JPH01303582A - Motion pattern analyzing device - Google Patents

Abstract

PURPOSE:To analyze the motion pattern of a body in motion easily and efficiently by indicating a position of the body in motion to be analyzed and supplying knowledge regarding the motion of the position. CONSTITUTION:An indication part 7 indicates the position to be analyzed in an initial image of the moving body displayed on a display part 6 and knowledge information on characteristic property regarding the motion of the position is indicated and inputted. For the purpose, image processing technique and the knowledge regarding the motion are used to make the object positions correspond to each other between continuous image pickup input images of the body in motion. Consequently, the motion of the object positions can be tracked and the motion pattern can be analyzed. Consequently, the motion patterns of various bodies in motion are easily and efficiently in a natural state.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a movement pattern analysis device that can easily analyze the movement pattern of a specific part of a moving body.

(Prior Art) Analysis of motion patterns of various moving bodies has conventionally been carried out solely by attaching marks to the moving bodies and tracking the movements of the marks. Specifically, for example, a small LE
A D light emitter is attached to the part of the moving body to be analyzed for the movement pattern, and the moving body is moved in a darkened environment. Then, this state is recorded and then the motion pattern is analyzed by playing back the recorded images one scene at a time and tracing the movement of the LED light emitter.

However, the work of reproducing recorded images scene by scene and tracking the movement of the LED light emitters requires a great deal of effort and processing time. Moreover, when attempting to simultaneously analyze a plurality of motion pattern analysis target parts, it is very difficult to associate the analysis target parts for each scene.

To make matters worse, it is often difficult to attach an LED light emitting device to a part of a moving body whose movement pattern is to be analyzed. For example, when analyzing the movement of a person or a machine, it is relatively easy to install an LED light emitting device, but when trying to analyze the movement of an arbitrary animal, it is extremely difficult to install it. Moreover, when the analysis target is an animal, if you attempt to record its movements in a darkened environment, there is a risk that the animal may adopt a movement pattern different from its original movement. For this reason, there have been problems such as it being extremely difficult to analyze the movement pattern of a moving body under natural conditions.

(Problem to be Solved by the Invention) In this way, in the past, for example, marks were placed on parts to be analyzed in the movement pattern of a moving body, and movements of the moving body were marked to facilitate analysis processing. There are various restrictions such as the need to darken the surrounding area when recording.
A major problem has been that it is difficult to easily analyze the movement pattern of any moving object in a natural environment.

The present invention was made in consideration of these circumstances, and its purpose is to enable various movements in a natural environment without the hassle of marking the analysis target part of a moving body. The purpose of the present invention is to provide a movement pattern analysis device that can easily and efficiently analyze body movement patterns.

[Configuration of the Invention (Means for Solving the Problems) A motion pattern analysis device according to the present invention includes an imaging means such as a television camera that continuously captures and inputs images of a moving body whose motion pattern is to be analyzed; a means for displaying at least an initial image of the image of the moving body, and a part (motion point) of the moving body to be analyzed with respect to the displayed image;
analysis of the moving body between images that are continuously captured and input. The present invention is characterized in that the movement pattern of the target parts is analyzed by sequentially associating the target parts according to the above-mentioned knowledge.

In other words, by using image processing technology and knowledge about motion, the movement of the analysis target part is tracked by correlating the analysis target part between consecutive images of a moving body. The present invention is characterized in that the motion pattern of the above-mentioned region to be analyzed is thereby analyzed.

(Function) According to the present invention, a part (motion point) to be analyzed for at least an initial image of images of a moving body that is continuously captured and input, and knowledge regarding the motion of that part (i.e., how it compares with other motion points). By using image processing technology and the knowledge regarding the above-mentioned motion, the parts to be analyzed are sequentially correlated between images obtained by continuously capturing and inputting images of a moving body. By associating the parts to be analyzed in the continuous images, it becomes possible to track the movement of the parts to be analyzed, and it becomes possible to analyze the movement pattern thereof.

Therefore, it is no longer necessary to mark the area to be analyzed in the movement pattern of a moving object or to darken the surrounding area when recording the movement of the moving object to facilitate analysis processing, as in the past. The motion patterns of various moving bodies can be analyzed easily and efficiently.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic system configuration diagram of a motion pattern analysis device according to an embodiment of the present invention, where ◯ indicates a TV left camera that continuously captures and inputs images of a moving body to be analyzed. The image (image data) of the moving object captured and inputted via this TV camera 1 is taken in to the control section (device main body section composed of a processor etc.) 3 via the A/D converter 2, and the image is The data are sequentially stored in the memory 4. This image memory 4 may be something like an optical disk file.

The image of the moving body inputted by the TV left camera and stored in the image memory 4 is outputted via the D/A converter 5 and displayed as an image on the display section G. This display section 6
The instruction unit 7
The part to be analyzed is specified from the ``controller'', and knowledge information such as specific properties regarding the movement of the part is inputted.

The instructions for the analysis target region are inputted from the instruction unit 7 by, for example, using a stylus pen to trace the boundaries of the moving body part in the displayed image and cutting out the moving body image as shown in FIG. 2(a). This is done by specifying the parts to be analyzed in the moving body image, such as the joints and their connections, as shown in FIG. 2(b). Then, information such as the unique motion properties known in advance of each of the above-mentioned analysis target parts is inputted as knowledge regarding motion.

