Bundle adjustment is one of the essential components of the computer vision toolbox. This paper revisits the resection-intersection\r\napproach, which has previously been shown to have inferior convergence properties. Modifications are proposed that greatly\r\nimprove the performance of this method, resulting in a fast and accurate approach. Firstly, a linear triangulation step is added\r\nto the intersection stage, yielding higher accuracy and improved convergence rate. Secondly, the effect of parameter updates is\r\ntracked in order to reduce wasteful computation; only variables coupled to significantly changing variables are updated. This leads\r\nto significant improvements in computation time, at the cost of a small, controllable increase in error. Loop closures are handled\r\neffectively without the need for additional networkmodelling.The proposed approach is shown experimentally to yield comparable\r\naccuracy to a full sparse bundle adjustment (20% error increase) while computation time scales much better with the number of\r\nvariables. Experiments on a progressive reconstruction system show the proposed method to be more efficient by a factor of 65 to\r\n177, and 4.5 times more accurate (increasing over time) than a localised sparse bundle adjustment approach.
Loading....