JP3480591B2 - Object tracking device - Google Patents

Description

Detailed Description of the Invention

[0001]

[Table of Contents] The present invention will be described in the following order. Field of Industrial Application Conventional Techniques Means for Solving Problems to be Solved by the Invention (FIG. 1, FIG. 10, FIG. 26) Action (FIG. 1, FIG. 10, FIG. 26) Embodiment (1) First Example (FIGS. 1 to 9) (2) Second Example (FIGS. 2, 5 to 8, and 10 to 2)
5) (3) Third embodiment (FIGS. 15 to 22, and 25 to 3)
4) (4) Effect of the invention of other embodiments

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an object tracking device, and can be applied to, for example, a device for tracking a target object by moving a television camera according to the motion of the target object.

[0003]

2. Description of the Related Art Conventionally, when a television camera is moved in accordance with the movement of a target object to track the target object, first, the movement of the target object in the image captured by the television camera is detected. A block matching method is generally used as this motion detection method. The block matching method divides the image frame into multiple blocks, calculates the correlation between the target block and the next temporally successive frame, and determines the motion with the block with the smallest error as the corresponding block. To do.

In the motion detection by this block matching method, since the matching is carried out without separating the target object from other objects or the background, the final judgment is decided by a statistical amount. That is, even if the correct correspondence cannot be obtained, the motion is determined by selecting the block having the smallest error.

[0005]

However, in the motion detecting method using the block matching method as described above, the amount of calculation is large because the correlation between the target block and the next frame that is continuous in time is calculated. Further, there is a problem that the movement of the patterned object cannot be detected.

Further, since the simple block matching detects the motion by the correlation between the target block and the next frame which is continuous in time, it cannot cope with the size change of the object and the target object is specified. Not necessarily. Therefore, there is a problem that even if the object is not in the field of view, the minimum location of the correlation calculation is output as the corresponding block and the wrong movement is detected.

The present invention has been made in consideration of the above points, and it is an object of the present invention to propose an object tracking device capable of detecting the movement of an object with high accuracy and reliability and automatically tracking the target object. .

[0008]

In order to solve such a problem, in the present invention, a designated area WA set in the screen is set.
In the object tracking device 1 for tracking the target object in the KU,
An image capturing unit 2 that captures an image of an object to generate a video signal, a histogram processing unit 12 that calculates a histogram of a video signal in a designated area WAKU on the initial screen, and a spatiotemporal image with respect to a high-frequency histogram range. The spatiotemporal image processing means 11 and the control means 13 for controlling the photographing means 2 based on the trajectory of the spatiotemporal image are provided.

Further, according to the present invention, in the object tracking device 1 for tracking the target object in the designated area WAKU set in the screen, a histogram processing means for calculating the histogram of the video signal in the designated area WAKU on the initial screen. 1
2, the spatio-temporal image processing means 11 for obtaining a spatio-temporal image and the control means 13 for controlling the designated area based on the locus of the spatio-temporal image are provided for the high frequency range of the histogram.

[0010]

When the tracking target object is placed in the designated area WAKU set on the screen and the tracking is started, the histogram in the designated area WAKU of the video signal is calculated, and the spatiotemporal image is obtained in the high frequency range of the histogram. By controlling the movement of the photographing means 2 or the designated area WAKU according to the locus of the spatiotemporal image obtained as a result, it is possible to detect and track the movement of various objects more reliably than the conventional block matching. You can

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings.

(1) First Embodiment FIG. 1 shows an object tracking device 1 according to the present invention as a whole.
The video signal from the television camera is output from the taking optical system 2 including the lens 2A, the zoom lens 2B, the diaphragm 2C, and the CCD image pickup device 2D. The video signal passes through the AGC circuit 3 and is converted into a digital signal by the analog-digital conversion circuit (A / D) 4, Y / C separated by the digital camera processing circuit 5, and a luminance signal is passed through the digital-analog conversion circuit (D / A) 6. It is transmitted as Y and chroma signal C. The luminance signal Y and the chroma signal C are supplied to a viewfinder (not shown) for display.

Further, the luminance signal Y and the color difference signals R-Y and B- obtained by Y / C separation in the digital camera processing circuit 5.
Each Y passes through a low pass filter (LPF) 7 and is image-compressed by a variable image compression circuit 8. At this time, the luminance signal Y and the color difference signals RY and BY are 1/4 or 1/8.
And the size is actually 256 × 256 pixels or 128.
× Compressed to the size of 128 pixels. This switching is controlled by a control signal S1 from the system control circuit 13 having a microprocessor configuration. Of the luminance signal Y and the color difference signals RY and BY compressed in this way, the Cartesian coordinate system color signals RY and BY are saturated (Satu).
ration) / hue (Hue) detection circuit 9 through polar coordinate system Sa
Converted to turation and Hue. The Y signal, H signal, and S signal for one field output in this manner are stored in the memory 1
It is temporarily stored at 0.

The Y signal, H signal, stored in the memory 10,
The S signal is input to the histogram processing circuit 12 and the spatiotemporal image processing circuit 11, and the color and luminance signal histograms in the area designated for tracking and the x- and y-direction spatiotemporal image for the color designated for tracking. And luminance signal x, y
Directional spatiotemporal images are generated. System control circuit 13
Calculates the motion vector and deformation amount in the x and y directions from the spatiotemporal image in the x and y directions for the color.

The system control circuit 13 also outputs a panning motor drive signal PAN, a zoom lens motor drive signal ZOOM, and a tilt motor drive signal TILT on the basis of the x- and y-direction motion vectors and the deformation amounts thus obtained. Output. These motor drive signals PAN,
ZOOM, TILT panning motor 14,
The zoom lens motor 15 and the tilting motor 16 operate so that the target object can be automatically tracked. Alternatively, the focus lock, AE, or the like can be controlled by focusing on the target object. The signal S2 from the system control circuit 13 to the memory 10 is a memory control signal for setting the area designation frame and adaptively changing the color area and the brightness area.

