Video behavior identification method based on space-time adversarial generative network

Abstract

The invention relates to a video behavior identification method based on a space-time adversarial generative network. The method comprises the following steps: extracting spatial features of an inputvideo containing human behaviors based on a spatial domain adversarial generation network, extracting time characteristics of an input video containing human behaviors based on a time domain adversarial generation network; and splicing the two dimension features extracted by the spatial adversarial generative network and the temporal adversarial generative network to obtain spatial-temporal fusionfeatures, and classifying the fused feature vectors through an SVM support vector machine to identify video behaviors. Based on a space-time generative adversarial network, learning characteristics,video characteristics and human action characteristics of the video are fully considered. M ain spatial-temporal characteristic information contained in the video is extracted by effectively combininghuman behavior characteristics for fusion, and spatial-temporal characteristics with higher representation capability are obtained based on complementarity among the spatial-temporal characteristic information, so that accurate behavior recognition is performed on the input video.

Description

technical field [0001] The invention relates to the fields of computer vision and pattern recognition, in particular to a video behavior recognition method based on spatio-temporal confrontation generation network. Background technique [0002] In recent years, with the explosive growth of image and video data in real life, it has become almost impossible to completely rely on manual processing of massive visual information data, and relying on computers to simulate human vision to complete tasks such as target tracking, target detection and behavior recognition Computer vision has become a research hotspot in academia. Among them, video behavior recognition has great application requirements in human-computer interaction, intelligent surveillance video system, video retrieval and other intelligent security, smart life and other scenarios, but due to practical problems such as occlusion, angle change, scene analysis, etc. It is still a challenging problem to analyze the beh...

AI Technical Summary

Problems solved by technology

[0003] With the development of deep learning methods and the substantial improvement of computing power, deep learning technology has also made some breakthroughs in the field of video behavior recognition, but it is still in its infancy.

Images