Text Detection of Power Equipment Nameplates Based on Deep Learning

Recognition of the nameplate text in power equipment by image recognition technology can effectively improve the equipment management in power system. Text detection is the first step of text recognition, which also affects significantly the accuracy of text recognition. In this paper, an Efficient and Accuracy Scene Text Detector (EAST) based text detection model is proposed for detecting the characters on power equipment nameplates. Data augmentation technology is used to expand the data set of nameplate images. The experiment results show that the proposed model achieves a detection precision of 95.48% and a recall of 97.08%. Compared with the traditional text detection method, the EAST-based model performs better on the nameplate images of power equipment and has stronger generalization ability. Moreover, the nameplate text detection model runs at 7 frames per second (FPS) on average, which is faster than the traditional methods.