We propose a new methodology to assess whether the shape of a measured object fits a known surface. The method is based on calculating uncertainty spheres for a specific confidence level obtained by kernel modelling through a weighted resampling of the point cloud. If these uncertainty spheres do not contain the theoretical surface the two surfaces are considered to be statistically different. The methodology was used to assess terrestrial laser scanning as a suitable technology for fast and precise geometric optimization of parabolic trough collectors when in operation. A parabolic trough collector was measured using time-of-flight and phase-shift terrestrial laser scanners and results were compared with those obtained using photogrammetry. The impact of point density and the choice of surface (front or rear) on geometric optimization quality were analysed, with the results indicating that terrestrial laser scanning based on data collected from the front surface of the collector is not suitable for geometric optimization of parabolic trough collectors. However, the precision achieved for high-resolution scanning of the rear surface of the collector is similar to that yielded by photogrammetry, with the advantage that data acquisition time is considerably faster.