Abstract: In the traffic surveillance system based on digital image processing, nonlinear distortion will be caused by the installation place of the camera. It also will lead to get the inaccurate parameter of traffic. An algorithm based on grid image can rectify this distortion by edge detection, Hough transformation, and least squares. It can get the real traffic informations from the 2D images. When the inner parameters change, this algorithm will regulate the revise coefficients automatically.

Key words: nonlinear distortion, edge detection, Hough transformation, camera self-calibration, least squares

摘要： 在基于数字图像处理技术的交通监测系统中，拍摄得到的图像存在非线性几何形变，影响正确的交通参数的获取和交通现象的判断。一种基于网格的图像非线性校正算法,通过图像二值化，边缘检测，Hough变换等自动识别图像中的交通网格线，进而根据特征点匹配寻找全局图像的对应关系，对图像的非线性形变能有较好的校正，能从二维图像信息得到实际道路上的交通信息。该算法还可根据摄像机自标定，在摄像机的内参数发生改变时自动地调整校正系数，在ITS视频监控中有较好地应用前景。

关键词: 非线性形变, 边缘检测, Hough变换, 摄像机自标定, 最小二乘法