The system control circuit 13 also supplies information S3 of coordinates and size of the area designating frame to the frame signal generating section 41. The frame signal generator 41 generates an area designation frame signal S4 based on the information S3. The area designation frame signal S4 is superimposed on the luminance signal Y output from the digital-analog conversion circuit 6, and is displayed together with the captured image on a visual finder (not shown).

Here, in the case of this embodiment, the saturation (Satura
/ Hue detection circuit 9 converts the rectangular coordinate system color signals RY and BY into the polar coordinate system Saturation and Hue. In practice, the rectangular coordinate system color signals RY and BY. The relationship between Y and the polar coordinate system (HSI color model) Saturation and Hue is expressed as shown in FIGS. 2 (A) and 2 (B). In addition, u shows the value of RY, V shows the value of BY. Here, the dynamic range of the rectangular coordinate system color signals RY and BY is (-128, 127), and the dynamic range of the polar coordinate systems S and H is (0, 255).

[Equation 1] Is converted by.

In this embodiment, first, the user inputs the input section 40.
Using the cursor keys or the like (not shown), move the area designating frame on the view finder to fit the target object to be tracked into the area designating frame. Then, when the user presses a confirmation key or the like (not shown) of the input unit 40 at a desired position, the area designation frame on the initial screen is determined. The system control circuit 13 supplies the coordinates and size of the area designating frame on the initial screen to the memory 10 as the memory control signal S2. The target object in this embodiment is the area designation frame WAKU.
The red car inside.

For reference, the color distribution in this area designation frame is shown in FIGS. 4 (A) and 4 (B). FIG. 4A shows the case in a rectangular coordinate system, and FIG. 4B shows the case in a polar coordinate system. In this embodiment, the histogram processing circuit 1
Hue histogram HST obtained in 2 (Fig. 4 (C))
Is used to specify the target object. It is assumed that the most frequent color area in the Hue histogram HST (FIG. 4C) in the area designation frame WAKU is the color of the target object, and the maximum range of the color area is limited to 20 at this time. This corresponds to about 28 degrees at the angle θ of the polar coordinate system. In practice, if the range of the color area is not limited, a plurality of colors will be tracked.

[Equation 2] Is limited to the range.

Further, in the case of this embodiment, the spatiotemporal image processing circuit 11 performs the video information 3 as shown in FIG.
Assuming a three-dimensional space-time rectangular parallelepiped, a space-time image is generated from the three-dimensional space-time rectangular parallelepiped. Generally, the spatiotemporal image is formed by arranging continuously input images in the time axis direction to form a three-dimensional space-time rectangular parallelepiped, and adding and averaging the images of the three-dimensional space-time rectangular parallelepiped at a certain time in the x direction and the y direction. Is calculated and arranged in the time direction. At this time, a light and shade stripe pattern appears in the spatiotemporal image according to the movement of the camera and the object.

Actually, the value i of the luminance signal of each pixel in the three-dimensional space-time rectangular parallelepiped is

[Equation 3] The Hue value h and the Saturation value s of each pixel in the three-dimensional space-time rectangular parallelepiped are expressed by

[Equation 4] It is represented by. Therefore, from the color region range (h1≤h≤h2, h1 = 65, h2 = 85 in this embodiment) determined by the color determination of the target object, the color spatiotemporal image (Hue) is shown in FIG. ) And as shown in FIG. 5 (C),

[Equation 5] And for the y direction

[Equation 6] Ask in.

The color spatiotemporal image at each time thus obtained is standardized in the color region range as shown in FIG.
In this embodiment, the spatiotemporal image based on the luminance signal is not used. The spatiotemporal image thus obtained is represented as shown in FIG. FIG. 7A shows a spatiotemporal image based on a luminance signal, and FIG. 7B shows a spatiotemporal image based on a color signal. In this example, the camera is fixed and the lens is pulled wide to photograph the target object, which is a red car, moving at a slow speed.

When a spatiotemporal image is taken on the entire screen, the amount of information is compressed too much and it is difficult to obtain a correct result. Therefore, in practice, the area is divided into eight in each of the x direction and the y direction, and the spatiotemporal image is generated in each of the areas. As an example, when the screen is divided into four parts in the x direction and the y direction, the result is as shown in FIG. The block in which the current subject is present can be obtained from the position of the designated frame that is designated or automatically controlled, and the amount of movement and the amount of deformation can be calculated using the information of the block. FIG. 7 described above is a spatiotemporal image of 3 blocks in which the subject exists within the 8 divisions.

Further, in the object tracking device 1 of this embodiment, the system control circuit 13 executes the procedure as shown in FIG. 9 for the color spatiotemporal image thus obtained, and the motion amount of the target object and The amount of deformation is detected. That is, FIG. 9 shows x
This is the process performed for the spatiotemporal image in the direction (FIG. 9A), and the same process is performed for the y direction.

First, the initial screen (t = 0, t = t-1 (see FIG. 9
Regarding (B)), the non-zero coordinate area (p0, p1) within the range of the area specification frame coordinates (w0 (t), w1 (t)) initialized in the color space-time image is integrated. Then

[Equation 7] The width width (t) represented by is the initial state of the subject.
By doing this, the distances d0 (t) and d1 (t) to the coordinates at both ends and the area designation frame are

[Equation 8] It is represented by. In this equation (8), if p0 <w0 (t), d0
(t) = 0, and if p1 <w1 (t), d1 (t) = 0. In reality, both the movement amount and the deformation amount are 0 on the initial screen.

