Research Of Shoe Soles Images Retrieval Based On Shapes Features

Shoe soles images are a class of special computer images, they can reflect the structural characteristics of soles'appearance. How to inspect soles quickly and exactly is a important task for theory and practice of forensic technology. The retrieval of content-based plane soles makes use of image processing and pattern recognition techniques to identify whether the soles are left by certain shoes, which is of great significance to the development of forensic technology.The paper focuses on designing and developing the plane shoe soles retrieval system through deeply researching the features of the soles images. The software developed breaks the limit of paying attention to experts'experiences. It applies quantitative methods, excludes individual subjective factors and scales data seriously for obtaining the result. The theory and technology of shoe sloes recognition are becoming more and more systematic and objective.In this paper, we aim at extracting shape features from plane shoe sloes and propose a series of practical methods based on shape features of outside contour and inside pattern. The main works are as follows.(1) We put forward a new method for slant correction of sloes, which is based on the angle shaped by the longest and horizontal axis. And then segment the shoe soles and construct coordinates for further features extracting.(2) Expatiate a depicting method for outside contour based on aspect ratio and feature points of each segment. The aspect ratio refers to the ratio between the width of top, palm, bow, heel of shoe sloes and the longest axis. And the features points of outside contour indicate inflexion and tangent point. The outside contour of shoe sloes mainly consists of convex arc, concave arc and line. The inflexion refers to the smooth transition point of convex arc and concave arc, and the transition points of convex, concave and line are called tangent point.(3) The inside pattern is also depicted through kinds of shapes. We propose a method to describe the pattern shapes based on distance and angle, which not only extracts the shape features of pattern but also considers the relationship of spatial position.(4) In order to improve space efficiency and the speed of retrieval, train the disconnected textural features of all sloes and then gets codebook. Use index value of codebook and centroid coordinate to describe each disconnected textural feature.This paper mines the features of shoe soles images on the basis of the existing shape-based image retrieval, proposes the method of extracting outside contour and inside pattern of shoe soles, and retrieves the shoe soles images using shape features. The experimental prototype system shows the method proposed works well.