Object motion pattern recognition method and device based on depth image sequence

Abstract

A method for object motion pattern recognition based on depth image sequences is proposed, which includes: a foreground determination step, which determines the foreground object area in each frame of depth image; a difference step, which calculates the depth value of each pixel in each frame of depth image and its neighbor pixels. The difference between depth values ​​is used to obtain the neighborhood depth value difference vector; the quantization encoding step determines the area of ​​interest based on the foreground object area, and quantizes and encodes the neighborhood depth value difference vector of each pixel in the area of ​​interest into a neighborhood depth value. Differential coding; the region division step, divides the area of ​​interest into multiple sub-regions; the distribution description step, based on the neighborhood depth value differential code of each pixel, obtains the neighborhood depth value differential code distribution feature vector of the sub-region; the identification step, based on The combination of neighborhood depth value differential code distribution feature vectors identifies the motion pattern of the object corresponding to the area of ​​interest. Correspondingly, an object motion pattern recognition device based on depth image sequences is also proposed.

Description

technical field [0001] The invention relates to an object motion pattern recognition method based on a depth image sequence and an object motion pattern recognition device based on a depth image sequence. Background technique [0002] With the development of computer vision, image processing, and human-computer interaction technology, there are more and more technical requirements for vision-based human-computer interaction systems, such as interactive projectors, interactive displays, interactive electronic whiteboards, and somatosensory game consoles. More urgently, in these application areas, not only the recognition of objects in images is often required, but also the recognition of object motion patterns. In this regard, certain progress has been made in related technical fields. [0003] Patent Document 1 (US7274800B2) proposes a technique of using a depth disparity map generated by a stereo camera and performing gesture recognition based on the depth disparity map. ...

AI Technical Summary

Problems solved by technology

Because motion can cause blurring of the depth map, thereby affecting the accuracy of the shape, the method is limited to shape description, making it unsuitable for motion gesture recognition

Images