【摘要】 针对某些道路交通视频场景下，获得用于检测消失点的特征较少时出现的检测消失点的误差和精度问题，提出结合基于平行坐标系的级联霍夫变换和DBSCAN聚类算法分别计算道路交通视频图像两个正交消失点。通过固定样本个数阈值并初始设置较大邻域距离阈值?,根据设定的最大聚类簇包含点与总数据点数的比例自适应降低阈值?,无须手工调整DBSCAN算法的两个预设参数且能够解决DBSCAN聚类算法难以处理消失点样本集密度不均匀的问题。实验结果表明，在视频车辆较少的情况下本文方法能够有效地对道路交通视频场景下的相机进行自动标定。更多还原

【Abstract】 Due to the insufficient accuracy of the vanishing point detection in some road traffic scenes, this paper proposes an novel method to calculate two orthogonal vanishing points in road traffic images by combining cascaded Hough transform based on parallel coordinate system and DBSCAN clustering algorithm. By fixing the threshold value of the number of samples, and properly setting a larger neighbourhood distance threshold ?, the two pre-set parameters of the DBSCAN algorithm can be adjusted automatically. Moreover, the problem of uneven density of vanishing point sample set in the DBSCAN clustering algorithm can be therefore solved. The experimental results show that this method can effectively calibrate the camera in the video scene of road traffic when there are few vehicles.更多还原