Substation peripheral potential safety hazard identification method and system based on target detection

The invention discloses a method for identifying potential safety hazards around a transformer substation based on target detection, and the method comprises the steps: constructing a foreign matter sample data set around the transformer substation, carrying out the image feature enhancement of the foreign matter sample data set through employing CLAHE, and obtaining a foreign matter intrusion data set; an attention mechanism module, an alpha-CIoU loss function and depth separable convolution are introduced into Backbone of a YOLOv5 network structure, and an improved YOLOv5 target detection model is constructed; inputting the foreign matter intrusion data set into the improved YOLOv5 target detection model for iterative training so as to optimize the improved YOLOv5 target detection model; and performing target detection on images or videos around the transformer substation by using the optimized and improved YOLOv5 target detection model, and outputting a potential safety hazard identification result. Accord