A cross-view human behavior recognition method and device based on dictionary learning

Abstract

The invention discloses a cross-view human behavior recognition method and device based on dictionary learning. The method includes: extracting spatio-temporal interest points from a training video, which is shot through multiple different viewing angles and contains multiple action classes. Video; use the k-means algorithm to cluster the spatio-temporal interest points to obtain the corresponding local features, and use the optical flow method to operate the training video to obtain the corresponding global features; use the K-SVD algorithm to initialize the dictionary, and based on the initialized The dictionary iteratively solves the local features and the global features to obtain the corresponding sparse coding and a complete dictionary, wherein the same action class under different perspectives has a sparse coding with a similarity greater than the preset value, and different Action classes have sparse coding whose similarity is less than a preset value. It can adapt to the situation of recognizing human behavior in the video when the viewing angle of the video is changed, and the recognition performance is high.

Description

technical field [0001] The present invention relates to the technical field of human behavior recognition, and more specifically, to a cross-view human behavior recognition method and device based on dictionary learning. Background technique [0002] With the development of modern information technology, people's mutual communication is no longer limited to traditional media such as text, voice and images. A large number of videos and even high-quality video signals are increasingly filling human society. A large amount of video data exists in our daily life, and it is still expanding rapidly beyond imagination. How to quickly and effectively understand and process these video information has become a very important issue. Human body movement is the core information in video, and the research on human behavior recognition in video has become the key key for computers to understand the meaning of video. [0003] The current technical method for realizing human behavior recog...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a cross-view human behavior recognition method and device based on dictionary learning to solve the problem of low recognition performance in the existing models used to recognize human behavior in videos

Images