Three-dimensional Footprint Of 3d Surface Segmentation And Recognition

In recent years, biometrics has been a research hotspot in the information technology domain. To use the abundant physical and the behavior characteristics more sufficiently and effectively, it is need to carry on the further exploration and research in more widespread domains.Under the conditions of standing or walking, the sole surface of the human body marks a footprint on the ground. Footprint is a type of important trace information belonging to biometrics. Although footprint verification is applied to the domains such as criminal detection for a long time, in which a certain expert experiences are accumulated, but the domestic and foreign research on footprint biometrics is far from sufficiency for the reason that the footprint's forming condition is variable, and lot of difficulty should be faced in data acquisition and measurement, characteristic description and recognition. There are massive foundational questions and application technologies to be researched.Researching on 3D footprint depth data, along with utilizing the range image analysis and pattern recognition technologies, this thesis addresses the scientificity of 3D footprint biometrics recognition. The main contributions are summarized as follows:1. A novel range image segmentation algorithm based on randomized Hough transform is proposed. The algorithm finds planar regions by utilizing randomized Hough transform and realizes the range image segmentation. Because the Hough transform can use the whole information insensitive to local noise, while avoids estimating the curve surface curvature which is vulnerable to the noise, the algorithm has good immunity to noise. Comparing with four other typical range image segmentation algorithms by experiment results, the proposed algorithm shows better performance. Here, a current popular range image database (ABW image database) is taken as test data.2. A new range image segmentation algorithm based on the principal curvatures and directions is proposed. The algorithm selects the peak regions (two principal curvatures are negative) of range image surface as the seed regions, and grows with the changing of the principal curvatures and directions, and finally ceases growth at local maximum positions of the principal curvature. Here, edge detection is applied in the region growth rule. In this way, this proposed algorithm takes merits of the combination of edge detection and region growth to realize the accurate boundary location of 3D curved surface.3.Aiming to the low resolution and serious noise of the range image, we integrates multi techniques, interpolation, smoothing, segmentation based on the principal curvatures and directions, multi-scale data fusion, morphological grayscale reconstruction, and watersheds segmentation, to constitute one practical 3D range image segmentation scheme. The segmentation method keeps the real boundary information, and at the same time eliminates the false edge caused by noise. The acquired region has obvious physic significance.