基于神经网络的视频序列车牌定位算法

在研究静态汽车牌照定位算法的基础上,利用时间序列多帧图像信息量大的优势弥补以往单幅图像牌照定位的不足,研究了视频序列下的汽车牌照定位方法.针对静态图像的牌照定位问题,提出了一种新的基于字符边缘特征和神经网络的汽车牌照定位算法.将上述定位算法应用到视频单帧图像中的牌照定位,并依此为基础,在动态视频序列中,对连续抽取的帧图像进行定位检测,各得到若干候选牌照区,比较相邻帧之间对应候选区的匹配程度,以得到真正的牌照位置.实验结果表明,该方法的定位精度高,实用性强.

Vehicle license plate location algorithm in video sequence based on neural network

Based on the studies of static license plate location,image-sequence-based plate location methods is reasearched which takes the advantage of the information in image sequences to overcome the drawbacks of those static-image-based location methods.A new algorithm based on neural network and image edge features is proposed.On the basis of this method,in the dynamic video sequence,by comparing the resemblance of the plates in adjacent frames,precise plate locations can be determined.Computer simulations show that this method has a very high location accuracy,and can be readily used in practical plate recognition systems.