Abstract: K-means clustering is widely used in image segmentation due to its simplicity and rapidity.
However, it is easy to fall into local optimum, leading to poor image segmentation results. In order to
overcome this disadvantage of K-means, this article proposes a mixed image segmentation algorithm
based on random weight particle swarm optimization (RWPSO) and K-means clustering. In the early
stages of the algorithm running, it can avoid falling into local optimal using the global search
capability of RWPSO. In the later stages of the algorithm running, it can achieve fast convergence
using the local search capability of the K-means clustering. Experimental results show that RWPSOK
algorithm can effectively overcome the weak global search capability drawback of the K-means
clustering. It can significantly improve the image segmentation results. Compared with traditional
particle swarm K-means clustering algorithm (PSOK), RWPSOK algorithm has better segmentation
results and higher efficiency.

Key words: random weight, particle swarm optimization, K-means clustering, image segmentation