Defect detection algorithm of transformer substation based on improved YOLOv11s

Abstract For the problems of insufficient feature extraction capability and unsatisfactory detection accuracy of small targets in the field of substation defect detection, an improved YOLOv11s model is proposed and applied to the background of substation defect detection. The model is based on YOLOv11s, replaces the C3K2 module in the backbone network with the DTBlock (Dynamic Transformer) module, uses the C2PSA module as the downsampling attention feature extraction part, and introduces ACLIP at the input end to strengthen the ability of the model to extract feature information; In the neck network, the MSF (Multi Scale Fusion) feature pyramid network (FPN) is introduced to improve the detection accuracy of the model for small targets. Experimental results show that for the improved model, accuracy increased 4.6% and the average accuracy increased 5%, greatly improving the model’s characteristic information extraction ability, strengthening the accuracy of the model for small target detection, and meeting the demand for complex scenarios and more intelligent substation inspection work.