Image-based transmission tower component identification for operation maintenance using deep learning

Operation maintenance work is an important part for the structural health assessment of transmission tower. The routine management and maintenance work mainly relies on engineers and technicians with practical experience to carry out visual inspection and fill in the questionnaire. However, human based visual inspection is an arduous and time-consuming task, and its detection results largely depend on the subjective judgment of human inspectors, as the same time the workers working at height are very dangerous. Aiming at the deficiency of artificial vision detection method, a detection method of transmission tower component recognition based on image recognition is proposed. UAV is used to detect the transmission tower in an all-round way, Thousands of images are used to train, verify and test the convolutional neural network (CNN) classifier based on Alexanet. Aiming at the problem of damage identification of transmission tower components, fast R-cnn based on improved ZF network is trained, verified and tested by using images. The performance of the trained CNNs and Faster R-CNN is tested on some new images which are not used for training and validation processes. The test results show that the method can accurately identify the components and damage of transmission tower.