Precision Relative Pose Estimation at Close Range Using SingleCamera and Laser Range Finder

摘要

As in close-range photogrammetry, space resection method is very frequently used in obtaining the exterior orientationdata of a single photograph by applying the collinearity equation, but it has relatively poor accuracy in determining theline-of-sight component, especially in the condition the camera has a small field of view (FOV). This paper proposes anew solution for relative pose estimation of a setup consisting of a single camera, a Laser Range Finder (LRF), threeoptical targets and a array of cube-corner reflectors. These optical targets are installed properly on the target object ascontrol points; the reflectors array is also fixed on the target object to make sure that sufficient signal energy could bereturned to the Laser Range Finder. Rotations of the target object produce some errors in the measurements of LRF, so ifwe want to get the precise data of the line-of-sight component, the two-dimensional position in the array plane must bedetermined where the signal is returned from. In this paper we develop a method to solve the problem coupling with thepose data obtained from photogrammetry. We derive a new collinearity equation form and improve the one photographspace resection algorithm for the target object pose estimation relative to the camera or LRF. Simulation experimentresults show that the algorithm can converge to a precision data during about 3 times least-squares iteration; the accuracyof position data achieve the level of sub-millimeter; and the RMS error of attitude data is less than 50 arcseconds.