JP2009217789A - Object detection-recognition method and image recognition technique by moving image using time luminance frequency analysis - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To detect an object and motion using time difference even when a plurality of objects exist on an image and a small number of pixels are occupied by each object, and stably detect a reference point location when recognition of each object and luminance of a bright spot vary in object recognition. <P>SOLUTION: In an image processing method for extracting and recognizing an object captured in a moving image, time luminance frequency analysis is performed between frames. In addition, an object to be detected in the image is efficiently separated and recognized from other objects in the same manner by outputting a known signal from the object to be detected. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image processing method for extracting and recognizing an object captured in an image from a moving image composed of continuous images.

  Conventional image processing methods for extracting and recognizing an object have used a method for recognizing features of the shape, shading, and color pattern on the image, and a method for using the amount of movement of bright spots on the image.

  In addition, in the conventional image processing technique, as a technique using the feature that the image is continuous in the moving image, the separation of the target and the background and the change in the target position for each image by detecting the time difference are detected. Motion prediction is used.

  In the conventional image processing method, when the features of the shape, shading, and color pattern are used for the recognition of the object in the image, the object is recognized when the number of pixels occupied by the object in the image is as small as 1 to several pixels. It was difficult.

  On the other hand, when the number of pixels occupied by an object in the image is small, bright spot movement information is used. However, it is difficult to improve the recognition accuracy because the recognition conditions differ depending on the movement state of the object.

  In addition, when there are multiple targets on the image and the number of pixels occupied by each target is small, it is necessary to detect the target using time difference and motion detection for each target. It was difficult to recognize the target.

  Further, in the conventional method, when the luminance of the bright spot varies with time, it is difficult to stably detect the base point position.

  Therefore, an object of the present invention is to enable stable recognition of a target even when the number of pixels occupied by the target in an image is small, or when the luminance of a bright spot varies with time.

  In order to solve the above problem, the invention of claim 1 adds luminance time frequency analysis to a conventional moving image processing method for the purpose of obtaining more target feature information.

  The invention of claim 2 recognizes an object using the feature of the temporal luminance frequency of the extracted signal in the invention of claim 1.

  According to the invention of claim 3, the analysis is limited to the area to be analyzed in the invention of claim 1.

  According to a fourth aspect of the present invention, in the first aspect of the present invention, analysis is performed while moving a region for temporal luminance frequency analysis on an image.

According to the first to fourth aspects of the present invention, in the image processing for a moving image, it is possible to obtain more information than the conventional analysis method by extracting the time-frequency characteristics of the luminance signal of the target from the detection target. When the conditions are bad, such as the number of pixels occupied by the object in the image is small, the accuracy of object detection and recognition can be increased.

  According to the fifth aspect of the present invention, when the time luminance frequency characteristic of the detection target is known, the image analysis using the time luminance frequency analysis is performed in the moving image of the detection target based on the known time luminance frequency characteristic. This is to detect the presence / absence / position.

  According to a sixth aspect of the present invention, in the fifth aspect of the present invention, a light source that outputs a known temporal luminance frequency characteristic in advance is set as a detection target.

  According to a seventh aspect of the invention, in the sixth aspect of the invention, a light source that outputs a known temporal luminance frequency characteristic in advance is installed on a plurality of objects.

  According to an eighth aspect of the present invention, in the sixth aspect of the invention, the gain for a known signal is improved in the time luminance frequency analysis by synchronizing the blinking of the light source and the acquisition period of the moving image.

  The invention according to claim 9 is the invention according to claim 7, wherein the light sources that output known temporal luminance frequency characteristics that are attached in advance to a plurality of objects have different characteristics, and the signal of the object from within the moving image is converted to the temporal luminance frequency. By extracting by analysis, each object in a moving image is distinguished and detected / recognized.

  According to the fifth to ninth aspects of the present invention, it is possible to efficiently select a signal and improve the accuracy of detection and recognition of the target by using the time-frequency characteristic of the known luminance signal possessed by the target.

Further, by outputting a known signal from the detection target, the detection target in the image can be efficiently separated and recognized from other targets.

The invention of claim 10 relates to an apparatus that specifically realizes the inventions of claims 1 to 9, and an imaging device that captures an object as a moving image, and the captured moving image with a frame time interval as a sampling period. It has a temporal luminance frequency analysis unit that can automatically perform luminance temporal frequency analysis for each pixel with a computer, etc., and a temporal luminance frequency feature recognition unit that recognizes features of temporal luminance frequency extracted by temporal luminance frequency analysis. .

Hereinafter, the present invention will be described in detail based on embodiments shown in the drawings. In this embodiment, an example in which a walking person is recognized as a target image having a temporal luminance frequency feature is recognized.

FIG. 1 is a conceptual diagram of the configuration of an apparatus embodying the present invention. In FIG. 1, 1 is a walking person to be recognized, 2 is an imaging device, 3 is an acquired silhouette video, 4 is an analysis unit that performs luminance time frequency analysis of the acquired video, and 5 is a luminance time that is an analysis result of the video It is a recognition part which recognizes a target from frequency information and 6 brightness time frequency information.

