Visual Simultaneous Localization and Mapping (SLAM) technology is widely adapted to autonomous vehicles,drones and augmented reality devices. However,the memory footprint and computing cost grow quadratically with map size. To reduce the map size and computation cost while maintaining the pose localization accuracy,based on the basic graph optimization feature sparsification algorithm,an information entropy weighting approach is introduced. And an adaptive connection between the sparsification module and the main module of SLAM is proposed. The effect of the algorithm under different parameter configurations is evaluated to determine the best parameters. By extensive experimental evaluations we demonstrate the proposed method reduce the average total system processing time by 16.2％,and increase the average absolute positioning accuracy by 6％.
simultaneous localization and mapping (SLAM)feature sparsificationparameter adjustmentadaptiveinformation entropy
NETL
NSTL
万方数据
