JPS60222981A - Object recognizing device - Google Patents

Abstract

PURPOSE:To discriminate a target object and its background even in case a luminance difference of both of them is not sufficient, and to recognize exactly the object by providing a color extracting device, a picture emphasizing device and a color information setting device. CONSTITUTION:Picture data of each separate color inputted to memories 7, 8 and 9 through a color signal separator 2, etc. from a color TV camera 1 are outputted to a color extracting device 10. The device 10 converts an input picture data so that it becomes color information which is not generated by the brightness, and outputs it to a picture emphasizing device 11. The device 11 forms a picture data related to whether both color information is similar or not from the color information inputted from the device 10 and the color information inputted from a color information setting device 12. This picture data becomes a larger value by the extent that the input picture is different from a color of a pattern picture set in advance to the setting device 12, and becomes smaller value as it becomes approximate. In this way, a difference of the pattern picture and its background is emphasized, and even in case when a luminance difference of a target object and its background is not sufficient, both of them can be discriminated, and the object can be recognized exactly.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an object recognition device that recognizes the position of a target object using an imaging device such as a television camera.

Conventionally, a black and white chimp camera sound is used to obtain one screen worth of image data including a target object, and when detecting the position of the target object on the screen from this image data, the image data of the target 9 is referred to. By sequentially extracting image data in the same area as the reference data from the image data for one screen and determining the correlation between the extracted image data and the reference data, image data similar to the reference data is extracted. region,
That is, there is an object recognition device which detects the position [4] of a target object on the screen (Japanese Patent Laid-Open Nos. 57-57096 and 39180-1988).

However, since the 1Ill mound data is a value proportional to the brightness of the target object and its background, if there is no brightness difference between the target object and its background, or if the brightness difference is small, the above correlation value No clear peak value appears,
It was difficult to distinguish between the target object and the background, making it impossible to perform accurate object recognition.

The present invention has been made in view of the above-mentioned circumstances, and provides an object recognition device that can distinguish between a target object and its background even when the brightness difference between the two is insufficient, and that can perform accurate object recognition. The purpose is to

According to this invention, the difference in color tone between the target object and its background is used as information to distinguish between the two, and the color information obtained from the video signal from the color television camera and the color tone of the target object set in advance are used. Based on the information, the video signal from the color television camera is intensified with respect to the target object.
The target object's position is recognized from this one-screen image data.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram showing an embodiment of an object recognition device according to the present invention, in which a moving object is assumed as a target object, and a device for tracking the moving object is shown.

In Fig. 1, the color television camera 1 photographs a moving object with a predetermined fixed field of view.
The standard video 1g signal is output to the color 1g signal separator 2ν and the sync separator circuit 3. The color signal separator 2 inputs the input video-1
G is separated into the three primary colors of light, red), green←) and blue-violet (B), and these signals are sent to the A7D converter 4.5.
and 6 in addition to the condition. Note that when the color television camera is a three-tube type television camera, the color signal separator 2 is unnecessary, and the video signals output from the R, G, and B terminals of the television camera are sent to the respective A/D i exchanger 4, 5,
and 6 directly.

The synchronization separation circuit 3 separates a synchronization signal from the input video signal, and the A//D converters 4, 5, and 6 synchronize the video signals separated into each color with the synchronization signal from the synchronization signal separation circuit 3. and converts each pixel into image data for each pixel, and outputs this to memories 7, 8 and 9, respectively. In addition,
The above-mentioned transformers 4, 5 and 6 each convert the luminance level into image data of, for example, 16 gradations according to the luminance of the input video signal.

The image data for each color temporarily stored in memories 7.8 and 9 is applied to color extraction device d10.

The color extraction value ratio is used to convert the above image data for each color into color information independent of brightness. Now, input image data that has been decomposed into R, G, and B is R(X + y) r G (x ry) +
B (x ry ), the color extraction device 10 is
Color information r (X r 3') which shows these as the following formula
r g (X r Y ) * b (X * 7)
and output to the image enhancement device 11.

