SAR Ship Target Detection Algorithm Based on Improved YOLOv7

A Synthetic Aperture Radar (SAR) SAR ship target detection algorithm with improved YOLOv7 is proposed for the problems of slow detection speed, high leakage rate, and poor recognition in complex scenes in SAR ship detection tasks. In the improved algorithm, ResNet50 with small size structure is used instead of the backbone feature extraction network of YOLOv7 to reduce the number of parameters of the network, and the Efficient Channel Attention (ECA) attention module is added to the feature fusion network to improve the feature expression capability of the network. The experimental results show that the improved algorithm achieves 91.7% accuracy (Precision, P), 82.5% recall (Recall, R), 24.64×10
⁶
number of parameters, and 39.27×10
⁹
floating point operations ( FLOPs) on the test set. Average Precision (mAP) is improved by 3.34% compared with YOLOv7. The number of parameters and FLOPs are reduced by 34.38% and 61.11%, respectively. The improved algorithm is used in real scene detection tasks. The improved algorithm effectively reduces the workload of detectors as well as the target miss detection rate and error detection rate, and shows good performance.