Research On Palmprint Line Feature Extraction And Classification Based On Steerable Filters

As a part of the feature on skin, palmprint is a kind of genetic gene in the human body, which is formed as a fixed model by the interaction of genetic and environmental factors in the first three months of embryonic development, and since it demonstrates high stability. Since many diseases are related to heredity and environment, so it is helpful for early detection and early treatment of disease by discovering the relation of palmprint characteristics and disease.The main line is a kind of specific feature of palmprint, which has the characteristics of stability and uniqueness. Therefore, we focus on how to extract the mainline from the palmprint and correctly classify palmprint patterns based on the mainline characteristics and feature points in order to facilitate the further research on the relation of palmprint and breast cancer. However, the current mainline extraction algorithms mainly focus on how to obtain feature points in the main line and do not extract the complete main line of palmprint. Therefore, in this paper, the palmprint image obtained by non-contact acquisition is used to extract the main line feature points, recover the complete palmprint line and classify the palmprint patterns based on the spatial location of mainline. The main work of this paper includes:The method of extracting the region of interest of palmprint image is studied. The extraction of palmprint region of interest is the key preprocessing for further palmprint feature extraction. The correct and effective ROI area can get the complete palmprint line. In the research, we first use a global iterative algorithm for palmprint image binarization, and then the maximum inscribed circle algorithm is applied to extract the minimal outer square of the maximal inscribed circle to obtain region of interest in palmprint. The algorithm can feasibly obtain a complete ROI region including the palm main line.A method of using directional tracking algorithm to extract the feature points of palmprint is proposed. First the steerable directional filter is applied to filter the ROI of palmprint image. The concluded angle of extracting palmprint mainline is 70° after experiments. Then, the feature points are extracted by the directional tracking algorithm, which utilizes the palmprint mainline space position information to extract the line features of palmprint. The algorithm can extract relatively complete feature points and effectively distinguish the palmprint feature points from noise points.Finally, the least square fitting algorithm is proposed to get the palmprint mainline and palmprint is classified based on the spatial position relationship between the mainline. The algorithm analyzes the palmprint lines treated as single pixel lines. Experiment results demonstrate that the algorithm can obtain continuous palmprint line, and effectively perform palmprint pattern classification. The correct rate of palmprint pattern classification is 82.12%.