Aiming at the problems of low recognition accuracy and poor recognition effect of traditional meter reading methods, this paper presents a numerical recognition method of intelligent meters based on Region-based Convolutional Neural Network (R-CNN). This method first collects the image information of the digital code disc of the ammeter by camera, then preprocesses the collected images, and finally automatically recognizes the preprocessed images through R-CNN. The experimental results show that the proposed method in this paper could identify the position information automatically and read information of the dial simultaneously using R-CNN, which will greatly reduce the pressure of manual meter recognition and improve the accuracy of meter recognition.