The Research And Implementation Of The Major Techniques In Automatic Shoeprint Analyzing And Matching System

Footprints are very important and most easily collected evidences in criminal cases. In the investigation of criminal cases, shoeprints taken from crime scenes are important in the process of serial and parallel cases. Effective shoeprint management and matching methods can shorten the time of handle a case and enhance work efficiency of criminal department. Because of manual manage and operation adopted, limited techniques of shoeprint collection in crime scenes, shoeprints cannot be used as effectively as fingerprints in criminal case investigation.In this paper, pretreatment, binarization and matching of shoeprint images in automatic shoeprint analyzing and matching system are investigated and realized based on the knowledge of image processing and analyzing. Based on concrete analysis of the shoeprint images, the feedback opinions of trail test operators in police and the research works in shoeprint image processing and matching in recent years, the algorithms are brought forward as following:(1) For the purpose of uniform shoeprint manage, improving the using efficiency of shoeprints, raising the processing speed of system and reducing the complexity of matching, the size of the original images is normalized and the direction of the original images is adjusted by means of graphical interactive mode.(2) In view of the characters of the noises in shoeprint images, a sort of adaptive mixed filter algorithm is designed. The pixels in image are classified into three classes, the pixels corrupted by pepper noises, the pixels corrupted by gauss noises and the pixels doing not be corrupted by noises. If one pixel is corrupted by noise, different filter methods are taken based on characters of the noise.(3) A sort of binarization algorithm based on blocks with many thresholds is designed. First the background regions in which the character information of shoeprint images is not included are removed, and then binarization algorithm with many thresholds is taken based on blocks.(4) A sort of binarization method making use of Laplace arithmetic operator is designed. It is the method that region growth is taken based on the edge information which is detected making use of Laplace arithmetic operator.(5) In shoeprint binary images the main noises are burr noises, hole noises and isolated point and line noises. Based on the character masks of the noises, a sort of filter algorithm for binary image is designed.(6) A sort of matching method for binary images is designed. First a two-step registration algorithm based on circle template is taken to the character images of study and one of the character shoeprints in storeroom, and the parameters of offset and rotating can be gotten. Then the character images are adjusted by the parameters. Finally matching is made and similarity is gotten.