Next, in the continuation screen (t = t (FIG. 9C)), a non-zero area connected to the subject width of t = t-1 is determined, and the coordinates at both ends are (p2, p3). .
The subject width at this time is

[Equation 9] The amount of deformation is calculated by

[Equation 10] Ask more.

Next, from the deformation amount obtained from the equation (10), a new area designation frame coordinate is calculated by the following equation.

[Equation 11] Further, the size and movement of the area designating frame can be tracked in this manner. This motion vector is

[Equation 12] You can ask more.

The system control circuit 13 uses the motion vector mvx (t) obtained by processing in the x direction,

[Equation 13] Further, a panning motor drive signal PAN for controlling the panning motor 14 is generated. Similarly, using the motion vector mvy (t) obtained from the y direction, the following equation

[Equation 14] Further, a tilt motor drive signal TILT for controlling the tilting motor 16 is generated. Furthermore, based on the change in the size of the area frame,

[Equation 15] The zoom lens drive signal ZOOM for controlling the zoom lens motor 15 is generated. In the equations (13) to (15), the coefficients A, B and C are values according to the state of the camera.

In this embodiment, the motion vector in the x and y directions and the size of the frame are detected from the coordinates of the new area designating frame, and the panning motor 14, tilting motor 16,
Although the zoom lens motor 15 is driven, the photographing optical system 2 is not driven as described above, and only the movement and enlargement / reduction of the area designation frame can be performed. In this case, the system control circuit 13 supplies the information S3 of the new area designation frame coordinates to the frame signal generator 41. The frame signal generation unit 41 determines the area designation frame signal S4 based on the new area designation frame coordinates.
To update.

According to the above construction, after the Y / C separation of the video signal taken in from the photographing optical system, the color of the target object is estimated from the color histogram of the tracking area, and the spatiotemporal image of that color is x and y. Compared with the conventional block matching, the motion amount and deformation amount in the x and y directions are detected from the trajectory and the relative size of the trajectory and the width and initial width of the trajectory. Thus, it is possible to realize an object tracking device that can reliably detect the amount of movement and the amount of deformation of even a patterned object with a small amount of calculation.

Further, according to the above configuration, the drive signal for controlling the camera lens, the camera body and / or the zoom lens is generated based on the movement amount and the deformation amount in the x and y directions. In addition to tracking the subject, it is possible to automatically respond to changes in the size of the subject, and it is possible to realize an object tracking device that can always shoot the target object with a preferable size, and it is easy to use Can be significantly improved.

Further, according to the above configuration, the area designation frame is moved based on the movement amount and the deformation amount in the x and y directions, and the size thereof is controlled. It can be tracked and can be applied to partial image processing for a target object.

(2) Second Embodiment In FIG. 10 in which parts corresponding to those in FIG. 1 are designated by the same reference numerals, 20 indicates an object tracking device according to a second embodiment of the present invention as a whole, and In the example, the saturation (Saturati
on) / hue detection circuit 9 outputs polar coordinate system
Of Saturation and Hue, the H signal having a hue is stored in the memory 10 through the color filter 21.

Y signal, H signal stored in the memory 10,
The S signal is the spatiotemporal image processing circuit 2 for each.
2, 23, and 24, the color and luminance signal histograms within the designated area frame for tracking are obtained, and the upper n-th color with a large output of the color filter 21 and the most frequent luminance range are targeted objects. It is assumed that Then, the spatiotemporal image in the x direction and the y direction for the color and luminance signal of the target object is obtained.

After the system control circuit 13 identifies the object from these spatiotemporal images, the panning motor drive signal PAN for tracking the target object is calculated by calculating the motion vector in the x and y directions, the state discrimination, and the size. , Zoom lens motor drive signal ZOOM, tilt motor drive signal TI
Output LT. These motor drive signals PAN, ZO
The panning motor 14, the zoom lens motor 15, and the tilting motor 16 operate in the OM and TILT, respectively, so that the target object can be automatically tracked.

In the case of this object tracking device 20, the saturation (Sa
As shown in FIG. 11, the turation) / hue (Hue) detection circuit 9 has u and v values (u and v are R−Y, respectively).
(Indicates values of BY)) to ROM-configured h matrix 9A
And conversion is performed by table lookup using the S matrix 9B. As a table used at this time, the distribution of u or v as shown in FIG. 12 is assumed to be an exponential function distribution.

Then, in order to reduce the memory capacity, the number of quantization levels is reduced from 128 to n steps. As shown in FIG. 13, two u-v planes formed by the u and v axes requantized in this way are prepared for H and S. For each part of the conversion map, the average value in the area is calculated in advance by the above-mentioned equation (1).

In the case of the object tracking device 20, the color filter 21 limits the color range of each of the H signals obtained from the u and v signals by the saturation / hue detection circuit 9 and limits each color. The integrated value of the histogram within the range is output.

Actually, the H signal is two-dimensionally arranged as shown in the video signal color plan view of FIG. Therefore, the colors are classified in the Hue area to configure the color filter 21. As the colors to be classified here, red, blue, green, yellow, cyan, purple, and flesh color are used as the colors generally seen in nature. Empirically, the color range is purple (Mg), red (R), flesh color, yellow (Y), green (G), cyan (C).
20-40 and 65-8 for y) and blue (B) respectively
5, 90-110, 115-135, 150-170, 195-215 and 225-24
Decided to 5.

Here, in the case of determining the color of the target object which is a red car for an image as shown in FIG. 15 as an initial image, for example, the Hue histogram described above is multiplied by the color filter 21 to determine the area of the histogram in each color range. Is output, and the color regions up to the upper nth place of the output (for example, n = 2) are assumed to be the color of the target object. Actually, the color filter 21 is provided with a gear gap between the respective colors so that the influence of the intermediate color can be effectively removed by the gear gap.

