Predator-Prey Biogeography-Based Optimization for Bio-inspired Visual Attention

摘要

Visual attention mechanism is one of the important techniques in computer vision field, and it can increase the effectiveness of computer image information processing. Biogeography-Based Optimization (BBO) is a bio-inspired algorithm for global optimization, searching for the global optimum mainly through two steps: migration and mutation. In this paper, a novel Predator-Prey Biogeography-Based Optimization (PPBBO) is utilized to solve the bio-inspired visual attention problem. In PPBBO method, BBO is combined with the mechanism of predator-prey, which can enhance the global convergence of the algorithm. The convergence property of the PPBBO is analyzed theoretically, and the detailed process is also given. Comparative experimental results with basic BBO, CBBO, CPPBBO, and Particle Swarm Optimization (PSO) demonstrate the feasibility and effectiveness of our presented PPBBO for adjusting combination of feature maps in visual attention.