摘要
依托不断发展的电子信息技术,对和车辆信息密切相关的驾驶证进行OC光学字符识别(OCR),成为实现智能交通的重要步骤。实际拍摄图像中,由于拍摄环境差别明显,光照不均匀和倾斜等情况普遍存在,识别时需要根据图像进行差异化处理,处理过程中很容易丢失文字结构信息,导致字符识别错误,降低系统识别率。本文针对驾驶证识别,采用"图片预处理+提取+识别"的整体思路,提出一种基于深度学习的机动车驾驶证检测与识别算法。首先,进行图片预处理,针对驾驶证图片存在的背景复杂、角度倾斜的问题,提出对边缘轮廓的位置进行定位以完成倾斜校正;其次,使用CTPN算法对图片中文字区域进行检测;最后,用RCNN算法对检测出的文字行进行文字检测,得到文字识别结果。
Relying on the continuous development of electronic information technology,OCR(Optical Character Recognition)for driving licenses closely related to vehicle information has become an important step to achieve intelligent transportation.In the actual captured images,due to obvious differences in shooting environment,uneven illumination and tilting,etc.,differentiation needs to be performed according to the image during recognition.It is easy to lose text structure information during processing,resulting in incorrect character recognition and reduced system recognition rate.The main research content of this article is for the recognition of driver's license,and the overall idea adopted is the combination of picture preprocessing,extraction and recognition.This paper proposes a deep learning-based vehicle driving license detection and recognition algorithm.The algorithm first preprocesses the picture.In view of the problem of complex background and angle tilt of the driver's license picture,it is proposed to position the edge contour of the driving license to complete the tilt correction.Then use CTPN algorithm to perform text detection on the text area in the picture on the corrected image,and finally the RCNN algorithm is used to perform character recognition on the detected character lines to obtain the character recognition result.
作者
姚砺
王昭丽
YAO Li;WANG Zhaoli(College of Computer Science and Technology,Donghua University,Shanghai 200000,China)
出处
《智能计算机与应用》
2020年第7期40-43,48,共5页
Intelligent Computer and Applications