The spatiotemporal image is obtained in the same manner as in the first embodiment. However, in this embodiment, if the spatiotemporal image is taken on the entire screen, the amount of information will be compressed too much and it will be difficult to obtain a correct result. As shown in FIG. 15 (B), the image area is limited to a strip area having a width slightly larger than the designated area frame WAKU in each of the x and y directions, and the spatiotemporal image as shown in FIG. 16 is obtained in the area. Has been done.

Further, in the case of this embodiment, the system control circuit 13 is adapted to obtain the time difference for the color spatiotemporal image thus obtained. In this way, the stationary part of the color spatiotemporal image is deleted, and the information of only the changed part is extracted. At this time, slight movements and edges are extracted as noise, so moving areas of a certain size or smaller are removed from the difference image.

In the above configuration, when actually starting the tracking of the target object, the photographer must first inform the system of the object to be tracked. Therefore, FIG.
As shown in (A), similarly to the case of the first embodiment, an object to be tracked is determined in the area designation frame WAKU in the initial screen, and tracking is started. By doing so, as shown in FIG. 17, the position of the target object can be known from the coordinates X0 and X1 of both ends in the x and y directions of the area designation frame WAKU.

The area to be searched is set to this range, and the coordinates (s0, s1) of the feature quantity within this range are obtained from the color-time space image and the color-time space difference image of the initial screen, and the following equation is obtained.

[Equation 16] When the condition of is satisfied, it is judged as the target object.
Alternatively, when a plurality of features are divided and exist in the area designation frame, if the intervals are smaller than a certain value, they are integrated and regarded as the same object.

The movement of the object using the time continuity from the coordinates of the target object specified in the initial state will be examined as shown in FIG. The coordinates of the object in the initial state t0 are (s0, s1). Next, the coordinates (s2, s3) continuous to the coordinates (s0, s1) at time t1 are obtained. Here, the size of the object is

[Equation 17] And the difference in size dw at this time is

[Equation 18] , Which determines the size of the object.

Further, in this embodiment, as the detection of the state of the target object, the intersection with other objects and the deletion from the screen are detected.
Actually, the intersection of other objects becomes a spatiotemporal image as shown in FIG. 19A, and it can be seen that the width of the object is hidden by the other objects. The crossing state is estimated by using this change in the object width. When the tracking object is specified by color, the color trajectory of the tracking object is a trajectory divided in the middle as shown in FIG.

Further, when the object disappears from the screen, the motion trajectory of the spatiotemporal image is interrupted, and hence the same algorithm as before cannot be used for tracking. Therefore, it is necessary to determine whether or not it disappeared from the screen. FIG. 20 shows how the spatiotemporal image disappears from the screen. If the coordinates of the object are (s0, s1) and the screen width is width, the coordinates of the object are

[Formula 19] Or

[Equation 20] In the case of, it is determined that it is about to disappear from the screen.

In the system control circuit 13 of this embodiment, the motion vector is detected from the color spatiotemporal image. Actually, as shown in FIG. 21, first, there is an object specified in the color spatiotemporal image. The coordinates at both ends at time t are (s0, s1). Both ends coordinates (s0
, S1) and the coordinates of both ends of the feature continuous to (s2, s3)
And Using this continuity, the motion vectors obtained from both ends coordinates and the center coordinates are mv1, mv2, and mv3. Further, as shown in FIG. 22, in the color spatiotemporal difference image, the motion vector is obtained from the temporal continuity of all the edges within the area designation frame. At this time, motion vectors mvi to mvj are set for all edges in the frame.

As for the luminance of the target object, as shown in FIG. 23, a luminance histogram in the area 21 in which the target object is designated is taken, and the portion with frequency deviation is defined as the luminance range (h0, h1) of the target object. decide. Then, in the same way as the color signal, the average value is obtained for this luminance range (h0, h1),
In the x and y directions, respectively,

[Equation 21] To generate a luminance spatiotemporal image. The calculation range is the band-shaped area described above with reference to FIG.

Subsequently, as shown in FIG. 24, an edge portion is detected for a feature in the area designation frame of the luminance spatiotemporal image, and pattern matching of a small area including the edge is performed to obtain an area having the largest correlation. Is regarded as the corresponding area,
Calculate the motion vector. At this time, the motion vectors mvk to mvl are calculated for all edges in the area designating frame.

In this way, the motion vectors mv1, mv2, mv3, mvi to mvj, mvk obtained from the color spatiotemporal image, the color spatiotemporal difference image, and the luminance spatiotemporal image, respectively.
.About.mvl are comprehensively determined, and from the motion vector histogram obtained from these motion vectors as shown in FIG. 25, the motion amount having the maximum frequency is determined as the motion vector at that time.

According to the above construction, after the Y / C separation of the video signal taken in from the photographing optical system, the target object is specified from the histogram of the tracking area, and the color spatiotemporal image, the color spatiotemporal difference image and the luminance thereof are identified. The motion vector is determined from each spatiotemporal image, and the majority of these motion vectors are used to determine the motion vector, so that even a patterned object can be reliably detected with a smaller amount of calculation compared to conventional block matching. A possible object tracking device can be realized.

Further, according to the above-mentioned configuration, the amount of information is smaller than that of the conventional block matching, so that the amount of calculation can be reduced, and thereby the processing can be speeded up and the memory capacity can be reduced. In addition, since the compressed features are used, the amount of motion can be calculated without being affected by the shape or unique pattern of the object, and tracking is performed after the target object is specified. Even if it disappears from, the false tracking can be effectively prevented.

