An Adaptive Splitting and Merging Clustering Algorithm of the Moving Target Segmentation

For the issue of multiple moving targets segmentation in intelligent monitoring system, an adaptive splitting and merging clustering algorithm of the moving target segmentation is proposed. First, it uses the time-domain information for foreground image segmentation through the sample variance background modeling algorithm, thus obtains the foreground image containing multiple moving targets. It defines pixel space connectivity rate and designs a perpendicular split method for the initial cluster adaptive splitting and merging. Without pre-set number of initial cluster, the self-organized iterative clustering segmentation algorithm can complete multiple moving targets segmentation. Experimental results show that the proposed algorithm is suitable for multiple moving targets segmentation, and the segmentation results are consistent with the human visual judgment. The use of space connectivity information improves the iterative algorithm convergence speed, thus it has good real-time.