Abstract:In order to solve the problem that intelligent mobile devices are difficult to be applied to the off-line environment without network in the image recognition of inspection and operation and maintenance of power transformer and other equipment, this paper designs an off-line image recognition system of power transformer based on neural rod, considering the actual needs and real situation of inspection and operation and maintenance of power equipment. First, the convolution neural network algorithm is introduced to build the deep learning image recognition model, and the real image data set is used to train and optimize the model; then, the model is innovatively integrated into the neural computing accelerator, and the off-line image recognition system of power transformer is built to further solve the constraints of the network on the system; finally, the power transformer based on the neural rod is separated from the system The line image recognition system is compared with the existing mobile device recognition system in 4G network environment and the mobile device recognition system in offline environment on the real picture data set provided by Chengbei Branch of Guiyang Bureau of Guizhou power grid. The results show that the off-line image recognition system of power transformer based on neural rod proposed in this paper has high recognition accuracy and stability.