Note that the conversion is not limited to the above-described method, and the three input image data may be converted to be normalized.

The color information setting device 12 sets color information α, which is information similar to the above, of a pattern image P (see FIG. 2(b)) that is a part of the screen of the moving object and has a different tone from the surroundings.

β and γ are set in advance, and these pieces of information are output to other inputs of the image enhancement device 11.

The image enhancement device 11 extracts both colors from the color information r (x, yLg(x, y), b(x+y) input from the color extraction device 110 and the color information α, β, r manually input from the color information setting device 12. Image data f (x
+y). Specifically, the above image data is determined by (electronic rise) of (2).In addition, in place of the above (2) equation, the following equation, r(x+y )=lα−r (X + y Monthly β−g(x
, yl+1r−b(x, yl ++·++ (3)).

As is clear from the above equation (2) or equation (32), the image data f (x ry ) has a larger value as the input image differs from the color of the pattern 1IijI image P.
The closer the approximation is, the smaller the value becomes, and the difference between the parable turn image P and its background is emphasized.

The 1111111 data indicating f(x, y) output from this image enhancement device 11 is divided into 3 pieces according to the color, and the image data becomes 11 pieces of data again. The same processing as image processing using the 14 degree level can be performed.

Below, a pattern image P in which the VC moves as the moving object moves.
We will explain the case of tracking the movement of.

Now, the image data emphasized as described above is applied to the changeover switch 13. The changeover switch 13 alternately switches and connects the movable contact 13a to the contact 13b and the contact 13.cJ according to the signal applied from the synchronization separation circuit 3 at the end of each frame, and switches the input emphasized image data to the first one. 1 image memory 14 and the first image memory 14.

The first image memory 14 and the second image memory 15 each store image data for one screen, and the stored contents of each memory are alternately updated with image data for a new screen every time the selector switch 13 is switched. be done. therefore,
The image data between the first image memory 14 and the second image memory 15 is -, assuming that one frame time is 1/30 second.
&, a time difference of 14r constant time Δt (-t/30 seconds) for loading E occurs.

Note that image data for one screen is sent from the image enhancement device 11 every l frame time.

From one screen worth of images stored in the valley memory, an image (partial image) including the pattern image P, which is a part of the image, is extracted (see FIGS. 2(b) and 3). The positions of the partial images extracted from the first image memory 14 and the second image memory 15 are controlled by a memory controller 16. That is, the memory controller 16
inputs a signal indicating the center position coordinates (x ry) of the pattern image from the moving position calculation circuit 18 as a signal indicating the moving position of the pattern image, and calculates the position (x, y) based on this position signal. A control signal for extracting a partial image of a predetermined central pixel is output to the first image memory 14 and the second image memory 15.

Now, at time 1o, the 72-tan image is located at the position (xo
+yo ) to υ, the screen at this point is the first image memory 1
4, and the time tl (=to+Δ
The screen at 1.t) is recorded in the second image memory 15. Let's explain the case where it is. First, the first image memory 1
4. Project the screen from the image data stored in 4 on a monitor TV (not shown) and use a light ben etc.
Specify the position (IQ + Fo) of the mother turn image.

This position (zo.

The signal indicating 3'o) is input to the movement position calculation circuit 18 as an initial value and is applied to the memory controller 16 as is.

The memory controller 16 inputs a position (x.

y), the first image memory 6 and the -second
A control signal is output to the image memory 7.

The data of the partial images extracted from the first image memory 6 and the second image memory 7 are sent to the FFs of the correlation calculator 20, respectively.
T units 21 and 22, where they are subjected to two-dimensional Fourier transform, and the results from both images are multiplied for each corresponding frequency component in a multiplier 23, and this result is applied to the IFFT unit 2.
Inverse Fourier transform is performed in step 4.

