Abstract: In order to deal with the low efficiency issue of traditional manual insulator defect detection,a method for insulator defect detection based on Faster-RCNN is proposed. Firstly,the data of insulator defect pictures taken by aerial photography is enhanced,and residual network structure is used in the algorithm and attention mechanism is introduced,to improve the detection effect and reduce the complexity of the model at the same time. Then,group normalization is used to replace batch normalization. Finally,Soft-NMS is used instead of NMS for result optimization. Experimental results show that the accuracy of the improved algorithm reaches 90.3%,which is 14.7% higher than that before the improvement,which improves the effectiveness and reliability of insulator defect detection.