Further, according to the above-mentioned structure, the drive signal for controlling the camera lens, the camera body and / or the zoom lens is generated based on the movement amount and the deformation amount in the x and y directions. In addition to tracking the subject, it is possible to automatically respond to changes in the size of the subject, and it is possible to realize an object tracking device that can always shoot the target object with a preferable size, and it is easy to use Can be significantly improved.

Furthermore, according to the above configuration, x and y
Although the photographing optical system 2 is controlled based on the amount of movement and the amount of deformation in the direction, it is also possible to move, enlarge, and reduce the area designation frame, as in the first embodiment. in this case,
The coordinates of the new area designating frame are obtained from the motion vector in the x and y directions and the width of the color spatiotemporal image.

(3) Third Embodiment In FIG. 26, in which parts corresponding to those in FIGS. 1 and 10 are designated by the same reference numerals, reference numeral 30 generally indicates an object tracking device according to a third embodiment of the present invention. , Y stored in the memory 10
The signal, the H signal, and the S signal are input to the spatiotemporal image processing circuits 22, 23, and 24, respectively. And this Y, H, S
A color and luminance signal histogram within a designated area frame designated for tracking the signal is obtained, and a spatiotemporal image in the x and y directions is obtained for the upper n-th color having a large output of the color filter 21. The spatiotemporal image obtained here is discriminated by the system control circuit 13 to determine whether it is the color of the target object.

Then, the coordinates of the target object are determined using the spatiotemporal image, and the segmentation is performed in the two-dimensional space. After that, the motion vector in the x and y directions, the state discrimination, and the size are calculated by using the temporal continuity of the color space-time image of the determined target object, and the panning motor drive signal PAN and the zoom for tracking the target object are calculated. The lens motor drive signal ZOOM and the tilt motor drive signal TILT are output. These motor drive signals PAN, ZOOM, TIL
At T, the panning motor 14, the zoom lens motor 15, and the tilting motor 16 operate, respectively, so that the target object can be automatically tracked.

In the case of this embodiment, the system control circuit 13 determines the state of the target object such as disappearance from the screen or temporary disappearance due to intersection with another object, size determination, and segmentation in a two-dimensional space. And so on. When it is determined that the target object has disappeared, the pattern matching circuits 31, 32, and 33 perform matching using the two-dimensional pattern of the object before disappearing using the Y signal, the H signal, and the S signal. It is designed so that it can be judged whether or not it appears again on the screen, and if it appears again, the tracking can be restarted.

Here, in the object tracking processing according to this embodiment, when the target object is first separated from the background or when the target object is specified, the user wants to track the object, as in the first and second embodiments. It is assumed that the system is started by designating an object in the area designating frame on the initial screen. As shown in FIG. 15, the input image size is (M,
N), the coordinates of the upper left corner of the window of the area designation frame WAKU are (x0, y0), and the window size of the area designation frame WAKU is (wx, wy).

FIG. 15A shows the target object as the area designation frame WA.
This is the initial state of the system designated by KU, and the target object in this example is a red car. Here, in the image at time t = 0 (initial screen), the Hue histogram in the area designation frame WAKU is multiplied by the color filter of FIG. 14 described above, and the area of the histogram in each color range is used as the filter output,
A spatio-temporal image is created for the upper n output color regions (for example, n = 2).

The spatiotemporal image generated here is represented by H (x, t
0). In this spatiotemporal image H (x, t0),
The point where H (x, t0) = 0 to H (x, t0)> 0 changes is defined as xsi, and H (x, t0)> 0 to H (x, t0).
The point at which) = 0 changes to xei. Where (xsi,
The range represented by xei) is that there is a feature, that is, there is an object in the color gamut range (h1, h2) at this position.

[Equation 22] I and i + 1 if there is a relation such as
Are integrated (FIG. 27 (A)).

At the coordinates thus obtained, as shown in FIG.
As shown in FIG. 7 (B), the coordinates of both ends of the window (x0,
The coordinates (xs0, xe0) concerning x0 + wx-1) are determined as the target object. Similarly, the y-direction window coordinate (y0
, Y0 + wy-1) to the coordinates (ys0, y) of the target object.
e0) is determined. Then the following equation,

[Equation 23] If the relationship is established, the color is judged to be the background color because the color exists on the entire screen. If not, the color of the target object is determined, and tracking is started using this color spatiotemporal image. Here, the coordinates (xs0, xe0), (ys
0, ye0) is stored as the initial coordinates of the target object.

Next, the color signal is set to H (x, y, t), and the spatiotemporal image is obtained from the color region range (h1, h2) determined as the color of the target object. At this time, the range of the spatiotemporal image calculation area is, as shown in FIG.
Is defined as a rectangular area in the x and y directions. If the region coordinates in the x direction are (axs, axe) and the coordinates in the y direction are (ays, aye), the color signal spatiotemporal image generation formula is

[Equation 24] Ask in.

When actually tracking the target object, t> 0
The color spatiotemporal image determined as the object color is obtained. The spatiotemporal image generated here is H (x, t). In this spatiotemporal image H (x, t), H (x, t)
From x = 0 to H (x, t)> 0, xsi, H
The point at which (x, t)> 0 changes to H (x, t) = 0 is x
ei. Here, the range represented by (xsi, xei) has features, that is, the color region range (h
It means that there is an object of 1, h2). In addition, as in the above equation (22), the following equation is added to xsi and xei.

[Equation 25] If there is a relation of, i and i + 1 are integrated.

