Study On Shoe Prints Segmentation Algorithm Under Complex Background

Footprints are very important traces of evidences and easily found in the criminal cases, which have important reference value in the investigation and detection of criminal cases. In the past, footprints are hard to play their roles in the criminal investigation because of the limit to the traditional manual operating and artificial management. With the development of computer technology and image processing technology, the identification technology of footprints is brought into a new level. Some shoe print recognition systems have been put into practices, which greatly improve the identification rate of the footprints in criminal scenes and play an important role in the detection of criminal cases. However, in these systems, the input images are required to be high quality and the image preprocessing functions are very simple, which are hardly to meet the need of complex criminal scenes. Therefore, the image preprocessing part of footprints automation identification system is studied in this paper.The characteristics of the shoe print images collected from the crime scenes are summarized. The backgrounds of the shoe print images are classified, and different filtering methods are taken to remove the noises according to different backgrounds. The visual perception of the images is improved by using the multi-scale top-hat transformation which is the concept of mathematical morphology. In order to solve the segmentation errors caused by the uneven light, thresholding method based on sub-image is applied to segment the images. In another word, the images are divided into several blocks before segmentation, and the blocks are segmented after each one gets its segmentation threshold. Images are segmented by using the fuzzy C-means clustering algorithm based on watershed transformation. This algorithm not only solves the problem of over segmentation by watershed transformation, but also enhances the running speed of the fuzzy C-means clustering algorithm. The color information of the images is used to segment the shoe print images. The images are segmented in terms of the differences between the object and background of the shoe print images on the values and variances of the three components of R, G and B. The proposed algorithm is valid for the shoe print images which have colorful background and partly gray shoe texture. In order to remove the interference of straight line of the image segmentation results, Hough transformation is applied to detect and remove them.All the proposed algorithms are carried out in the MATLAB platform. Experimental simulation results show that these algorithms can partly solve the problems of the shoe print images segmentation under complex background.