled color prediction function and a target in consideration of this function model, and the quality of the model setting and the setting of the allowable error range affects the characteristics of detection or false recognition. It has the drawback.

An object of the present invention is to provide a color target position detecting means capable of eliminating the drawbacks of each of the above methods.

[0016]

A color target position detecting device according to the present invention captures a target color image and outputs three color signals of red (R), green (G) and blue (B). By performing a predetermined operation on the image signals R, G, and B from the sensor, the tristimulus value signals X, Y, and X corresponding to the color signal input in real time for each pixel output from the sensor.
After the conversion into Z, the signals x, y of the chromaticity coordinate points not including the information of the magnitude of the stimulus value corresponding to the brightness of the image are calculated from the tristimulus value signals X, Y, Z, and this color is calculated. It is characterized in that a chromaticity difference signal ΔC is obtained from the signals x and y of the chromaticity coordinate points, and target detection is performed using the chromaticity difference signal ΔC.

Since the chromaticity coordinate points are directly related to the inherent spectral reflectance characteristics of the target, accurate determination can be made without being caught by the brightness of the target, and the target position with few errors can be detected. . Further, since the values of x and y are equal to or less than “1”, scaling of processing values required for circuit design in subsequent processing is facilitated.

In the present invention, the chromaticity signals x and y are combined with the stimulus value signal Y, and the target can be detected based on the chromaticity coordinate point signals x and y and the stimulus value signal Y.
Since the stimulus value signal Y corresponds to the brightness of the color light, more accurate detection determination can be made by combining the chromaticity values, x, and y using the Y stimulus value as it is for target detection.

Further, according to the present invention, the signals x, y of the chromaticity coordinate points are provided.
A method of subtracting after logarithmic compression can be used for the operation for obtaining. This is because color signals R, G, and B are converted into stimulus value signals X, Y, and Z, so long as all of R, G, and B are “0”, stimuli that are numerators of x and y Since the values X and Y never become “0”, it is possible to adopt the effect obtained by the stimulus value conversion into X and Y that stable logarithmic compression is possible in the area where the image is captured. Therefore, the dynamic range of stimulus values that can be handled to obtain x and y can be made extremely wide, and the division circuit can be omitted.

Further, according to the present invention, a positional deviation of a captured image from a coordinate reference in a chromaticity space with respect to a target chromaticity value to be detected is converted into one signal value corresponding to distance information. Then, a target point is detected by detecting a point at which this value is minimum. As a result, the scale of the processing circuit can be reduced.

Hereinafter, the features of the present invention will be described with reference to FIG. In the present invention, the color image sensor 101
The three image signals R, G, and B of red, green, and blue are input by the stimulus value converter 1 for each pixel in real time in response to a color signal input.

[0022]

(Equation 8) Is converted to tristimulus value signals of X, Y, and Z.

R, G, B from the color image sensor 101
As shown in FIG. 13, the signals are time-series signals R (α, β), G (α, β), and B in the imaging screen 501 over the entire frame of the screen one pixel at a time according to the signal output order 502.
(Α, β), the X, Y, and Z tristimulus value signals are also time-series signals X similar to the R, G, and B signals.
(Α, β), Y (α, β), and Z (α, β). Here, α represents the pixel position of the horizontal pixel row on the screen, and β
Represents the position of the vertical pixel row on the screen, and corresponds to the coordinates on the screen.

In the present invention, from the values of the X, Y, and Z tristimulus value signals,

[0025]

(Equation 9) Is converted to a signal of x, y chromaticity coordinate points not including information of the magnitude of the stimulus value corresponding to the brightness of the image, and the time-series chromaticity signals x (α, β), y (α, β) ) Is output.

The obtained chromaticity coordinate signal x (α,
β), y (α, β) and the stimulus value Y (α, β) are input to the designated point data holding unit 13, and the coordinates α, β
X of the pixel that matches the initial capture designated point Gto expressed by
(Α, β), y (α, β) and the stimulus value Y (α, β) are selected and held, and are updated for each image frame.

This held data is a designated point chromaticity signal xt
o, yto, and the designated point stimulus value Yto are output to the reference chromaticity updating unit 4 and the reference luminance updating unit 10 as outputs of the designated point data holding unit 13, respectively.

The reference chromaticity updating section 4 only outputs the designated point chromaticity signal xt when the setting completion timing signal To is input.
The values o and yto are directly updated to the values of the reference chromaticity signals xo and yo and output. On the other hand, the reference luminance updating unit 10 updates the designated point stimulus value Yto to the value of the reference luminance signal Yo only when the setting completion timing signal To is input, and outputs it. That is, the reference chromaticity signals xo and yo and the reference luminance signal Yo are the initial chromaticity and luminance values of the target point captured at the time of input of the setting completion timing signal To.

After receiving the setting completion timing signal To, the reference chromaticity updating unit 4 outputs the minimum minimum point chromaticity value output from the minimum point data holding unit 6 corresponding to the obtained chromaticity of the tracked target point. xpm and ypm are input for each frame, the reference chromaticity of the next frame is predicted or filtered, and new reference chromaticity signals xo and yo are output.

In the reference luminance updating unit 10, similarly,
The minimum minimum point luminance value Ypm of the minimum point data holding unit 6 in which the minimum minimum point chromaticity values xpm and ypm corresponding to the chromaticity of the target point being tracked are input, and prediction or filtering processing is performed. This is a new reference luminance signal Yo.

