Abstract: Curve fitting plays very important role in preprocess of object recognizing. A particle swarm optimization (PSO) based multi-object optimization algorithm was proposed in this paper to implement the smoothness fitting quickly for image with complicated noise around the target-area. The external repository and strategy of diversity were employed to prevent the PSO from converging too quickly. Moreover, the search policy of split and-merge made the selection of knots parameter more flexibly in l}spline bases computation while getting the discrete control points set of the target area. I}herefore, curve fitting can be achieved by the multi-resolution interpolation. As shown in experiments, this algorithm can get the approximation curve quickly, eliminate the noise from the target area, and satisfy the requirement of image based 3-D reconstruction as well.

Key words: Curve fit, Particle swarm optimization, B-spline curves, Multi-object optimization, Pareto optimal