提出了一种从全景影像中还原目标三维信息的测量方案. 通过全景相机所获取的影像只有GPS 坐标，而没有姿态信息，需要运用影像匹配技术以及光束法准确算出拍摄时刻全景影像的姿态，然后通过前方交会计算全景影像中目标的三维坐标. 针对全景影像畸变较大的特点，采用仿射不变特征匹配算法进行影像匹配，同时使用随机抽样一致算法剔除粗差点，以保证匹配点的数量及准确度. 根据全景影像的透视投影几何模型，能改进常规摄影测量中的光束法，可用来求解全景影像姿态. 提出一种针对全景摄影测量的前方交会算法，将空间直线方程变换后建立法方程，并进行平差解算. 实验表明，该方法相比于传统方法有更高的精度，可准确测算全景影像上的物点坐标.

This paper presents a measurement solution that recovers 3D information of
objects from a panoramic image. In case of having GPS position only, the panoramic
posture at the shooting time can be calculated by image match and bundle adjustment.
Thus we can use intersection to obtain the objects’ coordinates. As large distortion in
panoramic is an obstacle in image match, an affine invariant feature matching algorithm
is used for image match, and a random sample consensus algorithm used to eliminate
coarse errors, giving better point match in terms of both quantity and quality. For the unique geometric feature of panoramic image, the conventional photogrammetric bundle
adjustment is improved and used to solve posture of panorama. Finally, an intersection
method is proposed to build normal equations through transformation of space straight
line equations before adjustment. Compared with the conventional method, the proposed
algorithm has higher accuracy. Experimental results show feasibility of the scheme. It can
accurately measures object coordinates in panoramic images.