雷达故障表格处理系统

本课题来源于南京某研究所合作项目“雷达故障传递模型建立及验证系统”,由于研究所内资料具有很强的保密性,为避免网络传输造成的信息泄露,研究所提供的雷达故障资料大部分为纸质版表格形式。针对纸质版表格人工录入数据的工作量大、效率低问题,本文从表格图像处理、字符识别和表格还原三个方面进行研究,设计了一套完整的纸质版表格OCR系统。在表格图像处理部分,通过Shi-Tomasi角点检测算法和有向单连通链算法来提取表格图像四个顶点,并通过透视变换法实现图像矫正;采用K-means聚类算法和模板匹配法对传统的投影法进行改进,提高了字符分割的正确率。在字符识别部分,本文对比了VGG16和ResNet这两种经典的卷积神经网络模型,最终选取ResNet模型完成字符识别任务。实验表明,该系统具有识别率高、适应性强、速度快等优点,能够很好地将雷达故障表格转换为电子表格。

T his subject originates from a cooperation project of"The radar fault transfer model establishment and verification system"in a Nanjing research institute.Due to the strong confidentiality of the data in the institute,in order to avoid information leakage caused by network transmission,most of the radar fault data provided by the institute in paper form.Aiming at the problem of large workload and low efficiency of manual data entry in paper form,this paper studies from three aspects of form image processing,character recognition and form restoration,and designs a complete paper form OCR system.In the form image processing part,the Shi-Tomasi corner detection algorithm and directed single connected chain algorithm are used to extract the four vertices of the form image,and the perspective correction method is used to correct the image.The K-means clustering algorithm and template matching method are used to improve the accuracy of character segmentation.In the character recognition part,this paper compared two classic convolutional neural network models,VGG16 and ResNet,and finally selected the ResNet model to complete the character recognition task.Experiments show that the system has the advantages of high recognition rate,strong adaptability,and fast speed.It can effectively convert the radar fault form into an electronic form.