Therefore, using the values of the reference chromaticity signals xo, yo and the reference luminance signal Yo as reference values, the range of the target chromaticity coordinate difference distance and the distribution limit of the stimulus value Y are set, and both are limited. Since it has the chrominance difference signal binarization unit 7 and the luminance signal binarization unit 8 that can binarize an area in the target, the target existence area can be output as a binarized image. it can.

In the present invention, the reference chromaticity signals xo, yo
And the chromaticity signal x (α, β), y from the chromaticity coordinate calculator 2
(Α, β), and outputs the result as a chromaticity difference signal ΔC (α, β) (ΔC map). α, β) and a minimum point detection unit 5 for detecting a point at which this signal becomes minimum.
have.

Further, the present invention has a minimum point data holding unit 6 for stimulus values Y (α, β), chromaticity signals x (α, β), y
(Α, β) and the chromaticity difference signal ΔC (α, β) are input, and the chromaticity difference signal ΔC (α, β)
Every time a minimum point is detected, the minimum point detection timing signal p
m, the minimum point data holding unit 6 takes in the chromaticity difference signal ΔC (α, β) of the minimum point every time pm is generated, and compares this value with the value obtained in the past and the current value. Then, the chromaticity difference signal ΔC (α, β) of the smaller minimum point is held.

At this time, the held chromaticity difference signal ΔC (α,
The stimulus value Y (α, β) and the chromaticity signals x (α, β) and y (α, β) input corresponding to β) are also held at the same time. As a result, when the input of the color signal of one frame is completed, Y (α, β), x (α, β), y (α, β) corresponding to the minimum value of the minimum value of the chromaticity difference signal ΔC (α, β) are obtained. ),
And ΔC (α, β) can be obtained, and these are respectively calculated as minimum minimum point luminance value Ypm and minimum minimum point chromaticity value xpm, y
pm and the minimum minimum point chromaticity difference ΔCpm.

In the present invention, the chromaticity difference signal ΔC (α, β) and the stimulus
It has a ΔC map frame memory 71 shown in FIG. 4 and a luminance map memory 81 shown in FIG. 5 for temporarily storing (α, β) as an image map for one frame. By reading, the same chromaticity difference signal ΔC (α, α,
β) and the image map of the stimulus value Y (α, β) is ΔCpm
Binary binarization using a threshold that sets the range of the chromaticity coordinate difference distance of the target and the distribution limit of the stimulus value Y using the threshold values Yp and Ypm as the reference values, and binarizes an area where both are within the limits. Image outputs BΔC (α, β) and BY (α, β) can be obtained.

The logical product B (α,
The binarized image centroid calculating unit 11 that calculates the final surface centroid according to β) can output the binarized area centroid of the target existence region as the target detection position signal G.

As described above, in the present invention, since the chromaticity coordinate points directly related to the intrinsic spectral reflectance characteristic of the target are evaluated as the feature amounts in the target detection evaluation, the spectral conditions of the irradiation light with respect to the target are evaluated. If the change is small, it means that an accurate feature amount that is not captured by the brightness of the target is determined, and therefore, there is a feature that the target position can be detected with much less error.

In the chromaticity coordinate calculator 2 for obtaining chromaticity coordinates, the stimulus values X (α, β) and Y (α,
β) and Z (α, β) are logarithmically compressed by the following calculation to obtain chromaticity coordinates.

[0039]

(Equation 10) This is because if the stimulus values X and Y never become “0” unless all of R, G and B become “0”, stable logarithmic compression is possible in the area where the image is captured. It became possible to adopt it.

Therefore, when digital arithmetic is used for an extremely wide range of image signals R, G, and B by logarithmic compression, the number of data bits is greatly increased in consideration of the data range of the stimulus value. It also has an excellent feature that it is possible to prevent an increase in an error in chromaticity calculation without using it.

Further, as for the method of calculating the chromaticity difference, as shown in FIG. 8, the reference chromaticity signals xo and yo obtained by the reference chromaticity updating unit 4 are converted to the target designated point chromaticity coordinates 2 in the chromaticity coordinates.
03, and the chromaticity signals x (α, β) and y (α, β) of each pixel output in chronological order from the chromaticity coordinate calculation unit 2 are used as input image chromaticity coordinates 202 to determine the chromaticity coordinates between these two points. Coordinate direct distance r
Can be considered as a chromaticity difference, but the chromaticity difference calculating unit 3 of the present invention uses

[0042]

[Equation 11] Is used as the chromaticity difference ΔC. Note that ABS represents an absolute value.

In this method, the direct distance r between the target chromaticity coordinate point and the reference chromaticity coordinate point is always

[0044]

(Equation 12) When the direct distance between the target chromaticity coordinate point and the reference chromaticity coordinate point is not “0”, it is almost always larger than the distance r in the above equation between the coordinate points. In addition to the feature that it has a sharp target discriminating ability, it also has a feature that can be realized by a simple addition and subtraction circuit.

