Due to the complex background of insulators in transmission lines

traditional fault identification algorithms have problems such as false detection

missing detection and low recognition rate. Compared with the traditional convolution neural network

capsule network uses vector as input for the first time

which can well retain the direction

angle and other characteristic information of the target

and is more suitable for identifying insulator in complex background. Therefore

this paper proposes an insulator damage identification algorithm based on the combination of capsule network and Yolo text location. The convolution kernel of traditional capsule network convolution layer and main capsule layer is simplified to 3×3 convolution kernel

and the weight is optimized by genetic algorithm and random gradient descent method

which shortens the training time and keeps the angle and direction of insulator. Therefore

the faulty insulator can be identified in complex environment; at the same time

the YOLO’s text location algorithm is used to correct the size of the damaged part of the insulator to get a more accurate location of the broken insulator. Finally

compared with Alex Net

YOLOV2

Fast R-CNN recognition algorithms

the insulator recognition rate of the proposed method is increased to 95%

so that the insulator can be identified more quickly and accurately from the complex background

and the speed and accuracy of identifying insulator and accurately locating the damaged position of insulator under complex background are improved.