Based on the improved YOLOv3 under the Catenary Insulator Image Recognition

摘要

In railway related supporting facilities, the catenary insulator is most among them very important one annulus, put forward different methods to the pretreatment of data sets, expands the capacity of data, loss of function optimization, deep learning of YOLOv3 insulator detection algorithm was improved, after the sampling by spatial attention mechanism and channel attention mechanism combining mechanism of cascading double attention to fusion filtering characteristics, improve the ability of feature extraction, the introduction of gaussian function to the maximum inhibition method was improved, reducing the presence of keep out target miss rate, improve the accuracy of insulator detecting, The identification of OCS insulators under complex background is completed by shortening the detection time of insulators, and the non-maximum suppression NMS is introduced. The experiment proves that the improved network performs well in the identification accuracy and identification time, and the accuracy is improved by 0.0216 on the basis of the original one. In addition, a Web interface is added to realize the online identification of insulators and meet the real-time requirements.