装载有视频采集设备的巡检列车已被应用于铁路环境监测中。提出了一种从前向运动的列车车载视频中生成铁路环境立体全景图的方法，该方法仅使用一个摄像设备即可获取整个场景的立体全景信息。基于场景的几何先验和标定的相机参数从视频序列中连续地抽取一组梯形窄带作为完全采样拼接区域，窄带的宽度根据车速自动调节，将所有窄带经单应性变换修正后拼接得到全景图像，并据此构建准确的几何模型来生成全景立体像对。实验结果表明，该方法能够快速有效地生成具有立体感的全景图。与原始的视频格式相比，生成的全景图所需存储空间小，具有更广阔的视野和更强的真实感，可实现铁路场景的快速访问和沉浸式浏览。

The patrol train with picture pick-up device has been used for railway environment monitoring. A method to generate the stereo panorama from forward motion train-borne video using only one camera to acquire the panoramic stereo information of the entire scene is presented. On the basis of the geometric priori information of the scene and known camera parameters, a set of trapezoidal stripes is extracted from a video sequence continuously, whose width is automatically adjusted according to the vehicle speed, and the stripes are stitched into a panoramic image after aligning all the stripes by homography transformation. The accurate geometric model is built to generate the panoramic stereo pairs. The experimental results show that the method can effectively and efficiently generate a stereoscopic panorama. Compared with the original video format, the stereo panorama requires less storage space, possesses wider field of view and greater realism, and enables the inspector to perform quick examination and immersive browsing.