SFLA with PSO Local Search for detection sonar image

Abstract

To obtain accurate detection results, SFLA (Shuffled Frog Leaping Algorithm, SFLA) with PSO (Particle Swarm Optimization, PSO) local search is proposed to detect sonar image in this paper. K-means block clustering and clustering validity index SBWP (Simple Between-Within Proportion, SBWP) are integrated to determine the optimal clustering number of sonar image. And the encoding method based on cluster center is used to complete the initialization of the population. On this basis, in order to make full use of search experience in the past and information of related individuals, the updating method of PSO is introduced into local search of SFLA. At the same time, the individuals of sub-population all participate in the local search, not only the worst frog. Through the comparative analysis of the detection results, it can demonstrate that the proposed detection method can more accurately complete underwater object detection, relative to the traditional SFLA. Meanwhile, the proposed detection method has relatively high effectiveness and adaptability.