Identification of Class number of ship radiated noise based on RK-Means algorithm

Abstract

Ship radiated noise recognition is one of the basic problems in underwater target recognition. To identify the number of ship types according to ship noise, the ship scale can be predicted in advance and the next step of processing can be carried out. Aiming at the problem of identifying the number of ship types based on ship radiated noise, a RK-Means (Recurrent K-Means) is proposed in this paper. According to the short-time energy characteristics of the target signal, the dimension of the feature data is reduced by principal component analysis; The K-Means algorithm is used to cluster and identify the dimension-reduced signals through cyclic iteration; A single connection algorithm is used to determine the appropriate clustering center, which reduces the similarity value of data objects between classes and makes the difference value between classes large. Furthermore, the algorithm can automatically determine the K value, which is the number of ship types. Verified by the experimental data, compared with the contrast algorithm, the K-value recognition result and clustering effect of the RK-Means clustering algorithm are better.