Abstract:Traditional target detection algorithms are affected by the complex background of substations and cannot accurately recognize and locate the electrical equipment.Therefore,this paper proposes an infrared image recognition model for the electrical equipment based on an improved CenterNet algorithm.Firstly,to address the problem of insufficient target feature information in complex environments,the ResNeXt50 with stronger feature extraction ability is used as the backbone network of CenterNet.While keeping the original model parameters unchanged,the network width is increased to obtain more abundant feature information,thereby the detection accuracy is compared.Secondly,the channel attention mechanism is added to the prediction layer to enhance the model’s attention to important targets while suppressing irrelevant information interference,ensuring the robustness of model detection.Finally,to prove the effectiveness of the model,verification experiments were conducted on a self-made dataset.The results show that the mean average accuracy of the improved model reached 93.7%,an increase of 2.1% compared to the original model,which is better than several existing mainstream detection models.This method effectively improves the recognition accuracy of electrical equipment infrared images in complex substation environments.