3D Face Registration on Simulated Annealing Algorithm
Wang Xiao-bi · Dianzi Ke-ji Daxue xuebao · 2015
Based on simulated annealing algorithm, this paper uses depth information to register and recognize 3D faces. The simulated annealing algorithm(SA) with global optimization capability is applied to search the global extremes, the appropriate fitness-maximum likelihood estimation sample consensus(MLESAC) and surface interpenetration measure(SIM) are selected to control the registering process for obtaining the recognition results. Based on ‘coarse to fine', we use three steps to register the 3D faces and improve the fine alignment stage. By choosing appropriate regions and classifier, we can better respond to the expression. Simulation results show that simulated annealing algorithm can escape from local optimal solution, and converge to the global optimal solution quickly. Furthermore, MLESAC and SIM would help to effectively control the registering process, thus can improve the recognition accuracy.