Aiming at the problems of low accuracy, long calculation time and few training samples in the traditional image detection method of UAV inspection of transmission line, an improved YOLOv3 model for insulator defect identification of transmission line components is proposed. K-means++ algorithm is introduced to solve the problem of small target insensitivity, Focalloss function is introduced to solve the problem of sample imbalance, Mish activation function is introduced to improve the accuracy of the model, and Senet attention mechanism is introduced to improve the performance of feature extraction. Through the comparative analysis of the performance of the model before and after the improvement, the superiority of this method is verified. The results show that, compared with the traditional detection methods, the proposed method can meet the needs of real-time detection in terms of detection speed, and the detection accuracy is optimal. The detection time is 0.079 s, and the average detection accuracy is 94.40%. This research can meet the needs of automatic detection of image defects in transmission line UAV inspection.