中文摘要：

提出了一种结合车牌区域边缘特征和梯度方向特征的新型车牌定位算法,该算法将车牌图像变换为灰度图像后,首先利用Sobel算子和Gradienffaces方法分别获取车牌图像的垂直边缘图像和梯度方向图像,然后根据边缘长度、边缘点密度去除垂直边缘图像中的干扰,并根据梯度方向图像中的梯度方向消除更多的干扰边缘,最后利用一个矩形窗扫描边缘图像完成车牌区域的定位和分割.实验结果表明,该方法的定位准确度可达93.7％,同时对复杂背景或弱光环境下的车牌图像具有很好的鲁棒性.

英文摘要：

A novel algorithm for license plate location is proposed, which employs edge and gradient direction features of plate region. After the conversion into gray image for the input image, the proposed method first uses the Sobel operator and the Gradientfaces algorithm to obtain the vertical edges image and the gradient direction image respectively. Then it eliminates the background and noise edges in accordance with the length and the density of the edges and removes mere inference edges according to the gradient directions of the gradient direction image from the vertical edges image. Finally, scan the edge image by a rectangle window to locate and segment the real license plate region. The experiment results indicate that the proposed method is reliable and accurate and its positioning accuracy is 93.7%. Especially, the method is robust to the images with complex backgrounds or poor illumination.