In addition, since the present invention is not intended for display, the output of the image pickup device is output as it is without performing a masking process or the like that approaches an ideal characteristic having a negative spectral characteristic in consideration of color reproducibility. It is assumed to be used as In this case, since the output of the imaging device of the color image sensor has no negative sensitivity region, the chromaticity value of the coordinates located outside the chromaticity value range 201 based on the RGB signal of the color image sensor as shown in FIG. It will not be.

The chromaticity value range 201 of the color image sensor based on the RGB signals is represented by a straight line 301 where x + y = 1,
It is in the range surrounded by the x- and y-coordinate axes and the limit line 302 satisfying ΔC ≦ 1.0 with respect to the point O.
The magnitude of C is also always in the range of "0" to "1", which has the advantage that data scaling of the circuit for processing .DELTA.C as in r is easy.

[0047]

FIG. 1 is a block diagram showing an embodiment of a color image target position detecting device according to the present invention. 2 to 5 show an embodiment of the chromaticity coordinate calculator 2, chromaticity difference calculator 3, chromaticity difference signal binarizer 7, and luminance signal binarizer 8, respectively, of FIG. Is shown. FIG. 6 is a block diagram showing a connection example using the color image target position detecting device of FIG.

First, the outline of the target tracking operation using the color image target position detecting device of the present invention will be described with reference to FIG.

The color image target position detecting device 10 of the present invention
0 is a color image sensor 101 that captures the image of the imaging target 106 and outputs three image signals R, G, and B of red, green, and blue;
An image display device 103 that displays a captured image by image signals R, G, and B, an initial capture designation input device 102 that designates a target to be tracked by an operator from a display screen at an initial stage, and a target detection from a color target position detection device 100. A tracking signal output device 105 for outputting a target tracking drive signal DT for tracking a target based on the position signal G and a gimbal device 104 for changing its visual axis based on the target tracking drive signal DT are used.

In the tracking operation, first, the gimbal device 104 is moved by direct operation to adjust the visual axis of the color image sensor 101 to the approximate target direction, and the image display device 10 is operated.
The tracking target 108 is imaged on the third display screen 107.

The operator displays a target to be tracked on the display screen 107.
Then, the initial capture designation input device 102 sets a tracking target point and sets a window for prohibiting capture of a similar target to be excluded. On the display screen 107 of the image display device 103, the tracking target point and the window set by the initial capture designation input device 102 are displayed so as to be superimposed on the images captured as the initial capture designated point marker 107b and the initial capture window marker 107a, respectively. Therefore, it is possible to set the tracking target point and the window while monitoring this.

This set value is set in the initial capture designation input device 10.
Initial capture designated point Gto represented by coordinates α and β on the screen in 2
And an initial capture window WGto, which are sequentially output to the color image target position detecting device 100 of the present invention.

When the setting is completed, the operator performs a switch operation or the like.
Is generated, and is output to the color image target position detecting device 100 of the present invention.

The color image target position detecting device 100 receives the initial capture designation point Gto from the initial capture designation input device 102.
And an initial capture window WGto at the time of input of the setting completion timing signal To, and this point is set as an initial tracking target point. Hereinafter, the color image target position detecting device 100
A target position is detected based on the feature amount of the tracking target point obtained from the color image, and a target detection position signal G is sequentially output.

The tracking signal output device 105 outputs a target tracking drive signal DT for controlling the gimbal device 104 so that the target detection position signal G becomes a coordinate point at the center of the screen. As a result, after the generation of the setting completion timing signal To, the visual axis of the color image sensor 101 continues to follow the movement of the target for which capture designation has been performed.

Next, the configuration of the color image target position detecting device 100 of the present invention will be described with reference to FIGS.

The color image target position detecting device of the present invention comprises:
The color image signal (red) R, the color image signal (green) G, the color image signal (blue) B, which are the inputs from the color image sensor 101, and the initial which is the input from the initial capture designation input device 102 shown in FIG. The target detection position G is output by inputting six signals of the designated capture point Gto, the initial capture window WGto, and the setting completion timing signal To.

The color image signals (red) R, (green) G,
(Blue) Three of B are connected to three input terminals of the stimulus value conversion unit 1, and three output terminals of the stimulus value conversion unit 1 for the X, Y, and Z stimulus values are sent to the chromaticity coordinate calculation unit 2. It is connected. The outputs of the chromaticity coordinates x and y of the chromaticity coordinate calculator 2 are connected to a chromaticity difference calculator 3, and the chromaticity difference signal ΔC as the output is input to the minimum point detector 5.

At the input end of the minimum point data holding section 6, the minimum point detection timing signal pm from the minimum point detection section 5, the chromaticity difference signal ΔC from the chromaticity difference calculation section 3, and the chromaticity coordinate calculation section 2 , And the luminance signal Y (Y stimulus value) from the stimulus value converter 1 are input.

In the minimum point detecting section 5, the chromaticity difference signal ΔC is used as an evaluation area generating section 1 for limiting the detection area of the minimum point.
2 and the detection area signal WG, which is an input from FIG.
As shown in FIG. 2, the chromaticity difference signal ΔC input in time series is sequentially determined for the range of the pm generation range designation area 404, and the points ΔCi, at which the chromaticity difference signal ΔC satisfies the minimum value condition, When a j is detected and the pm generation point 403 is scanned, a minimum point detection timing signal pm is output.

