Study Of The Feature Extraction And Classification Of Footprint Images

In the need of case solution for police, the researches on feature extraction and classification of footprint image play a very important role in the footprint analysis and match system. This paper analyzed the characteristics of footprint image, finished the enhancement, segmentation and feature extraction algorithms according to different types.1. This paper gave an image enhancement algorithm based on wavelet transform and region segmentation. The contrast of footprint image can be improved, and the noise can be restrained by this algorithm.2. This paper used the k-means cluster algorithm to segment the footprint image. The cluster samples and center numbers are constructed based on the footprint image characteristics, then the footprint is extraction from the background.3. This paper used the two-dimensional histogram ostu method to segment the footprint image. In this method, the two-dimensional histogram of gray level and neighbor mean gray-level is constructed, and ostu method is extended to two-dimensional condition. The threshold is selected to segment the footprint image and the segment function is optimized to improve the capability of anti-noise.4. The features are extracted to classify the footprint image based on different types. The area, ratio of width and height, solidity, form factor and equiv-diameter of objects contained in the footprint image are extracted to recognize the circle, pot and block objects. The cross points are detected after thinning the cross, wave and line objects. The chain codes are used to calculate the numbers of corner point. The numbers of cross points and corner points are used to recognize the cross object, the wave object,and the line object.