Further, as shown in FIG. 28, the coordinates (xs0, xe0), (ys0, ye0) of the target object obtained at t = 0 are obtained.
) Continuous feature coordinates (xs, xe), (ys, y
e) is required. The coordinates of consecutive features at this time are

[Equation 26] It becomes a continuous relational expression like. The coordinates obtained here are the time t
It is determined to be the target object in. As a result, the zoom lens drive signal Z for controlling the zoom lens motor 15
OOM is the following formula

[Equation 27] Generated by.

Subsequently, (xs, xe) is newly added to (xs0,
xe0), the object tracking, the size determination, and the region identification are sequentially performed again in the same procedure. The Y, H, and S signals in the target object regions (xs0, xe0) and (ys0, ye0) obtained here are used to determine whether or not the object appears again when the object disappears, as shown in FIG. The reference signal is stored in the system control circuit 13 as a reference signal used for pattern matching.

Further, the reference signal at the time t-1 and the difference signal of the target object area at the time t are also stored as reference values to be referred to during the matching determination. Time t
The reference area coordinates at -1 are (xs0 ', xe
0 '), (ys0', ye0 '), the reference error signals ΔY, ΔH, ΔS of the Y, H, S signals are

[Equation 28] become that way. These signals are updated every hour unless a state determination such as intersection or disappearance is made.

Next, in this system control circuit 13, the time difference of the color spatiotemporal image between time t and t-1 is calculated by the following equation.

[Equation 29] Ask more. In this way, the static part in the color space-time image is deleted and the information of the part where the motion is generated is extracted.

Next, the system control circuit 13 subsequently detects the state. The state to be detected here is temporary disappearance due to intersection with another object (FIG. 19) or disappearance from the screen (FIG. 20). At the intersection in the spatiotemporal image, it can be seen that when the object intersects with another object, the width of the object is hidden by the other object and the width becomes smaller. By using this phenomenon, the object width at time t can be calculated as wx (t), wy (t)
And the object width at time t-1 is wx (t-1), wy
If (t-1),

[Equation 30] It is assumed that there is a possibility that an intersection has occurred when the relationship of. After that, if the time continuity is interrupted, it is determined that an intersection has occurred.

Further, as shown in FIG. 20, when the object disappears from the screen, the motion trajectory of the spatiotemporal image is interrupted, and thus the same algorithm as before cannot be used for tracking. Therefore, it is necessary to judge whether or not it disappeared from the screen. The coordinates of the object are (xs0, xe0) and the screen width is M
Then,

[Equation 31] When the relationship of exists, it is determined that it is about to disappear from the screen. Then, when the time continuity disappears, it is determined that the display disappears from the screen.

Next, the system control circuit 13 detects the motion vector from the color space-time image. That is, as shown in FIG. 21, the both-end coordinates of the object identified in the color space-time image at time t-1 are (xs0, xe0), and the both-end coordinates of the features continuous at (xs0, xe0) at time t are (xs0, xe0). , Xe). Using this continuity, the motion vectors obtained from the coordinates at both ends and the coordinates at the center are mv1 and mv1.
2 and mv3.

Further, as shown in FIG. 22, in the color spatiotemporal difference image, the motion vector is obtained from the temporal continuity of the motion edges in the area designating frame. At this time, the motion vector mvi is calculated for all edges in the frame. The motion vectors mv1, mv2, mv3 thus obtained
, Mvi are comprehensively determined to determine the final motion vector. That is, these motion vectors mv1, mv
From the histogram obtained from 2, mv3 and mvi, the motion amount having the maximum frequency is determined as the motion vector mv at that time.

From the motion vectors mvx and mvy in the x and y directions obtained here, the actual image size in the variable image compression circuit 8 and the size of the image to be processed are considered,
The panning motor drive signal PAN and the tilt motor drive signal TILT for controlling the panning motor 14 and the tilting motor 16 are expressed by the following equations.

[Equation 32] Based on.

In the system control circuit 13 of this embodiment, the reappearing object is tracked by using pattern matching. First, when it is judged by the state judgment that an intersection with another object has occurred. , The pattern matching is performed using the reference pattern of the target object stored in the memory 10 by the Y, H and S signals to check whether or not the object appears again. It is assumed that the size of an object when it disappears temporarily and then reappears does not change much from the size when it disappears. At this time, as shown in FIG. 30, the areas where the matching is performed are ± (mvx + α) and ± (mvy + α) centered on the disappeared position.

When it is determined by the state determination that the object disappears from the screen, Y of the target object stored in the memory 10,
Pattern matching is performed using the reference pattern by the H and S signals to check whether or not the object appears again. It is assumed that the size of the object when it reappears does not differ much from the size when it disappears. At this time, if it is assumed that the areas where the matching is performed are continuously photographed, it is assumed that they appear from any of the four sides of the image as shown in FIG. The error signal Yerror of the pattern matching obtained at this time,
As Herror and Serror,

[Expression 33] When, the target object is judged to have reappeared. Here, the variables α, β, and γ are values that are determined when the reference pattern is held and when the luminance level and color signal level of the current image are changed.

The system control circuit 13 of the object tracking device 30 can realize the above-described processing by executing the object tracking processing procedure SP0 shown in FIGS. 32, 33 and 34.

According to the above construction, after the Y / C separation of the video signal taken in from the photographing optical system, the target object is specified from the histogram of the tracking area, and the time of the color spatiotemporal image and the color spatiotemporal difference image are identified. The motion vector is calculated using continuity, and the motion vector is determined by taking the majority vote of these, so that even a patterned object can be reliably moved with a smaller amount of calculation compared to conventional block matching. It is possible to realize an object tracking device that can detect the.

