Shape Features Extraction And Analysis For Planar Barefoot Impression

With rapid development and wide employment of information technology, biometrics has been a research hotspot in the information technology domain in recent years.Footprint is a kind of trace information which is left on the ground by the sole surface when the human body stands or walks. The shape features of footprint depending on foot skeleton have relative speciality and stability, which are basis of distinguishing identity. As footprint has special relation with people's height, posture, gender, age and walking habit, this relation has been confirmed and successfully applied to long-term practice in forensic science.Researching on planar barefoot impressions, along with utilizing the forensic expert experiences, image processing and analysis theories, statistic analysis and pattern recognition technologies, this thesis addresses the scientificity of planar barefoot impression biometrics recognition in aspects of planar barefoot impression segmentation algorithm, shape feature extracting and describing methods, relations between shape features of planar barefoot impression and physiological features of human body, and the method of distinguishing identity based on planar barefoot impression. The main contributions are summarized as follows:1. A planar barefoot impression image segmentation method base on Gaussian fitting is proposed. As grey histogram of planar barefoot impression appears a multi-peak feature, the algorithm detects grey sections of the objects roughly based on the peaks and valleys of histogram. The sections of grey histogram are approximated by Gaussian functions separately. Finally the optimal segmentation thresholds are accurately calculated by Gaussian density functions. The algorithm effectively avoids not only the shortcoming of mass calculating of multi-thresholds segmentation, but also adjacent peak sections interfering when fitting the grey section of object by Gaussian functions. Experimental results show that the proposed method can be effectively applied to planar barefoot impression segmentation.2. Combining the expert knowledge with researches on planar barefoot impression image segmentation and foot skeleton structure, a group of geometric features are extracted, such as distances and angles, on the basis of establishment of an uniform coordinate system, contour subsection and region partition of planar barefoot impression. Morphological features based on regions and contour curves of planar footprint are extracted by utilizing image analysis theory.3. Considering the advantages of partial least-square regression(PLS) in analyzing multivariate data with few observations, the relations between shape features of planar footprint and physiological features of human body are analyzed by PLS method. Experimental results show that people's height and weight can be predicted roughly by shape features of planar barefoot impression proposed in this thesis. When the tolerance of height prediction error is [-5cm,5cm], correct prediction rate of height is 83.65%; When the tolerance of weight prediction error is [-5kg,5kg], correct prediction rate of weight is 71.63%.4. Aiming at the specific request of planar footprint biometrics recognition, we propose a...