JP3254241B2 - Moving object identification and tracking system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving object identification and tracking system for identifying and tracking a moving object.

[0002]

2. Description of the Related Art Conventionally, pattern matching using a camera has been performed by using a pattern matching method, and tracking of a moving object has been performed by using signals from an electromagnetic wave, light, ultrasonic sensor or the like. For example, a CCD camera is fixed and the operation of a certain place is stored in a video, and a method of identifying the operation by a human eye is used, and is used for crime prevention and the like.

[0003]

However, the tracking device using the conventional electromagnetic wave, light, ultrasonic wave sensor and the like tracks an object, but cannot identify an object from various types of objects. If the state of an object changes along the way, tracking cannot be performed, and there is a problem in human rights because surveillance by a person is required. SUMMARY OF THE INVENTION The present invention has been made in view of such circumstances, and enables a moving object to be easily identified and to be able to track freely and with high accuracy even if the state of the moving object changes. The purpose is to provide a system.

[0004]

According to the present invention, there is provided a semiconductor device comprising:
The moving object identification and tracking system according to the present invention learns reference image data of an object having a different shape to be identified in advance by a neural network, captures an image of a moving object with a television camera, and, from the image data captured by the television camera,
The object to be tracked is identified through the neural network, and the controller controls the deviation between the image data and the reference image and its change, and further, the fuzzy control of the deviation between the TV camera rotation angle data and the reference rotation angle data and its change. It is configured to calculate using rules and drive a moving object tracking mechanism via a motor drive circuit so that the TV camera tracks the moving object.

[0005]

In the moving object identification and tracking system according to the first aspect, the neural network learns in advance the reference image data of the object having a different shape to be identified, images the moving object with a television camera, and captures the moving object with the television camera. From the image data, an object to be tracked through the neural network is identified. Further, the controller calculates the deviation between the image data and the reference image and the change thereof, and further calculates the deviation between the rotation angle data of the TV camera and the reference rotation angle data and the change thereof, and calculates the reference image data and the reference rotation angle reference data. The moving object tracking mechanism is driven via a motor drive circuit so that the deviation of the moving object falls within a predetermined range, and the moving object is tracked by the television camera.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments embodying the present invention will now be described with reference to the accompanying drawings to provide an understanding of the present invention. Here, FIG. 1 is a schematic block diagram of a software surface of the moving object identification and tracking system according to one embodiment of the present invention, FIG. 2 is a schematic block diagram of the hard surface, and FIG. FIG. 4 is a sectional view of a moving object tracking mechanism used in the object identification and tracking system, FIG. 4 is an explanatory view of image data processing, and FIG. 5 is an explanatory view showing input and output of a fuzzy controller. FIG.
As shown in FIG. 1, a moving object identification and tracking system 10 used in an embodiment of the present invention is a CCD which is an example of a television camera.
A camera (charge-coupled device camera) 11, a microcomputer system 12 for connecting the CCD camera 11,
It has a motor drive circuit 13 connected to the microcomputer system 12, and a moving object tracking mechanism 14 connected to the motor drive circuit 13.

[0007] The microcomputer system 12
Is the video I / F board 1 connected to the CPU bus 15
6. A fuzzy controller 18 (see FIG. 1) which is an example of a controller provided with a parallel I / O board 17, and the present inventor described in Japanese Patent Application No. 3-155812, "Method of identifying shape of natural product or processed product". Has a neural network 19 (see FIG. 1) proposed in FIG. The CCD camera 11 is connected to a video I / F board 15. The motor drive circuit 13 is connected to the parallel I / O board 17. The moving object tracking mechanism 14
Is provided with a DC motor 20 for horizontal movement and a DC motor 21 for vertical movement, which are examples of motors connected to the motor drive circuit 13, and a horizontal movement interlocked with the DC motors 20 and 21 for horizontal movement and vertical movement. And a vertical movement encoder 23. The horizontal movement and vertical movement encoders 22 and 23 are connected to the parallel I / O board 17. The CCD camera 1
1, the microcomputer system 12, and the motor drive circuit 13 are connected to a power supply (not shown).

