Resarch Of Human Recognition Method Based On The Similarity-Grading Model

Extracting and recognizing external features is one of the key technologies in digital image processing. Exploring the recognition methods of external features can be applied in identifying all kinds of items including human being’s external features. This paper mainly focuses on the research of recognizing human being’s external features. The core issues and difficulties existing in this topic mainly included dealing with lights and noise interference, feature selection, the recognition accuracy, recognition accuracy and the rate and efficiency of recognition etc.Targeting on these requests and difficulties, the current paper recognition of human being’s external features as researching cases. The following are the specified contents of research in this paper:In view of the deficiencies in the existing methods of extracting and processing human images. This paper proposes the algorithm of self-adapting background updating rate to resolve the contradiction between foreground extraction stability and the background convergence rate in background modeling, the denoising algorithm of small rectangular is used in resolving the problem of foreground target image distortion when eliminating the foreground noise in the morphological algorithm, and the merging algorithm of minimum edge rectangular is employed in solving the problem of human image being parted caused by over-extent movement etc., ensure the integrity of the image.In view of the defects of lots of personal recognition methods, mainly represented by face recognition, which were characterized by harsh recognition requirements, weak anti-jamming capability and difficulty for characteristics to be described. This paper proposes the human recognition method based on the similarity-grading model. Adopting hierarchical fuzzy matching, this method extracted the person’s overall and local external features and evaluated the similarity grading, used ways respectively on extraction of bare limbs, removal and similarity evaluation based on skin histogram algorithm, personal image segmentation method based on color boundary segmentation algorithm, ways on color feature extraction and similarity evaluation based on HPV low saturation algorithm, and ways on texture feature extraction and similarity evaluation based on texture color feature fusion algorithm. This method provided solutions to defects of lots of personal recognition methods, mainly represented by face recognition. So, it expanded the scope of application of personal recognition.