Abstract: Distribution line safety is one of the keys to the safe operation of the power grid. Nowadays, distribution line inspection is mainly aimed at the existing faults and potential risks of a single object, with insufficient consideration of multi-objective risk factors. To solve this problem, a method for identifying hidden dangers of illegal fishing under distribution lines using Bayesian optimization RT-DETR (Bayesian optimization-based real-time detection transformer, BO-RT-DETR) is proposed. Initially, the image data is obtained by UAV, and the features of the image are extracted based on HGNetv2. Subsequently, the high-dimensional features are extracted using the efficient hybrid encoder. An IoU perception query is then employed to enhance model classification accuracy and IoU value, ensuring effective detection of illegal fishing hazards. Then, a multi-objective hazard identification strategy is developed, integrating the Bayesian optimization algorithm into the hazard detection model to generate the hidden danger identification model with optimal parameters, thereby elevating model performance. The experimental results show that the F1 score of the proposed method is 97.5%, mPA50 is 0.972, and the false alarm rate is 4.7%, which is 0.8% and 0.01 higher than RT-DETR, respectively, and the false alarm rate is reduced by 1.5%.