Here, the moving object tracking mechanism 14 will be described in detail with reference to FIG.
It has an outer sphere 25 attached to the gantry 24, an inner sphere 26 rotatably mounted inside the outer sphere 25, and the CCD camera 11 fixed to the inner sphere 26. The gantry 24 fixes a horizontal movement DC motor 20 and a horizontal movement encoder 22 on both inner sides. At the ends of the DC motor 20 for horizontal movement and the encoder 22 for horizontal movement, a horizontal movement gear 2 is provided.
7 and 28 are fixed, and the horizontal movement gears 27 and 28 are fixed.
Are engaged with both ends of a horizontal gear 29, and the horizontal gear 29 fixes a horizontal shaft 30 rotatably fixed to the gantry 24 at the center thereof. Outer sphere 2
5 has a space inside, a window for an inner sphere (not shown) on the outer surface, the end of the horizontal shaft 30 is fixed at the lower part, and the vertical shafts 31 and 32 are fixed inside both sides. The vertical gear 33 is fixed to the vertical shaft 31. The inner sphere 26 has the CCD camera 11 fixed inside so that the outside can be imaged, the vertical movement encoder 23 fixed to the upper part of the CCD camera 11, and the vertical movement DC motor 21 fixed to the lower part. The ends of the vertical movement encoder 23 and the vertical movement DC motor 21 are connected to a vertical movement gear 34,
35 are rotatably fixed. Further, the vertical movement gears 34, 35 mesh with both ends of the vertical gear 33, respectively. The moving object tracking mechanism 14 is connected to a motor drive circuit 13 (see FIG. 2), and the motor drive circuit 13 is connected to the DC motor 20 for horizontal movement and the DC motor 21 for vertical movement.

Hereinafter, the moving object identification and tracking system 10 will be described.
A method for identifying and tracking a moving object using the method will be described.
As shown in FIG. 1, the moving object identification and tracking system 10
Is obtained by prefetching binary image data (12 × 15 pixels) of an object having a different shape to be discriminated into the neural network 19 and using a back error propagation method (back propagation method) as a reference image data as an example of the reference image data. 36 (see FIG. 4). When new image data 37 (see FIG. 4) is obtained by the CCD camera 11, the neural network 19 identifies the pattern to which the object belongs. And a fuzzy controller 1 which is an example of a controller
8, the horizontal and vertical patterns of the image data 37 captured by the CCD camera 11, or the patterns themselves, and the horizontal and vertical encoders 22 and 23.
The rotation angle data from FIG. 2 and FIG. 3 is compared with a reference rotation angle 39 which is an example of the reference rotation angle data, and a signal is sent to the motor drive circuit 13 so that the deviation is minimum or within a certain range. And the motor drive circuit 13
PWM (pulse width modulatio)
n) driving the DC motor 20 for horizontal movement via a circuit so that the deviation between the image data 37 of the moving object and the reference image data 36 is minimized or within a certain range;
Track moving objects. When the moving object moves in the vertical direction, the DC motor 21 for vertical movement is driven.

As shown in FIG. 3, the moving object tracking mechanism 14 is driven by a DC motor 20 for horizontal movement to rotate a gear 29 for horizontal direction via a gear 27 for horizontal movement. Then, the outer sphere 25 is rotated in a predetermined direction. In the vertical direction, the DC motor 21 for vertical movement is driven, and the rotation of the gear 35 for vertical movement causes the CCD camera 11 to rotate in the predetermined vertical direction together with the inner sphere 26 via the gear 33 for vertical direction, thereby obtaining the reference image. The deviation between the data 36 and the image data 37 of the moving object is made to fall within a certain range. Further, the movement of the CCD camera 11 is sent to the fuzzy controller 18 via the parallel I / O board 17 shown in FIG. 1 as a pulse signal by the horizontal movement encoder 22 in the horizontal direction and by the vertical movement encoder 23 in the vertical direction. The positional relationship of the CCD camera 11 is constantly grasped. Therefore, the moving object can be tracked by repeating this operation at each sampling interval of the image data 37.

Here, the function of the fuzzy controller 18 will be briefly described with reference to FIGS. 4 and 5, as to an example of image data, corresponding to a case where image data 37 of a rectangular object moves. When the rectangular image data 37 moves horizontally with respect to the reference image data 36 as shown in FIG. 4, a deviation e of the image before and after the movement and a change Δe thereof are calculated. These are added as inputs to the fuzzy controller 18 shown in FIG.
0 and a pulse for driving the vertical movement DC motor 21 (in the case of vertical movement) are obtained. That is, e and Δe are regarded as fuzzy quantities having a triangular membership function, and using a plurality (six in this embodiment) of fuzzy control rules, the fitness and the center of gravity are obtained by the minimax method, and the output fuzzy quantity is obtained. The value of the pulse of the movement amount in a certain horizontal direction and vertical direction is calculated and added to the PWM circuit of the motor drive circuit 13. Thus, the moving object of the rectangular image data 37 can be accurately tracked.