The minimum point data holding unit 6 captures the chromaticity difference signal ΔC of the minimum point every time pm is generated, and compares this value with the value obtained in the past and the current value to determine the smaller minimum point. The chromaticity difference signal ΔC is held. At this time, the luminance signal Y and the chromaticity signals x and y input corresponding to the held chromaticity difference signal ΔC are also held at the same time.

As a result, at the time when the input of the color signal of one frame is completed, Y, x, y, and ΔC corresponding to the minimum minimum value of the chromaticity difference signal ΔC can be obtained. Are output as the minimum minimum point luminance value Ypm, the minimum minimum point chromaticity values xpm and ypm, and the minimum minimum point chromaticity difference ΔCpm, respectively.

The chromaticity signals x and y obtained from the chromaticity coordinate calculation unit 2 and the luminance signal Y obtained from the stimulus value conversion unit 1 are input to the designated point data holding unit 13, and the color image The chromaticity x and y and the stimulus value Y of the pixel at the screen coordinate point that matches the initial capture designated point Gto, which is the input of the target position detecting apparatus 100, are held and updated for each image frame.

The chromaticity x, y and the luminance signal Y are designated as designated point chromaticity signals xto, yto and designated point stimulus value Yto from the designated point data holding unit 13 and the reference chromaticity updating unit 4 and the reference luminance updating unit 4, respectively. Output to the unit 10.

At the input terminal of the reference chromaticity updating unit 4, the designated point chromaticity signals xto and yto from the designated point data holding unit 13 and the minimum minimum point chromaticity value xp from the minimum point data holding unit 6 are input.
m and ypm are input.

The reference chromaticity updating unit 4 only outputs the designated point chromaticity signal xt when the setting completion timing signal To is input.
The values of the reference chromaticity signals xo and yo are updated as they are in o and yto, and in other cases, the minimum minimum point chromaticity values xpm and ypm
Is used to predict or filter the reference chromaticity of the next image frame, and update and output new reference chromaticities xo and yo. The output of the reference chromaticity updating unit 4 is connected to the input terminal of the chromaticity difference calculating unit 3.

On the other hand, the input terminal of the reference luminance updating unit 10
The specified point stimulus value Yto from the specified point data holding unit 13 and the minimum minimum point luminance value Yp from the minimum point data holding unit 6
m is input. The reference luminance updating unit 10 sets the designated point stimulus value Y only when the setting completion timing signal To is input.
The value is updated to the value of the reference luminance signal Yo as it is, and in other cases, using the minimum minimum point luminance value Ypm,
The reference chromaticity of the next image frame is predicted or filtered, updated to a new reference luminance signal Yo, and output.

The output Yo of the reference luminance updating section 10 is connected to one of two input terminals of the luminance signal binarizing section 8. A luminance signal Y to be binarized from the stimulus value conversion unit 1 is input to the other input terminal of the luminance signal binarization unit 8, and a luminance image 2 is output from the output terminal of the luminance signal binarization unit 8. The digitized signal BY is output and connected to one of three input terminals of the AND circuit 9.

One of the two input terminals of the chrominance difference signal binarization unit 7 receives the minimum minimum point chromaticity difference ΔCpm from the minimum point data holding unit 6, and the other input terminal receives the chromaticity difference Calculator 3
, The chromaticity difference signal ΔC to be binarized is input. An output terminal of the chromaticity difference signal binarization unit 7 outputs a ΔC map binarization signal BΔC, which is connected to another of the three input terminals of the AND circuit 9. A third input terminal of the AND circuit 9 is connected to a binarized window area signal w for limiting a binarized output image area, which is an output of the evaluation area generator 12.

In the AND circuit 9, as shown in FIG. 10, the binary image signal B obtained by the logical product of the luminance image binary signal BY, the ΔC map binary signal BΔC, and the binary window area signal w. And a binarized image centroid calculation unit 1
1 is input.

The binarized image center-of-gravity point calculating section 11 performs the processing shown in FIG.
As shown by, the area centroids (αG, βG) of the binarized area where B = 1 are obtained, set as the target detection position G, and at the same time, the binarized area S whose value is the number of pixels where B = 1. And outputs it from the binarized image centroid calculation unit 11. In addition,
The target detection position G is output from the present color image target position detection device.

An evaluation area generator 12 for generating a detection area signal WG for limiting the detection area of the minimum point and a binarization window area signal w for limiting the binarized output image area corresponding to this area. At the input end, a target detection position G and a binarization area S from the binarized image centroid calculation unit 11, an initial capture window WGto and a setting completion timing signal To which are inputs to the present color image target position detection device. Is entered.

When the evaluation area generating section 12 receives the setting completion timing signal To, the initial capture window WGto corresponding to the initial capture window marker 107a in FIG. 6 is set as the initial value of the detection area signal WG. A corresponding binarized window area signal w is generated to ensure initial acquisition.

After receiving the setting completion timing signal To, the evaluation area generating unit 12 calculates the target detection position G input from the binarized image center-of-gravity point calculating unit 11 and the number of pixels where B = 1 as a value. From the binarized area S, the target detection position G is set as the center position, and the ratio of the binarized area S to the area area of WGto is constant while the vertical and horizontal ratio of the initial capture window WGto is fixed. Then, the detection area signal WG and the binarized window area signal w generation area are updated.

