Research On Shoeprint Image Feature Extraction And Registration Based On SIFT

With the rapid development of science and technology, crime scene investigation has a higher demand of evidence collection and examination. The automatic recognizing technology of shoeprint image is used to process, analysis and recognize shoeprints at the crime scene. It helps police to investigate cases. This paper will do research on shoeprint image feature extraction and registration which is based on the algorithm of SIFT(Scale Invariant Feature Transform). The main research work is as follows:For the part of shoeprint images positioning, it proposed a positioning algorithm which is based on twice positioning of the image of shoe print. After the preprocessed, the image is rotated which is based on the algorithm of the vertex angle deflection according to the image feature. Furthermore, the algorithm of equidistant vertical line as the second position is imposed on the shoe image. Finally, the shoe image is in a vertical position.Aiming at mass of data and complex calculation in the algorithm of SIFT, this paper proposed an improved algorithm of SIFT to match the shoeprint image features: on one hand, the second position was imposed on the shoe image. This advance rotating can reduce the amount of computation at the part of rotation invariant’s maintaining in SIFT. On the other hand, at the part of features matching, use the minimum Euclidean distance as the standard of shoeprint features matching. And then use the ratio of the minimum Euclidean distance to the second minimum Euclidean distance to remove invalid points. It can improve the matching efficiency.The result of experiment shows that the algorithm will not only shorten the registration time but also improve registration accuracy under the condition of keeping the image registration rate and algorithm robust.