The information on the analysis target part and the knowledge information regarding its motion include, for example, as shown in FIG. It is given as the permissible rotation angle and the direction of rotation. Such instruction information is stored in the table 8 and used for subsequent movement pattern analysis processing.

The object recognition unit 9 cuts out a moving body image from the input image according to a predetermined image processing technique, extracts its features, and recognizes the moving body. Specifically, as described above, a background image is obtained from the initial image in which the moving body image is specified, and based on this background image, the moving body images are respectively cut out from each successively inputted image. Then, feature extraction processing is performed on each of these images, and corresponding parts of the designated analysis target region are sequentially bundled as described above by collating the feature extraction parts and judgment processing based on the knowledge information. ing.

FIG. 4 shows a schematic flow of motion pattern analysis processing in the embodiment device, and this processing is started by inputting continuously captured images (moving images) of a moving body (
Step a). When an image of a moving body is input in this way, first, the initial image is displayed (step b
), information on the region to be analyzed (observation point) is input to this display image as described above (step C). This instruction information is stored in the table 8 in the format illustrated in FIG. 4, for example.

After the region to be analyzed is specified as described above, first, feature extraction processing is performed on the initial image (step d). This feature extraction process is performed, for example, by cutting out a moving object image by removing the background image portion from the initial image and thinning the moving object image. Then, instructions are inputted to this thinned image as described above, and the table 8
Characteristic information is obtained by specifying each of the analysis target parts (joints) based on the information stored in the .

After that, the feature extraction process is similarly performed on the image of the next scene (step e), and the feature points are compared with the previously obtained feature points in the image (step f). This matching of feature points is performed by examining the correspondence of feature points between images using the knowledge regarding motion described above.

For example, for the feature points defined as shown in Figure 3(a) in the initial image, from the next screen to Figure 3(b).
) When the feature points shown in ) are extracted, it is first checked whether the connection relationships between the rotation axes correspond, and then the rotation angles of each rotation axis are indicated using the knowledge information described above. It is checked whether the rotation constraint conditions are satisfied. When these conditions are met, the feature point is associated with the feature point previously determined (step g). Information on the feature points associated in this manner is sequentially stored in the analysis data memory IO as a series of coordinate values that change as the rotation axis moves, for example, as shown in FIG.

Incidentally, when the information on the feature points is obtained as shown in Figure 3(c), since the rotation angle of the rotation axis C around the rotation axis B does not satisfy the rotation constraint conditions described above, these It is determined that the feature points do not correspond to each other,
Excluded from the mapping.

The above feature point matching process is repeatedly performed on the feature point information obtained in the previous screen as long as there is a next screen for the image of the moving object (step h). When the correspondence between the feature points between the images of each scene is completed, the movement of the feature points is displayed using the display section 6, and the motion pattern is resolved (step i).

Through the processing procedure described above, the feature points of the analysis target region extracted by feature extraction from each scene of the input image are associated, and information on the movement of each associated feature point is obtained. . By sequentially tracing the positions of the associated feature points on the image, it becomes possible to effectively analyze the motion pattern of the feature points.

Therefore, the movement of a moving object placed in a natural environment can be continuously imaged as it is without the hassle of marking the moving object or inputting images of the moving object in a darkened environment as in the past. Just by inputting the information, you can easily and efficiently analyze the movement pattern. Therefore, when analyzing the movement patterns of moving bodies such as animals that are difficult to mark, the natural environment is not disturbed.
Great effects can be achieved, such as being able to analyze movement (behavior) patterns in a natural state. In addition, the knowledge of the motion properties specific to the moving object used when associating feature points can be provided depending on the moving object to be analyzed and its parts to be analyzed, so the motion patterns of various objects can be Effects such as making it possible to analyze effectively can be achieved.

Note that the present invention is not limited to the embodiments described above. For example, it is of course possible to simultaneously analyze the motion patterns of a plurality of moving bodies shown in an image inputted, and the present invention can be implemented with various modifications without departing from the gist thereof. It is.

[Effects of the Invention] As explained above, according to the present invention, it is possible to easily and efficiently analyze the motion patterns of various moving bodies in a natural environment without imposing equivalent constraints on the moving bodies. A great practical effect can be achieved, such as:

[Brief explanation of the drawing]

The figures show a movement pattern analysis device according to an embodiment of the present invention. Fig. 1 is a schematic system configuration diagram of the embodiment device, and Fig. 2 shows a moving body image to be analyzed and its region to be analyzed. FIG. 3 is a schematic diagram for explaining feature points extracted from an image and their correspondence, and FIG. 4 is a diagram showing the flow of a movement pattern analysis process procedure. l...TV camera, 3...control unit, 4...image memory, 6...display unit, 7...instruction unit, 8...table, 9...object recognition unit. Applicant's agent Patent attorney Takehiko Suzue Figure 1

Claims (1)

[Claims] means for continuously capturing and inputting images of a moving body whose movement pattern is to be analyzed; means for displaying the image of the moving body that has been imaged and input; A motion characterized by comprising means for instructing and giving knowledge regarding the movement of the part, and means for sequentially associating the analysis target part of the moving body between continuously captured and input images according to the knowledge. Pattern analysis device.