Therefore, the minimum point detecting section 5 can detect the minimum minimum value of the chromaticity difference signal ΔC only in the area near the tracking target, and converts the binary image signal B into the binary image centroid. It is possible to input to the point calculation unit 11.

In the configuration of the present apparatus, the chromaticity coordinate calculating section 2
, One three-input addition circuit 21, three logarithmic compression circuits 22, 23, 24, and two two-input subtraction circuits 2
5 and 26, and two exponential function generating circuits 27 and 28.

A signal line for X, Y, and Z stimulus values is connected to the three-input addition circuit 21 and the output terminal thereof is connected to the logarithmic compression circuit 2.
2 are connected. The logarithmic compression circuit 23 is connected to an X stimulus value signal line, and the logarithmic compression circuit 24 is connected to a Y stimulus value signal line.

An output terminal of the logarithmic compression circuit 22 is connected to subtraction value input terminals of two-input subtraction circuits 25 and 26. An output terminal of the logarithmic compression circuit 23 and an output terminal of the logarithmic compression circuit 24 are respectively an exponential function generation circuit. 27 and an exponential function generating circuit 28. Exponential function generation circuit 27
The signal of the chromaticity coordinate x is output to the output terminal of, and the signal of the chromaticity coordinate y is output to the output terminal of the exponential function generation circuit 28.

In the configuration of the present apparatus, the chromaticity difference calculator 3 comprises two two-input subtraction circuits 31, 32, two absolute value conversion circuits 33, 34 and a two-input addition circuit 35, as shown in FIG. ing. A signal line of chromaticity coordinates x and reference chromaticity coordinates xo is connected to an input terminal of the two-input subtraction circuit 31, and a signal line of chromaticity coordinates y and reference chromaticity coordinates yo is connected to an input terminal of the two-input subtraction circuit 32. Is connected.

The output terminals of the two-input subtraction circuits 31 and 32 are connected to absolute value conversion circuits 33 and 34, respectively, and their output terminals are connected to the input terminal of a two-input addition circuit 35. 2
A chromaticity difference signal ΔC is output from the output terminal of the input addition circuit 35.

In the configuration of the present apparatus, the chrominance difference signal binarizing section 7 has a ΔC map frame memory 71, ΔC
It comprises a Cmax calculation processing circuit 72 and a binarization circuit 73. A signal line of the chromaticity difference signal ΔC from the chromaticity difference calculator 3 shown in FIG. 1 is connected to an input end of the ΔC map frame memory 71, and an output end thereof is connected to an input end of the binarization circuit 73. I have.

The input terminal of the ΔCmax calculation processing circuit 72 is connected to the signal output of the minimum minimum point chromaticity difference ΔCpm from the minimum point data holding unit 6 shown in FIG.
It is connected to the binarization threshold setting input terminal of the binarization circuit 73. The minimum minimum point chromaticity difference ΔCpm is the chromaticity closest to the chromaticity of the target designated point, and can be handled as the latest target chromaticity at the time of imaging.

Therefore, by setting a value larger than this value by the target judgment limit value of the target model and using this value as a threshold value and binarizing ΔC below this value, the target feature existence area can be binarized. Will be able to

The ΔCmax calculation processing circuit 72 converts the value of the minimum minimum point chromaticity difference ΔCpm held in the minimum point data holding unit 6 into the target judgment limit value in order to binarize the chromaticity difference signal ΔC. Binarization threshold ΔCma obtained by addition
Output x. From the output terminal of the binarization circuit 73, the chromaticity difference signal ΔC read from the ΔC map frame memory 71 is output.
, The ΔC map binary signal BΔC smaller than the binary threshold value ΔCmax is set to “1” and output as a binary image.

In the configuration of the present apparatus, the luminance signal binarizing section 8 includes a luminance map frame memory 81, Ym, as shown in FIG.
ax, Ymin calculation processing circuit 82 and binarization circuit 83
It is composed of A luminance signal Y from the stimulus value converter 1 shown in FIG.
Is connected to the output terminal, and a binarizing circuit 8 is
3 is connected to the input terminal.

The input terminal of the Ymax / Ymin calculation processing circuit 82 is connected to the line of the reference luminance Yo from the reference luminance updating unit 10 shown in FIG.
3 is connected to a binarization threshold setting input terminal. An output terminal of the binarization circuit 73 outputs a luminance binarization signal BY.

The minimum minimum point luminance value Ypm is also ΔCpm
Is considered to be the target feature value in the same manner as
pm, a reference luminance value Yo is determined, and the luminance binarization limit values Ymax, Ym taking the characteristic of the target into consideration centering on this reference luminance value Yo
In, a binary image of the target area can be obtained with a high degree of accuracy by binarizing the value of Y in this range and taking a logical product with the ΔC map binary signal BΔC.

In order to binarize the luminance signal Y, the reference luminance updating section 10 shown in FIG. 1 predicts or converts the luminance signal Y from the target minimum minimum point luminance value Ypm held in the minimum point data holding section 6 shown in FIG. Reference luminance Yo obtained by filtering processing
In the Ymax and Ymin calculation processing circuit 82,
A luminance binarization limit value Ymin obtained by adding the target determination limit upper limit width in consideration of the target characteristics from this Yo and a target binarization limit lower limit value Ymin is obtained.

