Intelligent ship detection method based on spatial-spectrum features

Abstract

The hyperspectral remote sensing earth observation technology provides an important way for marine ship detection and identification. In order to solve the problem that the conventional detection and recognition methods mainly use single information source, and the reliability and accuracy can not be effectively improved, this paper proposes an intelligent detection method based on spatial spectrum features. Firstly, for hyperspectral single band images, the lightweight deep learning convolution model is used for target detection and recognition. From the perspective of spectral physical characteristics, the method of combining spectral anomaly detection and matching detection is used for target detection and recognition. Finally, the detection results of the two approaches are fused and screened to achieve intelligent detection and discovery. The experimental results show that based on the model and processing method designed in this paper, the detection rate of sea scene is better than 93.5%, the detection rate of port scene is better than 85.5%, and the false alarm rate is less than 40%, which can effectively ensure the intelligent monitoring and management of ships.