Performance evaluation of several high-quality digital cameras

The accuracy of object reconstruction was tested for six high quality digital cameras. Nikon D80, D200, D2X and D3 with the same 24 mm Nikon lens, a Canon EOS 5D with a normal as well as a stabilized 35 mm Canon lens and an Alpa 12 WA with special fixation of a digital camera back and lens were evaluated. All cameras were investigated on a test cube that was designed according to VDI/VDE 2634, a German standard for evaluation of optical 3-D measuring systems. The measuring volume was approximately 2000 mm x 2000 mm x 1600 mm (length x width x height). Seven scale bars with up to ten calibrated distances on them were placed in the cube to evaluate the length measurement error (LME). Images were imported into a PC and analyzed with AICON 3D Studio software. Calibration was performed with a standard set of parameters as well as with an image variant set of parameters to account for geometric instabilities and to reveal the potential accuracy of the cameras. The smallest LME for calibration with standard parameters was achieved with 0.047 mm using the Canon EOS 5D and a stabilized 35 mm lens. For calibration with image variant parameters the Alpa 12 WA performed best with a maximum absolute LME of 0.029 mm. Geometric accuracy was significantly reduced in cases where a ringflash was screwed into the filter thread of the lens. Only the Canon EOS 5D with stabilized lens was unaffected by the extra load of the ringflash. Subsequent evaluation of the Nikon D2X and D3 with ringflash mounted to the cameras’ tripod mount revealed similar LME of 0.052 and 0.059 mm, respectively, as for the Canon camera when calibrating with the standard set of parameters. The other Nikon cameras and the Alpa camera were only evaluated with the ringflash mounted to the lens.