Ship Target Detection in SAR Images Based on FPGA

Since Synthetic Aperture Radar (SAR) has the characteristics of all-day and all-weather operating, it is significant to carry out maritime ship target detection based on SAR images. Although deep learning algorithms have been widely used for target detection, there are problems of high model complexity and slow detection speed, which cannot meet the requirements of embedded deployment and real-time detection in certain complex environments. To address the above problems, this paper investigates the SAR image ship target detection method based on deep learning technology and FPGA devices. Firstly, a lightweight target detection network based on You Only Look Once (YOLO) is proposed, which combines EfficientNet with the YOLO backbone network to reduce the parameters and size of the network and increase the speed of detection. Secondly, the model is compressed and lightweighted by operations such as quantization to facilitate deployment on resource-constrained devices. The results demonstrate that the improved algorithm proposed in this paper can improve the detection speed while guaranteeing certain accuracy, and realize the deployment on FPGAs, demonstrating the feasibility of the deployment scheme, which can better meet the needs of terminal deployment and real-time detection.