With the development of space remote sensing technology,the need for rapid detection and identification of marine vessels is growing.This paper proposes a ship target detection method combined with high- and low-resolution remote sensing images based on deep learning.Firstly,the YOLO v3 model is used to quickly screen out the target ships in wide-range,low-resolution satellite remote sensing images.To the high-resolution satellite remote sensing image information,a RetinaNet model based on attention mechanism is proposed for exact matching and classification of the target ship.Simulation experiments show that the improved RetinaNet model has a good effect in target detection,and the use of the satellites with two kinds of resolution for collaborative work can greatly improve the working efficiency.