A detection method of human motion features in cluttered background

Abstract

The invention discloses a detection method of human body action features in a cluttered background. The specific steps are as follows: a salient point detection method based on information entropy proposed by Kadir is used in the airspace to detect airspace interest points; the detected airspace interest points are Background interest point suppression; use 1D Gabor filter in the time domain to obtain candidate spatiotemporal feature points; after non-maximum suppression of the spatiotemporal feature point response function, the point at the local maximum of the response function is the final feature point . The present invention adopts the interest point detection method based on information entropy in the air domain, so more abundant feature points can be detected. By suppressing the background interest points, most of the detected feature points are located in the human body, effectively improving the feature points. The accuracy of point detection provides a guarantee for the subsequent correct recognition of human actions.

Description

technical field [0001] The invention relates to the technical field of machine vision, in particular to a method for detecting human action features in a cluttered background. Background technique [0002] Vision-based human action recognition technology is an important research topic in video understanding. It has far-reaching theoretical research significance and wide application in the fields of intelligent video surveillance, content-based video annotation and retrieval, intelligent human-computer interaction interface and animation synthesis. prospect. [0003] At present, the recognition of simple human actions in simple scenes has made some progress, but the recognition of human actions in complex scenes still faces many difficulties. In the real environment, due to the influence of factors such as cluttered background, occlusion and illumination changes, the accuracy of human action recognition is not high. Common features can be divided into static features, dynam...

AI Technical Summary

Problems solved by technology

[0004] Although spatio-temporal features are a sparse feature detection method, too few feature points will lose important motion information, which will affect the accuracy of behavior recognition in subsequent steps

When performing feature detection on human actions in a cluttered background, most of the feature points detected by most spatio-temporal feature detection methods are located on the background, and the false detection rate is high.

Images