K-means without K in multi-view SAR clustering
Xiaoyan Zhou, Tao Tang, Gangyao Kuang · 2021 CIE International Conference on Radar (Radar) · 2021
K-means algorithm realizes clustering through the correlation between data, which is a typical unsupervised algorithm. But the K-means algorithm needs to set the corresponding number of clusters K first, and the accuracy of K greatly affects the performance of the algorithm. In multi-view SAR images, the similarity value of adjacent azimuth images of the same category is higher than the similarity of adjacent azimuth images of the different category. Therefore, the K can be estimated through the relationship between the azimuth angles, and the negative impact of clustering caused by the inaccurate setting of K can be avoided. Therefore, in this paper, we propose a method for estimating the number of clusters K in SAR images based on the similarity relationship of multi-view SAR images. This method simplifies the use of K-means in multi-view SAR images.