基于RANSAC的道路影像控制点选取方法研究

Research on selection method of road stereo image control points based on RANSAC

道路立体影像控制点的提取是影像匹配、图像拼接及利用道路立体影像进行车辆定位的前提。由于道路立体影像中地物平面层次多、背景复杂、光线影响较大等因素,基于SIFT算法直接获取的特征点均匀性较差,针对该问题提出了基于RANSAC和均匀度理论的控制点筛选方法,保证了控制点选取的准确性和均匀性。在野外、乡村和城市3个道路场景下的实验结果表明,该方法能够从道路立体影像中获取稳定、准确、均匀分布的控制点,为后续的立体影像匹配、控制点运动轨迹的估算及车辆自主定位导航的研究奠定基础。

The extraction of control points from road stereo image is the premise of image matching,image stitching and vehicle positioning using road stereo image.Because of the factors such as many levels of ground objects in road stereo images,complex background and large influence of light,the uniformity of feature points obtained directly based on SIFT algorithm are poor.To solve this problem,a control point selection method based on RANSAC and uniformity theory is proposed,which ensures the accuracy and uniformity of control point selection.The experimental results of field,rural and urban road scenarios show that the method can obtain stable,accurate and evenly distributed control points from road stereo images,which lays a foundation for subsequent research on stereo image matching,control point motion track estimation and vehicle autonomous positioning and navigation.