Research on Ship Detection in the SAR Image Algorithm Based on Improved SSD
Feng Min, Peng Liu · 2021 4th International Conference on Artificial Intelligence and Pattern Recognition · 2021
With the rapid development of deep learning, target detection technology has been widely used in various fields. At this stage, ship detection method in the SAR image based on the SSD algorithm, the accuracy was more improved than traditional methods, but the SSD algorithm takes a long time, and it was difficult to detect and locate the ship target in real time in the missile-borne SAR image. In addition, the SSD model has a large amount of calculation, and has a deviation in the detection of small targets. A ship detection method in the SAR image is proposed in this paper. The method improves the shallow network structure of the SSD algorithm, increases the network of different modules for feature fusion, reduces the loss of information in the training process, improves the feature extraction ability of small ships in the SAR image, and increases the receptive field of shallow features. Experiments was carried on SSDD and SAR-Ship-Dataset datasets. Compared with SSD algorithm, It was increased about 6% and 7% of mAP, which reduced the calculation amount of convolution, and the model size was reduced by about 7M and 10M.