Research On Scene Shoeprints Contour Extraction And Estimation Algorithm

Shoeprint is one of the marks that often can be found in crime scenes. People often use shoeprints as evidences in criminal cases. With the development of computer vision technology, shoeprints play an increasingly important role in solving the criminal cases. However, the shoeprints that left at crime scenes are always not satisfied due to the complex scene environment. Such as: incomplete, unclear, etc. These conditions aren’t conducive to analyze or utilize shoeprints’ information. Therefore, how to use shoeprints more effectively has become a hot research topic.The scene shoeprints are often used for shoeprint matching retrieval. The retrieval process includes extracting feature of the scene shoeprints and matching feature with the shoeprints in database. Shoeprint contour is not only an important feature of shoeprint, but also an important feature needed by shoeprint matching retrieval. Shoeprint contour can be divided into shoeprint pattern contour and shoeprint outer contour. With the help of shoeprint contour, we can analyze the shape and spatial distribution of the shoeprint more deeply and reduce the impact of adverse factors so that we can provide a good foundation for shoeprint matching retrieval. In this paper, our main research content is the extraction and estimation of the shoeprint contour. The research work includes the following three aspects:1) The scene shoeprint pattern contour extraction algorithm based on level set evolution. Due to the low quality of scene shoeprints and the difficulty of shoeprint pattern extraction, we provide a shoeprint pattern extraction algorithm based on level set evolution. The proposed algorithm can overcome the adverse factors of environment, improve efficiency and quality of the scene shoeprint pattern contour extraction, and provide a good foundation for the shoeprint outer contour estimation.2) The complete shoeprint outer contour estimation algorithm based on iterative fitting. Based on the extracted shoeprint pattern contour, we propose an outer contour estimation algorithm to get the outer contour of the complete shoeprint. Firstly, we use an iterative fitting algorithm to select real edge points. Then, we estimate four edges of the shoeprint outer contour based on real edge points. After correcting upper and lower edges’ contours and connecting four edges’ contours, we can finally get a smooth outer contour of the shoeprint by using Fourier descriptors.3) The incomplete shoeprint outer contour estimation algorithm based on curvature feature. We propose an outer contour estimation algorithm based on curvature feature to get the outer contour of the incomplete shoeprint. Firstly, we manually select a high significance and high reliability contour fragment from the incomplete shoeprint outer contour. Then, we use part shape matching algorithm based on curvature features to select the best match complete shoeprint contour in database as the estimated incomplete shoeprint outer contour.The experiments prove that all the proposed algorithms have good effects. The extraction of the scene shoeprint pattern achieve a good effect with some human intervention; The estimation of complete shoeprint contour can achieve a good estimation result for most complete shoeprints; Based on the shoeprints’ contours in database, incomplete shoeprint contour estimation can also achieve a good estimation effect by using selected contour fragment which meet the requirements.