Standard Human Body Image Spectrum Clustering Segmentation And Measurement Based On Adaptive SLIC

Image segmentation,as the basis of pattern recognition and image analysis,has always been a challenging topic in the fields of digital image processing and computer vision,and has important research significance.At present,a large number of image segmentation algorithms have been produced,some of which have also been applied to standard human image segmentation,providing theoretical support for the development of three-dimensional human reconstruction,human motion tracking,virtual fitting and other fields.On the basis of summarizing the existing typical image segmentation algorithms,this paper proposes a new standard human body for the difficulty of standard human body image segmentation in different backgrounds and the inaccurate matching of human feature points due to lack of prior conditions.Image segmentation algorithm.As an application,the extracted key points of the human body are fitted with three-dimensional data to obtain more realistic human body data.The main research work is as follows:1.Aiming at the problem that the initial seed points need to be set artificially when the simple linear iteration(SLIC)algorithm is segmented,an adaptive SLIC algorithm is proposed.The improved algorithm divides the image into several very small regions,and introduces CV energy segmentation in each region,which makes better use of the gray information of the image,and automatically sets the initial seed points,so that the segmented super-pixel blocks are more suitable for the edge of the image color block.2.In order to avoid the influence of complex environment on the segmentation precision of human body image,the human body template is used to mark the region of interest in the image,and the similarity function is rewritten by the template information in the region,and the super pixel block is clustered and the super pixel block is completed.Spectral clustering of human image segmentation improves image segmentation accuracy.3.Aiming at the problems of local fitting and too many iterations in traditional ASM algorithm,an ASM feature point extraction algorithm based on human contour is proposed.The improved algorithm defines a set of template datum points on the basis of human contour lines,and improves the accuracy and speed of feature points extraction from human front and side by means of datum point calibration,point set calibration and changing the direction of feature points search.After getting the feature points of the key parts of human body through the above steps,BP neural network is used to learn the relationship between multiple parts of human body.On the collected human body data,the neural network achieves the fitting of human body three-dimensional data within the error range of 2 cm.Finally,this paper summarizes the contents of this study and prospects the future work.