Transformer oil leakage detection with sampling-WIoU module

In order to detect the oil leakage problem of transformers in time and avoid the abnormal operation of power system caused by transformer oil leakage, this paper proposed a transformer oil leakage detection method based on improved YOLOv5. Firstly, by adding 4 × sampling layer to the FPN structure, the feature information was fused across layers to improve the detection accuracy of the model. Secondly, the Wise-IoU (WIoU) bounding loss function with dynamic non-monotony focusing mechanism was introduced to accelerate the training and reasoning of the network, and the overall performance of the model was further improved by balancing the learning of low and high-quality samples. Finally, the model was improved from Anchor-based to Anchor-free, which greatly reduced the number of parameters and calculation amount of the model, and had better performance than the original model. Used the self-built outdoor transformer oil leakage data set for training and testing, the results showed that compared with the original model, the improved network precision was increased by 7.3%, the recall was increased by 5.0%, the mAP0.5 was increased by 6.4%, the number of parameters was decreased by 28.2%, and the reasoning speed increased by 51.7%, which was beneficial to engineering deployment.