A complete solution for effectively, automatically, and accurately reconstructing the three-dimensional (3-D) panoramic terrain from aerial imagery is presented. With enough premeasured and identified georeferences, we first estimate every camera’s accurate intrinsic and extrinsic parameters by implementing bundle adjustment, which is introduced in detail. Afterward, the adjacent relationship of imagery is acquired from the cameras’ position parameters. In addition, the formulas for corresponding area prediction and image rectification are derived according to the camera parameters. Subsequently, feature-based matching is conducted between adjacent image pairs to provide much more constraints for bundle adjustment. Area-based matching is applied to pairs of horizontal epipolar imagery for dense correspondence to produce dense spatial point cloud. Eventually, the mosaicked digital ortho map and digital elevation model of the whole imaging area are produced automatically by a series of steps including spatial intersection, Tin generation, differential correction, and color blending. Experimental results show that the root mean square (RMS) residual errors of check points in planimetry and altitude are, respectively, 0.039 and 0.170 m, demonstrating the high accuracy of camera orientation. The visualized panoramic 3-D realistic scene validates the feasibility and effectiveness of the proposed solution.