Fast Simultaneous Localization and Mapping Algorithm with Point and Line Feature Based on Image Entropy

Abstract

To address the problem of feature information redundancy caused by visual simultaneous localization and mapping algorithm with point and line features in high-texture environment, a fast simultaneous localization and mapping algorithm with point and line feature based on image entropy is proposed. In this paper, we first propose a new feature extraction strategy, which determines the parameters of the feature extractor by image entropy; then, the idea of weighting is introduced in pose estimation, and the point and line features are weighted by the image entropy; finally, we test our method using the KITTI and EuRoC dataset, and demonstrate that our method improves the real-time performance of the system while ensuring the accuracy and robustness of the system.