Abstract: A new particle swarm optimizer based method for detecting features of pointbased models is presented in this paper. It solves the fast displaying of the characteristics of large scale models. It applies modified particle swarm optimizer (PSO) to the feature detecting of object space so as to complete the search of multiobjective regions. By redefining the particle, fitness function, initial terminal condition, g-Best, p-Best and update rule of PSO, the method presented in this paper combines the local search with global search for the purpose of finding multiple targets fast. It detects feature points by fitness function which is able to estimatelocal surface variation and marks feature points to display the characteristics of model fast. The experiments show that the improved algorithm is suitable for fast detecting features of point-based large scale models.

Key words: particle swarm optimizer, multiobjective, feature, curvature 