Multi-object detection method based on YOLOv9 for labor protection equipment wearing condition of offshore platform operators

Based on the work process monitoring requirements of offshore platform operators, the multi-object detection method for labor protection equipment wearing condition based on YOLOv9 model is proposed, the optimization method for YOLOv9 model based on CBAM attention mechanism and Focaler-iou bounding box loss function is studied, and the superiority of the optimized model is confirmed by a series of ablation experiments. Aiming at the missed detection problem of small targets of labor protection equipment, SAHI algorithm is introduced to slice the input image into multiple pieces, and the very small labor protection targets has been detected by the slicing detection method. A positive and negative samples combined detection method is designed to effectively reduce the false detection or missed detection rate of each labor protection equipment. The pictures with real labor protection samples in offshore platform operating scenes are used to train and test the models, and the research results have been successfully applied in the offshore platform, and the ideal detection results have been achieved.