Method of Human Skeleton Extraction Based on Depth Image

Abstract

The invention relates to a method for extracting human bones based on a depth image, which includes extracting the outer contour of a human body from the depth image, discretizing the outer contour curve into an outer polygon, and extracting bones from the outer polygon; After extracting the inner contour of the human body, the inner contour curve is discretized into inner polygons and bone extraction is performed in the inner polygons; adjacent bone points are connected into continuous bone lines and all bone lines are connected into a complete picture. skeleton. The method for extracting human bones based on depth images can extract the bones in the self-occluded area of ​​the human body, effectively remove the redundant branches in the bone lines, and maintain the original topology of the human body. The extracted bone model can be used for The information describing human motion can be used in video surveillance, human-computer interaction, virtual reality and other fields, and has a wider range of applications.

Description

technical field [0001] The invention relates to the field of image processing and calculation, in particular to the field of human skeleton extraction, and specifically refers to a method for realizing human skeleton extraction based on a depth image. Background technique [0002] With the gradual improvement of computer vision and image processing technology capabilities, human motion analysis is widely used in video surveillance, human-computer interaction, virtual reality and other related fields. Human bones and joints are located from unmarked image sequences for human body analysis. Movement research has also received extensive attention. Since the connectivity and topological structure of the object shape can be well reflected by the skeleton model, the human skeleton model contains key features that can be used for action recognition, and can be used to describe the target area of ​​the human body and the relevant information of the movement, as a real-time tracking ...

AI Technical Summary

Problems solved by technology

[0007] To sum up, the existing bone extraction methods are usually based on three-dimensional human motion data or scanning data. When these data are lacking, they cannot solve the problem of unextractable occluded parts of the bones or the occluded parts that exist when the human body poses self-occlusion. The bone extraction effect is not good, there are redundant branches, and the original topology of the human body cannot be maintained.

The 3D somatosensory camera can provide depth images and bone tracking, but its bone tracking is only suitable for large-scale body movements, and is not accurate enough for self-occlusion movements

Images