Aiming at the dependence on initial clustering centers of the K-means clustering algorithm, an improved algorithm is proposed. In the improved K-means algorithm, the initial clustering centers are selected according to the distribution of data spatial. The distance between two samples, the average difference of each sample, and total average difference of sample set are defined. Then the average difference of each sample is sorted. The sample with larger average difference is selected as the initial clustering center if its difference from the selected cluster is larger than average difference. Experimental results show that the stability and accuracy of the clustering results are increased by using the improved algorithm, and the convergence speed is also accelerated.