In the binarizing circuit 73, Ymax, Ym
The luminance image binary signal BY is set to “1” for the Y value read from the luminance map frame memory 81 in the range between “in” and is output as a binary image.

Hereinafter, the operation of the color image target position detecting device of the present invention will be described with reference to FIGS.

First, three received color image signals R, G,
B is input to the stimulus value converter 1 at the same rate as the output of the sensor one pixel at a time,

[0092]

(Equation 13) Are sequentially output.

This output is output to the chromaticity coordinate calculator 2 as X, Y, and Z stimulus values. Note that, among these three stimulus values, the stimulus value Y corresponds to the stimulus value of the primary color representing the luminance of the image. These X, Y, and Z signals are summed up by the three-input addition circuit 21 of the chromaticity coordinate calculation unit 2 shown in FIG.

The X and Y signals are respectively supplied to a logarithmic compression circuit 2
3 and input to a logarithmic compression circuit 24, and a two-input subtraction circuit 2
5 and 26. The two-input subtraction circuits 25 and 26 subtract the output value of the logarithmic compression circuit 22 from the input value,

[0095]

[Equation 14] Get. These x 'and y' signals are output by an exponential function generator 2
7, 28, where

[0096]

(Equation 15) Is obtained as an output signal from the chromaticity coordinate calculator 2. The values of the chromaticity signals x and y are

[0097]

(Equation 16) Of the color image signals R, G, B
Shows the chromaticity coordinate points for.

On the other hand, the chromaticity signals x and y from the chromaticity coordinate calculator 2 and the stimulus value Y from the stimulus value converter 1 are both input to the designated point data holding unit 13, and when the target is designated here X, y and the stimulus value Y of the pixel that coincides with the initial capture designated point Gto given to the target point, and update them for each image frame to obtain the designated point chromaticity signals xto, yto,
The designated point data holding unit 13 outputs the designated point stimulus value Yto.

The designated point chromaticity signals xto and yto are input to the reference chromaticity updating unit 4. The reference chromaticity update unit 4 receives the specified point chromaticity signals xto and yto as they are when the setting completion timing signal To indicating that the initial capture designated point Gto has been set as the initial capture point is received. 4 as initial reference chromaticity signals xo and yo.

The specified point stimulus value Yto is input to the reference luminance updating unit 10 and is set as the initial reference luminance signal Yo of the reference luminance updating unit 10 when the setting completion timing signal To is received. In addition, the evaluation area generation unit 12 generates a detection area signal WG having an initial capture window WGto given at the time of receiving the setting completion timing signal To as an initial value.

Next, the values of the chromaticity signals x and y from the chromaticity coordinate calculator 2 are subtracted from the xo and yo by the two-input subtraction circuits 31 and 32 of the chromaticity difference calculator 3 shown in FIG. Value conversion circuit 3
After obtaining the absolute values at 3 and 34, they are added by a two-input adding circuit 35,

[0102]

[Equation 17] Is obtained at the same rate as the image signal from the sensor.

This chromaticity difference signal ΔC is input to the minimum point detection unit 5, and every time a minimum point of ΔC is detected, a minimum point detection timing signal pm indicating the occurrence point is stored in the minimum point data holding unit 6. Is output.

In the minimum point data holding unit 6, the chromaticity difference signal ΔC from the chromaticity difference calculating unit 3, the chromaticity coordinate signals x and y from the chromaticity coordinate calculating unit 2, and the stimulus value converting unit 1 Although the luminance signal Y (Y stimulus value) is input, the minimum point detection signal p
At the time when m comes, it is determined whether ΔC smaller than the previously held chromaticity difference signal ΔC has been input, and if a smaller value has been input, all the input signals are updated / held.

Therefore, at the time when the input of one frame on one screen is completed, ΔC, x, y, and Y of the minimum point having the minimum value of the chromaticity difference signal ΔC are stored in the minimum point data holding unit 6. Are retained, and these are respectively represented by ΔC
Output as pm, xpm, ypm, Ypm.

The signal of ΔCpm from the minimum point data holding unit 6 is a binary threshold value ΔCmax obtained by adding a target determination limit value modeled from a target characteristic by the chromaticity difference signal binarization unit 7 shown in FIG. Calculated by the calculation processing circuit 72, and ΔCmax is 2
The value is set in the value conversion circuit 73.

Similarly, the luminance signal binarizing section 8 shown in FIG. 5 adds, based on the reference luminance signal Yo from the reference luminance updating section 10, a target determination limit upper limit width in consideration of target characteristics. The luminance binarization limit value Ymin obtained by subtracting the luminance binarization limit value Ymax and the target determination limit lower limit width is determined as a binarization range by the Ymax and Ymin calculation processing circuit 82 and set in the binarization circuit 73.

When the setting of ΔCmax, Ymax, and Ymin is completed, the map data of ΔC for one frame is stored in the ΔC map frame memory 71 shown in FIG. 4, and the map data of Y is stored in the luminance map frame memory 81 shown in FIG.
, And two binarized image signals BΔC and BY for ΔC and Y are obtained by the binarization circuits 73 and 83, respectively.

In the evaluation area generating unit 12, if the detection area signal WG having the initial capture window WGto given at the time of receiving the setting completion timing signal To as an initial value is within the area on the corresponding image frame, A binarized area signal w which is "1" is generated and output. The binarized area signal w and the binarized image signal BΔ
C and BY are ANDed by an AND circuit 9 to obtain a final target binary image B.

