The secondary device of power plants and substations bears important functions. The information on the version of the device should be compared during the commissioning and acceptance. The existing practice process is complex, the labor cost is high and the efficiency is low. Given this phenomenon, this paper inventions a method of obtaining image information of the secondary equipment version based on ORC technology, studies the image preprocessing of various functions such as Resize, and EAST algorithm convolutional neural network to extract the input images, uses the convolution loop neural network to end-to-end character recognition, and finally realizes text detection and character recognition. The experiment shows that PaddleOCR processing based on the mkl mathematics library test platform has a fast response time and high accuracy, which is suitable for scenes with low-resolution field device screen. Based on this scheme, the output of the key information of the protection device is realized.