The Studies Of Positioning The Human Body Joints Based On RGB And Depth Information

The technology of human motion detection and analysis is a rapidly developing research direction in recent years.This technology is widely used in Virtual Reality,intelligent transportation,somatosensory game and other fields.Human joints positioning is one of the key issues in the technology of human motion detection and analysis.This thesis works on three aspects: extraction and classification of corner points of the human skeleton line,head joint positioning,other joints positioning except head joint.(1)The general method only gets the human skeleton line without further processing,this thesis increase the work of extraction and classification of corner points of the human skeleton line in order to find those end points and turn points.Firstly,we preprocess the human body depth image with filtering,dilating and eroding,and binarization.Then thinning the preprocessed depth image with Zhang-Suen refinement algorithm to get the human skeleton line.Detecting and finding the corners of human skeleton line with Shi-Tomasi corner detection algorithm.At the end we filter these corners into two parts: the set of turn points and the set of end points.Because the human joints is included in the end points and turn points of the human skeleton line,we positioning the human joints without head joint from these two sets.(2)In the general method,set the head end point of the human skeleton line as the head joint.But the most accurate position of the head joint is located in the center of head.In order to positioning the head joint at the center of head,this thesis method is: we shoot the RGB image with kinect,then make the depth image's coordinate and RGB image's coordinate consistent,and erase parts of the RGB image's background in order to reduce the amount of data.Detecting and tracing the face with camshaft moving target tracking algorithm.At last,positioning the head joint at the center of face.This method positions the head joint at the center of head proximately,and the deviation is more smaller.(3)In the general method,positioning the human joints just according to the general human body size ratio.This size ratio is universal and general,but everybody has its own special body size ratio different from others.In this thesis,we position the neck joint from the set of turn points firstly.Then calculate the distance of head joint and neck joint as l(head-neck),and calculate other distances among joints based on l(head-neck).Positioning the human body joints from the sets of turn points and end points in the area of those distances ± 20%.This method positions human body joints at the real location proximately.