Aiming at the low efficiency and potential safety problems of insulator detection in power industry in China, an improved insulator detection method based on convolutional neural network is proposed in this paper.The method to solve the detection caused by insulator size scale diversification inaccurate detection problem and caused by insulator number or Angle problem between shade so that the residual error detection problem, the Faster-RCNN anchor in the scale of the width to height ratio and made changes, to NMS post-processing algorithm was improved at the same time, the introduction of phased punishment factor, to adapt to the overlap in multi-scale, more proportion, insulator block such as complicated situation.The improved Faster R-CNN algorithm can accurately detect the insulator and output the coordinate box and the probability value at the same time.Experimental results show that the improved Faster R-CNN detection method improves the AP from 0.7977 to 0.9058, and significantly improve the situation of missed and false detection of the insulator.