一种基于SIFT特征提取的车牌定位方法

针对传统车牌定位算法对车牌图像质量要求较高、鲁棒性较差、准确率较低等不足，提出了一种基于尺度不变特征（SIFT）特征提取的车牌定位新方法．它利用车牌中汉字字符的局部特征属性，以SIFT特征提取方法进行抽取，并用之构建特征模板库，然后把待识别车牌图像的SIFT特征与之相匹配，用RANSAC算法剔除误匹配点后，便得到仿射变换矩阵，从而实现对车牌较准确的初定位和初倾斜校正．进一步对提取的车牌区域图像二值化，用Radon变换求得倾斜角度后，可生成精确的仿射变换矩阵，并实现对车牌的精确定位和倾斜校正．实验表明：与传统算法相比，本方法不仅能够实现准确的车牌定位及倾斜校正，而且对图像亮度、污损、倾斜、尺寸变化等具有良好的适应性和鲁棒性．

An approach to vehicle license plate location based on SIFT

In order to overcome the drawbacks such as high demands for image quality, poor robustness and low accuracy in traditional vehicle license plate location algorithms, this paper presents a new license plate location method based on scale invariant feature transform （SIFT） algorithm. By using the local features of the Chinese characters in the license plate, the approach extracts SIFT features of the license plate in the sample template database, matches the SIFT features of the license plate image to be identified with them, and then eliminates the error matching points with RANSAC algorithm and calculates the affine transformation matrix of the template, thus the approach achieves accurate initial license plate location and initial tilt correction, then transforms the extracted license plate region to binary image, obtains the incline angle with the Radon transform to calculate the accurate affine transformation matrix, finally achieves the precise vehicle license plate location and tilt correction. Experiment shows that compared with traditional algorithms, the proposed method not only achieves accurate license plate location and tilt correction, but also has good adaptability to the image brightness, defilement, tilt, scale change and so on, so it has a good performance of robustness.