An Improved YOLO-Based Method for River Oil Spill Detection and Drift Velocity Estimation

Oil spills on water surfaces represent significant environmental pollution events that pose serious threats to aquatic ecosystems and public health. Currently, most oil spill detection studies have focused on marine environments and remain insufficient for addressing narrow, complex flow conditions found in riverine environments. However, oil spills in inland rivers can have severe consequences, including direct threats to drinking water safety and urban ecosystems. Therefore, this study proposes a real-time oil spill detection approach tailored to riverine conditions. The method employs a lightweight object detection network for rapid identification of oil surfaces, integrating it with a target tracking algorithm to extract bounding box centroids and compute oil drift velocity. To collect training data, a river oil spill simulation environment was constructed, and ablation, as well as comparative experiments, were conducted to assess detection accuracy and algorithmic feasibility. The results demonstrate that the proposed method effectively identifies riverine oil spills and measures drift velocity in real-time, providing a practical solution for monitoring inland water pollution.pollution.