The obtained binarized image B is subjected to a labeling process in the binarized image centroid calculating section 11 to discriminate / separate into a plurality of independent binarized areas. , One of the binarized areas is selected, the surface centroid is calculated, and output as the target detection position G from the apparatus.

At the time of moving to the processing of the next frame,
Before starting the input of the next image frame, the reference chromaticity updating unit 4 inputs the latest target chromaticities xpm and ypm held in the minimum point data holding unit 6, and stores the currently held reference chromaticity value xo. , Yo, the reference chromaticity of the next frame is predicted or filtered to obtain a new reference chromaticity value x.
o and yo.

Similarly, the reference luminance updating section 10 uses the target luminance Ypm held in the minimum point data holding section 6 to set a new reference luminance Y by prediction or filtering.
Update to o.

The evaluation area generation unit 12 sets the target detection position G as the center position based on the target detection position G and the binarized area S from the binarized image centroid calculation unit 11, and sets the vertical and horizontal directions of the initial capture window WGto. WGto while keeping the ratio of
The detection area signal WG is updated so that the ratio of the binarized area S to the area area is constant, and the binarized area signal w corresponding to the WG updated in the next frame can be generated. .

The above description is based on the color image signals R,
By repeating from the input of G and B, the target of the initial capture designated point when the setting completion timing signal To is first received is detected, and the position is output for each frame. When the setting completion timing signal To is next input, the initial capture designated point Gto given at that time is input.
Goals become new goals.

In the above embodiment, the minimum point data holding unit 6
Holds the data group corresponding to the minimum point having the minimum ΔC for only one point and outputs it as information for tracking the target. However, as another embodiment, the minimum point data holding unit 6 can also hold the position coordinates of the minimum point on the screen, and can be configured to hold a data group for a plurality of minimum points in the form of a list.

In this configuration, if the chromaticity to be searched is set as the target designated point chromaticity, the coordinates of the target candidate points to be searched on the screen are ranked in ascending order of a plurality of ΔC, and the data group is set. And a so-called target candidate search device that can output these.

[0117]

According to the present invention, three image signals R, G, and B of red, green, and blue are converted into tristimulus signals X, Since the signals x, y of the chromaticity coordinate points which are directly related to the spectral characteristics of the target reflected light are calculated from the tristimulus value signals X, Y, Z, the signals are converted into Y, Z. Accurate target detection without being caught can be performed.

Furthermore, since the values of x and y are always less than or equal to "1", the scaling of the processing values required for the circuit design in the subsequent processing is facilitated. Further, since the Y stimulus value obtained by the stimulus value conversion unit corresponds to the brightness of the color light, it is more accurate to combine the chromaticity values, x, and y with the Y stimulus value as it is for target detection. Detection judgment can be performed.

Further, according to the present invention, the chromaticity coordinate calculation unit calculates x,
Since a method of subtraction after logarithmic compression can be used for the operation for obtaining y, the dynamic range of the stimulus value that can be handled for obtaining x and y can be extremely widened, and the division circuit can be eliminated. Has an effect.

Further, the chromaticity difference calculating section 3 converts the positional deviation of the captured image from the coordinate reference in the chromaticity space into one signal value called distance information based on the target chromaticity value to be detected. Since the conversion is performed, the target point can be set as the target point detection point only by detecting the point at which this value is minimum, and the scale of the processing circuit can be reduced. Furthermore, when obtaining the distance information of the positional deviation from the coordinate reference on the chromaticity space, the chromaticity difference is calculated, so that the distance information can be easily obtained without calculating the square operation and the square root. , Always ΔC
Since there is a relationship of ≧ r, it can be obtained as a value of a chromaticity difference larger than the value of r.

Further, according to the present invention, since the reference chromaticity updating unit 4 is provided, the target distance changes with time over time,
Even when the imaging chromaticity changes due to the influence of atmospheric scattered light and the spectral characteristics of transmittance, tracking of the reference chromaticity is possible, and the effect of suppressing the influence of the imaging distance is provided.

The final target position is calculated by the binarized image center-of-gravity point calculation unit 11, which outputs the result as the surface centroid after selecting the binarized area by the target motion prediction determination. It is also possible to determine the intersection of objects of the same color, and the detection positions are averaged within the area, so that the fluctuation of the detection point position or the width of the jump can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a block diagram showing an example of a chromaticity coordinate conversion unit 2 in FIG.

FIG. 3 is a block diagram illustrating an example of a chromaticity difference calculator 3 in FIG.

FIG. 4 is a block diagram illustrating an example of a chrominance difference signal binarization section 7 in FIG. 1;

FIG. 5 is a block diagram showing an example of a luminance signal binarizing section 8 in FIG. 1;

FIG. 6 is a block diagram showing an example of the connection used in FIG. 1;

FIG. 7 is a diagram for explaining the operation of each block in FIG. 1;

FIG. 8 is a diagram for explaining the operation of the present invention.

FIG. 9 is a diagram for explaining the operation of the present invention.

FIG. 10 is a diagram for explaining the operation of the present invention.

FIG. 11 is a diagram for explaining the operation of the present invention.

