Research on optical remote sensing image detection technology based on improved YOLOv5s

To address the problem of false positives and false negatives caused by complex scenes and large scale differences in optical remote sensing image detection technology, a detection algorithm based on improved YOLOv5s is proposed. This algorithm achieves accurate detection of targets without sacrificing detection speed. The algorithm incorporates the ECA attention mechanism into the main feature extraction network to improve the algorithm's ability to extract target features. Experiments were conducted on publicly available domestic datasets. The results show that the improved algorithm's average precision mean has increased by 4.9%. The average precision for airplanes, oil barrels, and overpasses has increased by 0.4 percentage points, 0.4 percentage points, and 19.3 percentage points respectively. This effectively improves detection accuracy while maintaining a similar detection speed.