关键词: 遥感图像;K-Means;聚类原理;学生t分布混合模型;烟花算法;

Abstract: Using appropriate image processing techniques and data models is the key to accurately interprete satellite remotely sensed river images. Aiming at the current technical problems, this paper starts from improving the accuracy and noise resistance of remote sensing river image segmentation, and proposes a firework-based optimized K-Means clustering and student's t-distribution Mixture Model (TMM). The algorithm first uses the Fireworks Algorithm (FA) to solve the initial clustering center of K-Means clustering, which improves the clustering effect and can obtain the preliminary segmentation results of remote sensing images. By using the preliminary segmentation result as initial value, the TMM is established, the EM algorithm is used to determine the final value of the parameters, and the Bayesian formula is used to complete the secondary image segmentation. The validation results show that the proposed method is better than existing algorithms in terms of segmentation accuracy and stability, and can achieve the interpretation of remotely sensed river images more effectively.

Key words: remote sensing image; K-Means; cluster method; student’s t distribution mixed model; fireworks algorithm