Further, according to the above configuration, the motion tracking is performed using the time continuous information as compared with the conventional block matching, thereby preventing the detected motion amount from becoming discontinuous. You can do natural tracking. Also, the amount of information is smaller than before, so the amount of calculation can be reduced,
This makes it possible to speed up the process and reduce the memory capacity. Furthermore, since the compressed feature quantity is used, the amount of movement can be calculated without being affected by the shape of the object or the unique pattern, and tracking is performed after the target object is specified. Even if it disappears from, the false tracking can be effectively prevented.

Further, according to the above configuration, the target object is specified and then tracked as compared with the conventional method.
It is possible to prevent erroneous tracking when it disappears from the field of view, and it is possible to obtain reference data for pattern matching because the segmentation is performed sequentially while specifying the target object area. It is possible to realize robust tracking in various scenes by using both the time-axis direction tracking based on time continuity and the pattern matching method in the spatial direction.

Further, according to the above configuration, the drive signal for controlling the camera lens, the camera body and / or the zoom lens is generated based on the movement amount and the deformation amount in the x and y directions. In addition to tracking the subject, it is possible to automatically respond to changes in the size of the subject, and it is possible to realize an object tracking device that can always shoot the target object with a preferable size, and it is easy to use Can be significantly improved.

Furthermore, according to the above configuration, x and y
Although the photographing optical system 2 is controlled based on the amount of movement and the amount of deformation in the direction, it is also possible to move, enlarge, and reduce the area designation frame, as in the first embodiment. in this case,
The coordinates of the new area designation frame are obtained from the motion vectors mvx and mvy in the x and y directions and the width of the color spatiotemporal image.

(4) Other Embodiments In the above-described embodiments, the movement of the target object in the image is configured by using the luminance signal and the color difference signal which are configured in the television camera as the object tracking device and obtained from the digital camera processing circuit. I described the case of detecting and tracking
Instead, it is configured separately from the television camera and detects the movement of the target object in the image using the luminance signal and color difference signal obtained from the output signal of the television camera, thereby controlling the television camera to perform tracking processing. Even if the above is realized, the same effect as that of the above-described embodiment can be realized.

In the above-mentioned embodiments, the case where the object tracking device according to the present invention is applied to the one for tracking the target object by moving the photographing optical system of the television camera in accordance with the movement of the target object has been described. However, the present invention is not limited to this, and may be applied to, for example, moving a moving object on which a television camera is mounted in accordance with the movement of the target object to track the target object.

[0084]

As described above, according to the present invention, when the tracking target object is placed in the designated area set on the screen and the tracking is started, the histogram in the designated area of the video signal is calculated and the frequency of the histogram is calculated. The spatiotemporal image is obtained for a high range of, and the movement of the photographing means or the designated area is controlled by the trajectory of the spatiotemporal image obtained as a result, so that even various objects can be compared with the conventional block matching. An object tracking device capable of surely detecting and tracking a motion can be realized.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of an object tracking device according to the present invention.

FIG. 2 is a schematic diagram used for explaining a video signal color plan view.

FIG. 3 is a schematic diagram provided for explaining the designation of a target object by a region designation frame on a screen.

FIG. 4 is a schematic diagram for explaining the color and saturation of a target object.

FIG. 5 is a schematic diagram for explaining generation of spatiotemporal images in the x direction and the y direction.

FIG. 6 is a schematic diagram for explaining determination of a color range.

FIG. 7 is a schematic diagram illustrating an example of a generated spatiotemporal image.

FIG. 8 is a schematic diagram illustrating area division of a screen when generating a spatiotemporal image.

FIG. 9 is a schematic diagram for explaining movement and deformation amount in a spatiotemporal image.

FIG. 10 is a block diagram showing a second embodiment of the object tracking device according to the present invention.

FIG. 11 is a block diagram showing a configuration of a saturation / hue conversion circuit.

FIG. 12 is a characteristic curve diagram for explaining the relationship between the distribution of color difference signals and the quantization step.

FIG. 13 is a schematic diagram illustrating a configuration of a matrix that converts a color difference signal into a saturation / hue.

FIG. 14 is a characteristic curve diagram for explaining a color filter.

FIG. 15 is a schematic diagram used for designating a target object and explaining a spatiotemporal image calculation area.

FIG. 16 is a schematic diagram for explaining a color spatiotemporal image.

FIG. 17 is a schematic diagram used for identifying a target object.

FIG. 18 is a schematic diagram for explaining the detection of the size of a target object.

FIG. 19 is a schematic diagram for explaining the intersection with another object in the spatiotemporal image.

FIG. 20 is a schematic diagram for explaining disappearance of a spatiotemporal image from the screen.

FIG. 21 is a schematic diagram for explaining motion vector detection based on a color spatiotemporal image.

FIG. 22 is a schematic diagram for explaining the motion vector detection using the color space-time difference image.

FIG. 23 is a characteristic curve diagram for explaining a luminance histogram.

FIG. 24 is a schematic diagram for explaining motion vector detection using a luminance histogram.

FIG. 25 is a histogram showing a motion vector histogram.

FIG. 26 is a block diagram showing a third embodiment of the object tracking device according to the present invention.

FIG. 27 is a schematic diagram used for specific description of a target object.

FIG. 28 is a schematic diagram illustrating a method of determining a target object coordinate based on time continuity.

FIG. 29 is a schematic diagram used for explaining a reference area serving as a reference for performing pattern matching when an object disappears.

FIG. 30 is a schematic diagram showing a matching area in which pattern matching is performed when an object disappears by crossing.

FIG. 31 is a schematic diagram showing a matching area in which pattern matching is performed when an object disappears from the screen.

FIG. 32 is a flow chart for explaining an object tracking processing procedure.

FIG. 33 is a flowchart provided to explain the object tracking processing procedure subsequent to FIG. 32.

