Abstract: Because there are so many types of multimedia image data resources, dimension is bigger, directly on the clustering of fusion result is not ideal, need to optimize the clustering fusion algorithm. The existing based on neighborhood remain embedded multimedia projection clustering image data fusion algorithm, the first to use method of neighborhood remain embedded multimedia projection image data sample dimension reduction processing; Then the reduction process of multimedia image projection for clustering data fusion. The present clustering algorithm performance is poor, poor fusion effect.In this paper, a multimedia projection image data clustering based on particle swarm optimization algorithm fusion, first using the gradient descent of error back propagation training, selection of multimedia projection image data clustering members, provide accurate data foundation for the subsequent clustering fusion; A comprehensive evaluation of the definition of diversity and correctness and then use standard, realize the selectivity of multimedia projection image data clustering fusion; Finally on this basis using the particle swarm algorithm of global optimization for its optimization. Through the experiment analysis, the results show that the proposed algorithm can improve the quality of multimedia projection image data fusion and clustering accuracy.