Soft-NMS-Enabled YOLOv5 with SIOU for Small Water Surface Floater Detection in UAV-Captured Images

In recent years, the protection and management of water environments have garnered heightened attention due to their critical importance. Detection small objects in unmanned aerial vehicle (UAV) images remains a persistent challenge limited pixel values interference from background noise. To address this challenge, paper proposes an integrated object detection approach that utilizes improved YOLOv5 model for real-time surface floaters. The proposed effectively detects by better integrating shallow deep features addressing issue missed detections and, therefore, aligns with characteristics floater dataset. Our has demonstrated significant improvements detecting floaters when compared previous studies. Specifically, average precision (AP), recall (R), frames per second (FPS) our achieved 86.3%, 79.4%, 92%, respectively. Furthermore, original model, exhibits notable increase both AP R, 5% 6.1%, As such, is well-suited on water’s surface. Therefore, method will be essential large-scale, high-precision, intelligent monitoring.