Abstract: In order to improve autonomous sensing perception of underwater vehicle on classification of seabed sediments and solve the problem of features redundancy,using genetic algorithm to optimize texture features of seabed sediments was studied.In the background of the classification and identification of seabed sediment based on a variety of seabed sediment visual texture features that are extracted based on gray level co-occurrence matrix and fractal theory,the reduction of feature dimension has been realized by using the genetic algorithm to optimize the texture features,and the texture features after dimension reduction are trained by a self-organizing mapping neural network as inputs for vi-sual classification of seabed sediments,improving the environmental awareness of underwater vehicle on underwater operation.The experimental results show that with respect to the texture features that are not optimized,optimized texture features have better classification effect in seabed sediment classification and recognition.

Key words: Seabed sediments,Genetic algorithm,Texture feature analysis,Gray level co-occurrence matrix,Fractal theo-ry,Self-organizing map