Research on nameplate image recognition algorithm based on R-CNN and SSD deep learning detection methods
Ning Baifeng, Huang Ganzi, Yang Yu · 2022 IEEE 2nd International Conference on Power, Electronics and Computer Applications (ICPECA) · 2022
In order to prevent sudden equipment failures and accidents, it is necessary to conduct regular and irregular regular inspections on important substations and lines. Nameplate is an important power facility, which is usually installed in a prominent position of tower. Due to its long-term exposure to wind and rain, it is prone to damage, scratches and other defects. In this paper, Faster-RCNN based on region suggestion is adopted as the detection algorithm, and the nameplate in the image is identified and located by this algorithm. By selecting the improved LeNet-5 proposed in this paper, the classical Convolutional Neural Network (CNN) and ResNetl01 as the basis to extract the network, and comparing the models, the detection results under different framework models are discussed. optimize SSD algorithm to improve the speed and accuracy of image recognition, deeply study the parameter adjustment strategy of deep learning (DL), and build a rich nameplate database for training. The experimental results show that the improved fast SSD algorithm has better performance.