Human motion tracking method based on second generation Bandlet transform and top-speed learning machine

Abstract

The invention discloses a human motion tracking method based on second generation Bandlet transform and top-speed learning machine, which is mainly used for solving the problems of inaccurate video image feature representation, high time complexity in regression function learning and inaccurate tracking result in the human motion tracking of the prior art. The realization process of the method comprises the following steps of: firstly, preprocessing the video image to obtain an original joint point three-dimensional coordinate matrix Y; extracting the second generation Bandlet 2 image feature X of the processed video image; taking the extracted Bandlet 2 image feature X as input, taking the three-dimensional coordinate matrix Y in the human body as the output, and using a top-speed learning machine to learn the regression function; using the regression function obtained by learning through the top-speed learning machine, taking the bandlet 2 feature X of the new video image as the input, estimating the three-dimensional attitude data of human body in motion. In comparison with the conventional human motion tracking method, the method provided by the invention is fast in the training process and accurate in image feature representation, and can be used for motion capture, man-machine interaction, video monitoring, human object recognition and three-dimensional attitude recovery.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a method for realizing human body motion tracking in the field of computer vision, which can be used in the fields of sports training, animation production, and video monitoring. technical background [0002] Human motion tracking is one of the major hotspots in the field of computer vision in the past two decades. Human motion tracking has been initially applied in many fields such as motion capture, human-computer interaction, and video surveillance, and has great application prospects. Accurately recovering 3D human pose from video sequences and realizing human motion tracking is a long-standing problem in the field of computer vision. The realization of human body motion tracking mainly includes two steps: the first step is to realize the accurate representation of video image features, and the second step is to learn the regression function from video image fe...

AI Technical Summary

Problems solved by technology

However, most of the image feature representations describing the human body are based on contour and edge information, which is not rigorous in theory, and it is difficult to accurately describe the internal information of the image.

These edge-based image feature representation methods also face a major problem: the rapid transformation of the video image often jumps along the edge curve discontinuity, which on the one hand will cause the gray level discontinuity of the closed boundary to be blurred, and on the other hand will cause texture changes. Does not gather along geometric curves

These regression methods need to use a large number of training samples and a large amount of training time in the process of learning the regression function, and the computational complexity is high. The obtained regression function depends to a certain extent on the complexity of the database, that is, the obtained results are low in stability.

Images