34 is a flowchart provided for explaining the object tracking processing procedure subsequent to FIGS. 32 and 33. FIG.

[Explanation of symbols]

1, 20, 30 ... Object tracking device, 2 ... Imaging optical system,
3 ... AGC circuit, 4 ... analog-digital conversion circuit,
5 ... Digital camera processing circuit, 6 ... Digital-analog conversion circuit, 7 ... Low-pass filter, 8 ... Variable image compression circuit, 9 ... Saturation / hue detection circuit, 10 ... Memory, 11, 22, 23 , 24 ... Spatiotemporal image processing circuit,
12 ... Histogram processing circuit, 13 ... System control circuit, 14 ... Panning motor, 15 ... Zoom lens motor, 16 ... Tilting motor, 21 ... Color filter, 31, 32, 33 ... Pattern matching Processing circuit.

Claims (18)

(57) [Claims] 1. An object tracking device for tracking a target object in a designated area set on a screen, a photographing means for photographing an object to generate a video signal, and the image in the designated area on an initial screen. Histogram processing means for calculating a histogram of a signal, spatio-temporal image processing means for obtaining a spatio-temporal image in a high-frequency range of the histogram, and control means for controlling the photographing means based on the trajectory of the spatio-temporal image An object tracking device characterized by comprising: 2. The control means detects a first movement amount based on the trajectory of the spatiotemporal image, and controls the photographing direction of the photographing means based on the first movement amount. The object tracking device according to claim 1. 3. The object tracking device according to claim 1, wherein the control means controls zooming of the photographing means based on a change in the width of the trajectory of the spatiotemporal image. 4. A Hue detecting means for detecting a Hue of the video signal, wherein the histogram processing means comprises the Hue.
The object tracking device according to claim 1, wherein the histogram of is calculated. 5. The object tracking device according to claim 1, wherein the control means determines whether the target object intersects or disappears from the screen based on the trajectory of the spatiotemporal image. 6. A Hue detecting means for detecting a Hue of the video signal, and a color filter means for filtering the Hue, wherein the spatiotemporal image processing means is a high order n output of the color filter means. The object tracking device according to claim 1, wherein the spatiotemporal image is calculated for each color of rank. 7. A Hue detecting means for detecting a Hue of the video signal, wherein the histogram processing means comprises the Hue.
And the luminance histogram, the spatiotemporal image processing means generates a luminance spatiotemporal image and a color spatiotemporal image for the high frequency ranges of the Hue histogram and the luminance histogram, respectively, and the control means. Detects the second and third movement amounts based on the loci of the luminance space-time image and the color space-time image,
The object tracking device according to claim 2, wherein the first motion amount is determined based on the first motion amount and the third motion amount. 8. A Hue detecting means for detecting a Hue of the video signal, wherein the histogram processing means comprises the Hue.
Of the Hue histogram, the spatiotemporal image processing means generates a color spatiotemporal image and a color spatiotemporal difference image, and the control means controls the color spatiotemporal image and the color. The second and third motion amounts are detected based on the trajectory of the spatiotemporal difference image, and the first motion amount is determined based on the second and third motion amounts. The object tracking device according to 1. 9. The control means performs pattern matching between the pattern of the target object area and the current screen when the target object intersects or disappears from the screen, and based on the error in the pattern matching. The object tracking device according to claim 5, wherein the reappearance of the target object is determined. 10. An object tracking device for tracking a target object in a designated area set on a screen, histogram processing means for calculating a histogram of a video signal in the designated area on an initial screen, and frequency of the histogram. The object tracking device, comprising: a spatio-temporal image processing means for obtaining a spatio-temporal image in a high range and a control means for controlling the designated area based on the trajectory of the spatio-temporal image. 11. The control means detects a first motion amount based on the trajectory of the spatiotemporal image, and moves the designated area based on the first motion amount. Item 10. The object tracking device according to item 10. 12. The object tracking device according to claim 10, wherein the control means controls the size of the designated area based on a change in the width of the trajectory of the spatiotemporal image. 13. A Hue for detecting a Hue of the video signal.
The histogram processing means includes a detecting means.
2. A histogram of e is calculated, wherein
0. The object tracking device according to 0. 14. The object tracking device according to claim 10, wherein the control means determines whether the target object intersects or disappears from the screen based on the trajectory of the spatiotemporal image. 15. A Hue for detecting a Hue of the video signal.
The spatio-temporal image processing means computes the spatio-temporal image for the upper n colors of the output of the color filter means, the detection means and the color filter means for filtering the Hue. The object tracking device according to claim 10. 16. A Hue for detecting a Hue of the video signal.
The histogram processing means includes a detecting means.
The histogram of e and the histogram of brightness are calculated, and the spatiotemporal image processing means generates a brightness spatiotemporal image and a color spatiotemporal image with respect to the high frequency ranges of the Hue histogram and the brightness histogram, respectively. The means detects second and third movement amounts based on the loci of the luminance spatiotemporal image and the color spatiotemporal image, and determines the first movement amount based on the second and third movement amounts. The object tracking device according to claim 11, wherein: 17. A Hue for detecting a Hue of the video signal.
The histogram processing means includes a detecting means.
The histogram of e is calculated, and the spatiotemporal image processing means is
A color spatiotemporal image and a color spatiotemporal difference image are generated, and the control means detects the second and third motion amounts based on the loci of the color spatiotemporal image and the color spatiotemporal difference image, and the second
The object tracking device according to claim 11, wherein the first motion amount is determined based on the third motion amount and the third motion amount. 18. The control means performs pattern matching between the pattern of the target object region and the current screen when the target object intersects or disappears from the screen, and based on the error of the pattern matching. The object tracking device according to claim 14, wherein the reappearance of the target object is determined.