A Human Shape Modeling Method Based on Graph Theory

Abstract

The present invention provides a human body shape modeling method based on graph theory. A graph theory-based human body shape modeling method includes: using line segments to describe the various components of the human body shape; describing the various components of the human body shape The line segments of are connected smoothly, and the whole human body shape is described by the undirected graph connected by line segments. The method of the invention can complete the automatic detection of the human body without manual intervention, has a simple recognition algorithm, and can save a large amount of calculation workload.

Description

technical field [0001] The invention relates to the technical field of human target recognition, in particular to a graph theory-based human body shape modeling method. Background technique [0002] In recent years, the following methods for human body detection from images have been mainly proposed. [0003] 1) A human body recognition model based on a large number of pose expressions is proposed. The starting point is that the human body contour is quite different from the contours of other objects, and the human body contours in continuous video images have great similarity, and based on this A method for detecting and classifying different humanoid individuals in video color optical images is presented; [0004] 2) Through the method of describing the appearance characteristics of the human body based on the dense growth random tree, the weak classifier is constructed into a strong classifier by using the method of adaptively improving the classifier, so as to complete ...

AI Technical Summary

Problems solved by technology

[0005] 3) Use Markov logic network to deal with the problem of human detection under the condition of occlusion;

[0009] When the above methods are used to automatically detect the human body, the recognition algorithm is complex and requires a large amount of computing workload.

Images