The function of the fuzzy controller 18 will be described in more detail. An arbitrary point of the input image having (12 × 15) pixels is defined as (i, j). The reference image data 36 of the rectangular object is f (i, j), and the image data 37 of the moving image after the T sampling time is f (i + Δi, j). The image of the hatched portion, that is, the deviation e (i, j) and its time change Δe (i, j) are calculated using the following equations (1) and (2). e (i, j) = f (i + Δi, j) −f (i, j) (1) Δe (i, j) = (e (i + Δi, j) −e (i, j)) / T .. (2) DC motor 2 for horizontal movement by fuzzy controller 18 using the two inputs of e and Δe and the rotation angle input
0, and outputs a pulse width for moving the vertical movement DC motor 21 (including a pulse width for moving the vertical movement DC motor 21 when vertical movement is also performed) and applies the pulse width to the PWM circuit of the motor drive circuit 13 to output the target moving object. Can be tracked freely.

The fuzzy controller 18 has e and Δe
In addition, the design is made so that the rotation angle has a predetermined specification, for example, a specification that takes a value within a certain range as an example. Further, six control rules are used as fuzzy control rules. The tracking using the fuzzy controller 18 has been described for the case of horizontal movement, but the principle is the same for a general moving object tracking method. As described above, the moving object identification and tracking system 10 according to one embodiment of the present invention
In the prior art, a neural network 19 can be used to identify an object that has been conventionally identified by the human eye, and furthermore, a moving object tracking mechanism 14 by a fuzzy controller 18 is provided. The device can be tracked, and its field of application is wide as a new identification tracking device that has never been seen before, and its ripple effect is large.

[0014]

In the moving object identification and tracking system according to the first aspect, the neural network learns in advance the reference image data of the object having a different shape to be identified, captures the moving object with a television camera, and uses the television camera to capture the moving object. From the captured image data, the object to be tracked is identified through the neural network.On the other hand, the deviation between the image data and the reference image and the change thereof, the deviation between the rotation angle data of the TV camera and the reference rotation angle data, and the like are determined by the controller. Calculate the change, drive the moving object tracking mechanism via the motor drive circuit, and track the moving object with the TV camera, so it can easily identify the moving object,
Even if the state of the moving object changes, tracking can be performed freely and with high accuracy.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram of a moving object identification and tracking system according to an embodiment of the present invention.

FIG. 2 is a schematic block diagram of the hard surface.

FIG. 3 is a sectional view of a moving object tracking mechanism used in the moving object identification and tracking system according to one embodiment of the present invention.

FIG. 4 is an explanatory diagram of image data processing.

FIG. 5 is an explanatory diagram showing input and output of a fuzzy controller.

[Explanation of symbols]

10: Moving object identification and tracking system, 11: CCD camera, 12: Microcomputer system, 13: Motor drive circuit, 14: Moving object tracking mechanism, 15: CPU bus, 16: Video I / F board, 17: Parallel I / O
Board, 18: Fuzzy controller, 19: Neural network, 20: DC motor for horizontal movement, 21:
DC motor for vertical movement, 22: encoder for horizontal movement,
23: Encoder for vertical movement, 24: Stand, 25: Outer sphere, 26: Inner sphere, 27: Gear for horizontal movement, 28: Gear for horizontal movement, 29: Gear for horizontal direction, 30: Shaft for horizontal direction, 31 : Vertical axis, 32: Vertical axis, 33:
Gear for vertical direction, 34: gear for vertical movement, 35: gear for vertical movement, 36: reference image data, 37: image data,
38: identification processing, 39: reference rotation angle

Claims (1)

(57) [Claims] 1. A neural network learns in advance reference image data of an object having a different shape to be identified, captures a moving object with a television camera, and wants to track the image data captured by the television camera through the neural network. The object is identified, and the controller calculates the deviation between the image data and the reference image and its change, and further calculates the deviation between the TV camera rotation angle data and the reference rotation angle data and its change using fuzzy control rules. A moving object identification and tracking system, wherein a moving object tracking mechanism is driven via a motor drive circuit, and the moving object is tracked by the television camera.