FIG. 12 is a diagram for explaining the operation of the present invention.

FIG. 13 is a diagram for explaining the operation of the present invention.

[Explanation of symbols]

Reference Signs List 1 stimulus value conversion unit 2 chromaticity coordinate calculation unit 3 chromaticity difference calculation unit 4 reference chromaticity update unit 5 minimum point detection unit 6 minimum point data holding unit 7 chromaticity difference signal binarization unit 8 luminance signal binarization unit Reference Signs List 9 AND circuit 10 Reference luminance update unit 11 Binary image centroid calculation unit 12 Evaluation area generation unit 13 Designated point data holding unit 21 3-input addition circuit 22, 23, 24 Logarithmic compression circuit 25, 26 Subtraction circuit 27, 28 Exponent Function circuits 31, 32 Subtraction circuits 33, 34 Absolute value conversion circuit 35 2-input addition circuit 71 ΔC map frame memory 72 ΔCmax calculation processing circuit 73 Binarization circuit 81 Luminance map frame memory 82 Ymax / Ymin calculation processing circuit 83 Binarization Circuit 100 Color image target detection device 101 Color image sensor 102 Initial capture designation input device 103 Image / designation position display device 104 Gimbal device 105 Tracking signal output device 106 Imaging target 107 Display screen 108 Tracking target 201 Chromaticity range by RGB signal of color image sensor 202 Chromaticity coordinate of input image 203 Target specified point chromaticity coordinate (reference chromaticity coordinate) 301 x + y = 1 Straight line 302 = 1 Limit line satisfying ΔC ≦ 1.0 based on 0 point 401 ΔC map 402 Minimum point 403 pm generation point 404 pm Generation range designation area WG 501 Imaging screen 502 Signal output order

 ──────────────────────────────────────────────────続 き Continued on the front page F-term (reference)

Claims (3)

[Claims] 1. An image signal R, G, B from a color image sensor that captures a target color image and outputs three color signals of red (R), green (G), and blue (B) is expressed by 1) And a stimulus value conversion unit that converts the stimulus value signals into X, Y, and Z stimulus value signals in real time for each pixel output from the sensor in response to a color signal input; The tristimulus value signal of A chromaticity coordinate computing unit that converts the chromaticity coordinate signals into x and y by calculating the following chromaticity coordinate signals x and y:
From, A chromaticity difference calculating section for calculating the chromaticity difference ΔC by performing the above calculation; and detecting a point at which the chromaticity difference ΔC is minimized in a detection area set by a detection area signal from the evaluation area generating section. A minimum point detection unit that outputs a minimum point detection timing signal pm, and a chromaticity difference ΔC of the minimum point and a chromaticity coordinate corresponding to the chromaticity difference ΔC of the minimum point each time the minimum point detection timing signal pm is generated. The signals x and y are held, and when the color signal input of one frame is completed, the smallest one of the chromaticity differences ΔC of the minimum points and the chromaticity coordinate signals x and y corresponding to the minimum ΔC are converted to the minimum minimum points. Chromaticity difference ΔCpm and minimum minimum point chromaticity value xpm, y
a chromaticity difference signal for inputting the chromaticity difference ΔC and the minimum minimal point chromaticity difference ΔCpm, and converting the chromaticity difference ΔC into a binary chromaticity difference BΔC AND for outputting a logical product signal B of a binarization unit and a binarization window area signal w for limiting the binarization output image area output from the evaluation area generation unit, A binary image center-of-gravity point calculation unit that receives the AND signal B and calculates a center-of-gravity point G of the binary image and a binarized area S whose value is the number of pixels where B = 1 A designated point data holding unit that selects and holds the chromaticity coordinate signals x and y of the pixel that coincides with the initial capture designated point and outputs the designated point chromaticity signals xto and yto; and the designated point chromaticity signals xto and yto And the minimum minimum point chromaticity values xpm and ypm, and the reference chromaticity signal o,
and a reference chromaticity updating unit for outputting yo. 2. The minimum point data holding unit further holds a stimulus value signal Y corresponding to the chromaticity difference ΔC of the minimum point each time the minimum point detection timing signal pm is generated, and stores a color of one frame. Means for outputting a stimulus value signal Y corresponding to the minimum ΔC among the chromaticity differences ΔC of the minimum points as a minimum minimum point stimulus value signal Ypm at the time of completion of the signal input, and the designated point data holding unit further comprises: Means for selecting and holding a stimulus value signal Y of a pixel which coincides with the initial capture designated point, and outputting the selected value as a designated point stimulus value signal Yto. Further, the minimum minimal point stimulus value signal Ypm and the designated point A stimulus value signal Yto, a reference brightness update unit that outputs a reference brightness signal Yo, and a stimulus value signal Y and the reference brightness signal Yo,
A luminance signal binarization unit that converts the stimulus value signal Y into a binarized luminance signal BY and outputs the binarized luminance signal BY to the AND circuit, wherein the AND circuit outputs the binarized chromaticity difference BΔC and the luminance signal BY 2. The color image target position detecting device according to claim 1, wherein a logical product signal B of the binary image signal and the binarized window area signal w is output. 3. The color coordinate conversion section converts the chromaticity coordinate signals x, y into 3. A color image target position detecting device according to claim 1, further comprising a calculating means for obtaining the target position.