A Novel Method for Traceability of Crude Oil Leakage on Offshore Platforms Based on Improved YOLOv5 Model Multi-Target Identification and Correlation Analysis

In order to meet the requirement of oil leakage monitoring on offshore platforms, The paper investigates a method for identifying oil spill targets and potential associated targets that may lead to oil leakage based on the YOLOv5 model. In response to practical issues such as the complex structure of offshore platform scenes and the overlap of associated oil spill targets, a deformable convolution DCNv2 module is introduced to improve the model’s recognition accuracy for multi-shaped targets. Additionally, the CBAM attention module is incorporated to enhance the model’s ability to learn target features. The EIOU loss function is used to improve the model’s convergence speed and its ability to identify small or partially occluded targets. Based on the positional relationship between the spilled crude oil target and the source target detection anchor frame, the possible source targets are classified into three association levels and assigned different association weights. By comprehensively applying parameters such as source-target association weight, leakage risk probability, and target overlap area, a calculation method of the association degree between the source target and the leaked crude oil target is proposed. Using the method mentioned above, the traceability analysis of crude oil leakage can be carried out based on the association degree size. The efficacy of the optimized model for multi-target detection in offshore platforms is substantiated through ablation studies and comparative analyses with analogous models. The effectiveness of the proposed crude oil spill traceability method for offshore platforms is verified by detecting and analyzing the experimental scenarios of offshore platforms.