基于视频流的车牌识别研究及系统设计

该文深入研究从视频流中提取运动车辆进行车牌识别的问题,提出了一种车牌识别算法,根据该算法开发了基于MFC的视频流车牌识别可视化系统。算法采用三帧差分与背景消减相结合的方法提取含有运动车辆的关键帧,对关键帧进行灰度化,采用Sobel算子进行边缘检测,融合形态学处理对关键帧进行去噪,从而实现车牌的定位,用投影法结合字符特征对车牌字符进行精确分割,采用BP神经网络算法提取字符特征,输出字符识别结果,由此完成车牌字符识别。通过对50组不同的车辆视频进行检测识别,识别率(包含汉字)达到96%以上。

Research on vehicle license plate recognition and system design based on video stream

In this paper,the problem of license plate recognition by extracting vehicle license plate from the video stream is studied,and a license plate recognition algorithm based on which a visualization system of vehicle license plate recognition based on MFC is developed is presented. For this algorithm, three frames difference and background subtraction are applied to extract key frames containing moving vehicles. Then gray - scale of the key frames,edge detection using Sobel Operator,and fusion of mor-phological processing of key frames for denoising are carried out for location of license plate. Next,the characters of the license plate are segmented by the projection method combining with and the features of the characters and the BP neural network algorithm is used to extract the features of the characters and then character recognition results is output. Thereby,license plate characters are recognized. By recog-nizing 50 sets of different vehicle video,it was shown that recognition rate (including Chinese charac-ters)is up to more than 96% .