The recognition target 1 is picked up by the image pickup device 2 and an image 3 serving as input data of the present invention is acquired. The image 3 is analyzed by four analysis units, and as a result, luminance time frequency information of 5 is obtained.

The analysis unit 4 uses a frequency analysis method such as filtering, discrete Fourier transform, or discrete cosine transform.

The recognition unit 6 recognizes the object from the luminance time frequency information of the object 5 obtained by analyzing the image 3 by the analysis unit 4 from the luminance time frequency characteristic of the object.

In this embodiment, for example, when the total luminance signal of the entire image is an analysis target in the analysis unit 4, since the analysis target is a silhouette of a walking person, the total luminance signal increases or decreases at regular time intervals. To do.

A characteristic of a specific frequency appears in the luminance time frequency information of 5, which is a result of analyzing the increase / decrease.

From the luminance time frequency information of 5, the recognition unit 6 can recognize that there is a person walking in the image by a method such as information matching.

Further, in this embodiment, it is possible to detect the position of a walking person by limiting the range 7 to be analyzed as shown in FIG.

Furthermore, by moving the range 7 to be analyzed and detecting a change in the image of the walking person, additional information such as the distance to the target can be obtained from the moving speed of the target, the walking speed and the moving speed, etc. It is possible to obtain.

In addition, this method is applicable even when the target is captured as very small as one or several pixels in the image.

Hereinafter, the present invention will be described in detail based on embodiments shown in the drawings. In this embodiment, as an object having a known temporal luminance frequency characteristic, an example is described in which an image arranged together with an object that is not a recognition object is analyzed with a light source that is lit at a certain period as a recognition object.

FIG. 3 is a conceptual diagram of the configuration of an apparatus embodying the present invention. In FIG. 3, 8 is a recognition target composed of a light source that is turned on at a constant period, 9 is a target that is not a detection target without a light source, 2 is an imaging device, 10 is an acquired movie, and 4 is a luminance time frequency analysis of the acquired movie. An analysis unit 11 performs output luminance frequency frequency analysis.

Specifically, the frequency analysis method 4 uses a frequency analysis method such as a filter that transmits a known frequency component emitted from a detection target.

Ten moving images are frequency-analyzed by four analyzing units. As a result, in the eleven output images, a detection target having a known temporal luminance frequency feature can be selected.

Further, in this embodiment, the gain in analysis can be improved by synchronizing the light emission of the light source 9 and the period of the imaging device 2 by a method such as radio wave reception.

1 is a conceptual diagram of a configuration of a device that embodies Example 1 of the present invention. The conceptual diagram which shows the example which limits the analysis range in Example 1 of this invention. The conceptual diagram of a structure of the apparatus which actualizes Example 2 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Walking person who becomes recognition object in Example 1 2 Imaging device 3 Captured moving image 4 Time luminance frequency feature analysis unit 5 Luminance time frequency information in Example 1 6 Time luminance frequency feature recognition unit 7 Limited analysis position 8 Recognized object consisting of a light source that is turned on at a constant period in Example 2. 9 An object that is not an object of detection without providing a light source in Example 2. 10 An output image subjected to luminance time frequency analysis.

Claims (10)

  A method for detecting and recognizing an object based on a moving image, wherein temporal luminance frequency analysis is performed between frames in an image processing method for extracting and recognizing an object captured in the moving image.   The object detection / recognition method according to claim 1, wherein the object is recognized using a characteristic obtained from a temporal luminance frequency meeting of the image luminance signal.   The object detection / recognition method according to claim 1, wherein analysis is performed by limiting a region where temporal luminance frequency analysis is performed.   4. The moving object detection / recognition method according to claim 3, wherein the region for temporal luminance frequency analysis is analyzed while moving on the image.   In an image processing method for extracting and recognizing an object captured in a moving image, a known temporal luminance frequency related to the recognition target is extracted by temporal luminance frequency analysis to detect the presence / absence / position in the moving image to be detected. Object detection / recognition method using moving images.   The object detection / recognition method according to claim 5, wherein a light source that blinks at a known temporal luminance frequency is installed in the detection target, and a temporal luminance frequency analysis is performed on the acquired moving image, whereby the detected moving image Object detection / recognition method using moving images, characterized by detecting the presence / absence / position of the object.   The object detection / recognition method according to claim 6, wherein a light source is installed on a plurality of detection targets.   The object detection / recognition method according to claim 6, wherein the blinking of the light source is synchronized with the acquisition period of the movie.   8. The object detection / recognition method according to claim 7, wherein the time luminance frequency signals of the plurality of light sources are different from each other, and the respective objects can be distinguished from each other.   An imaging device that captures an object as a moving image, and a time luminance frequency analysis unit that can automatically perform luminance time frequency analysis for each pixel of the moving image with a computer or the like using a frame time interval as a sampling period. 10. An image photographing / recognizing apparatus having a temporal luminance frequency feature recognizing unit for recognizing a characteristic of temporal luminance frequency extracted by temporal luminance frequency analysis and capable of realizing the object detection / recognition method according to claim 1.

Images