The Human Face Contour Extraction And Authentication

At present,digital face detection is a hot topic of scientific research,widely applied in the identity authentication,the video surveillance,the human-computer interaction and image data retrieval and so on.Among of them,the person face examination and the localization,the human facial contour characteristic extraction and technologies and person face recognition,are key technologies links in the identification authentication system.Active contour model(also known as Snake model) is a hot research in computer vision,which can achieve the object boundaries with the energy minimum function,the especially qualify in the contour characteristic's extraction,has laid the foundation for the next step of person face recognition and the identification authentication.This article has conducted the thorough research to the correlation theories and the key technologies,and has developed the corresponding software system.Firstly,the paper presents a new contour extraction algorithm based on PSO and GVF.PSO is proposed to find the optima of snake points for rapidly converging near image edge;Then the interpolation algorithm is applied to gain the object's rough contour, that is used as the next initial position for the GVF convergence,finally get the accurate contour lines.Experimental results showed that the new algorithm can not only extractive the accurate contour but also has some anti-noise performance, easy to implement and faster.Secondly,this article proposes a match method for the fuman facial contour based on the Bayesian classifier.First it establishes the initial control point around the target's image, make use of the algorithm that the front introduced to converge to these control points,withdraws the person facial contour characteristic;Then it uses directly smallest ellipse fitting technology to establish ellipse template;Finally it completes the fuman facial contour's match based on the smallest error rate Bayesian decision-making,gives the experimental result and analysis,the experiment proved that this method has high recognition rate of accuracy.Finally,this article introduce the system's organization frame, under the MATLAB 7.0 and VC + +6.0 environment complete the person face recognition system.Experiments indicate that the method was simple and effective.