SAR Image Ship Detection Based On Deep Learning

China is a maritime power, and ship detection is particularly important under complex sea conditions. At present, deep learning plays an important role in the SAR image ship detection field. An improved yolov4-Tiny detection algorithm is proposed in this paper. The improved algorithm introduces the attention mechanism unit to enhance feature extraction and make the target feature more pro-minent. The Batch normalization optimization data set is used to increase the robustness of the training model and effectively reduce the gradient disappearance or gradient explosion. Cosine annealing is used to optimize the learning rate and speed up the fitting of deep learning model. On the basis of realizing real-time detection, the whole network further improves the detection accuracy. The experimental results show that the MAP of improved Yolov4-Tiny algorithm is 75.56%, and the FPS is 30.