Conical scanning radiometric imaging system is good at large field view but suffers from visual nonlinear distortion. The distortion is caused by azimuth and elevation sampling in sphere coordinate, especially for short range and large views. An outdoor experiment is carried out on a building, and the raw image is obtained with obvious distortion. The key to correct distortion is solving the range in relationship between sphere coordinate and Cartesian coordinate. For a specific building, it is approximately treated as a plane object, and its height is assumed known to solve the range and parameters for plane fitting. Once the coordinates of all pixels are determined, the object is represented in Cartesian coordinate, and the nonlinear distortion is corrected. If any size information for object is unknown, an arbitrary plane is also competent for distortion correction. The difference is that the correcting result is a projection onto this plane instead of real location. However, the projection is also compatible with human vision.