低成本光学字符识别读表系统研制

Develop of Low Cost Meter Reading System by Optical Character Recognition

摘要: 为了实现低成本的光学识别以及对识别数据的云传输和储存，基于Raspberry Pi建立了通过光学字符识别的读表系统。并对该系统所采用的光学识别、数据存储、图像处理等算法进行研究。首先从摄像头获取一张带有电表数字的图像，利用OpenCV对其进行归一化尺寸、灰度处理、高斯模糊、二值化处理、提取连通区域等图像预处理，最后使用Tesseract进行OCR识别得到光学字符数据并进行记录和储存。实验结果表明：在系统设计的光照环境下，图像预处理后的数字光学识别成功率达到100%，识别时间低于1秒。基本满足光学字符识别系统的准确率高、速度快、稳定性好等要求。

Abstract: In order to realize low cost optical character recognition and storage of recognition data, the optical character recognition meter reading system based on Raspberry Pi was built. At the same time, we studied the algorithms of optical recognition, data storage and image processing used in this system. First, an image with a meter digital is obtained from the camera. Second, using Open CV to perform image preprocessing such as normalized size, grayscale processing, Gaussian blur, binarization processing, extraction of connectivity areas. Finally, using Tesseract for OCR recognition to obtain, record and store optical character data. Under the light environment designed by the system, the experimental results show that the accuracy of rate of digital recognition can achieve 100% and recognition time is less than 1 second after preprocessing. The optical character recognition system basically meets the requirements with high accuracy, fast speed and good stability.