The inverse Fourier transformed data is applied to the maximum value detector 17]1, where the data of the second image memory 15 is calculated for the partial image of the first image memory 14 where the cross-correlation of both partial directional image data is the highest. Partial image shift i (ΔX.

.DELTA.y)GAMEH-1 A signal corresponding to this is sent to the movement position calculation circuit 18 (see FIG. 3).

The movement position calculation circuit 18 adds hit (ΔX, Δy) to the initial value (XQl yo), and outputs a signal indicating this added value to the position (x+) of the turn image after Δ time.
output as a signal indicating y).

A signal indicating this point tf(x,y)k is applied to the memory controller 16. The memory controller 16 controls the first image memory 1 based on the input signal in the same manner as described above.
4 and the second image memory 15, respectively, to output control signals to extract partial images from the second image memory 15. In this case, the 113T movable contact piece 13a of the changeover switch 13 is switched to the contact 13b, and the first image memory 14 stores image data of a new screen.

From these partial images, the correlation between the two is calculated again, and the second image memory 150 has the maximum correlation.
Portion 1IiIII&! of image memory 14 deviation 1i (Δ
X, Δy) and add this to the current movement position (x * y) to find a new movement position.

A moving object can be tracked by repeating and executing the above processing.

Note that in this embodiment, the target object and its background are distinguished only by the difference in color tone, but the difference in brightness level is also taken into account to clearly distinguish the target object and its background. be able to.

In this case, for example, if the higher the brightness is, the higher the level of image data is, and the lower the brightness is, the image data is of a lower level, an object with lower brightness than the background is selected as the target object, and such image data is sent to the image enhancement device. What is necessary is to superimpose it on the output image data.

In addition, the image data formed by the present invention that emphasizes the difference in color tone is not limited to the method of tracking a moving object in the above embodiment, but can also be obtained by detecting the outline of a target object from a point where the image data suddenly changes, for example. This can be applied to various image processing methods such as object recognition methods.

As explained above, according to the present invention, even if the luminance difference between the target object and its background is not sufficient, if there is a color difference between the two, image data that emphasizes this can be obtained.
6. Can easily recognize objects.

[Brief explanation of the drawing]

Figure 1 is a diagram showing an embodiment of the object recognition device according to the present invention, Figure 2 (a) is a diagram showing an example of a screen imaged by a color television camera, and Figure 2 (b) is a diagram showing an example of a screen imaged by a color television camera. 2 is an enlarged view of a partial image extracted from FIG. 2(a), and FIG. 3 is a diagram showing a partial image used for detailed explanation of the present invention.

Claims (5)

[Claims] (1) A color television camera that photographs a target object in color in a predetermined field of view, a color information setting device in which color information of the target object is set, and color information obtained from the image No. 16 from the color television camera. a color filter device that converts the video signal into image data that emphasizes the target object based on the set color information; and recognizing the target object using one screen worth of image data from the color filter device. An object recognition device comprising recognition means. (2) The object recognition device according to claim (1), wherein the color information set in the lust setting device is information indicating the level of each color of R, G, and B in the original three primary colors. . (3) The color filter device includes a color extraction device that converts the video signal from the color television camera into a color for each color independent of brightness, and a color extraction device that converts the video signal from the color television camera into a color 1 * color obtained from the color extraction device. and an image strength 11 device that forms image data related to the similarity of both color information from the color information set by the color information set by the color information set by the =+1 color information setting device. recognition device. (4) The recognition means is configured to combine the image data of 11111 screens from the color filter device or the image data of a predetermined area in one screen with the image data of the target object or the image data of a predetermined area including the target object. The object recognition device according to claim 1, which calculates the cross-correlation and recognizes the position of the target object from the position on one screen where the correlation value is maximum. (5) The recognition means detects the contour of the target object from a level change in the image data from one screen of image data from the color filter device, and thereby recognizes the object. The object recognition device described in section 1).