Investigating SLAM Algorithms for Robustness in Illumination-Varying Scenarios

Indirect methods for Simultaneous Localization and Mapping (SLAM) are susceptible to illumination variations, which can compromise the algorithm's robustness. To enhance the performance of SLAM algorithms in low-light or illumination-varying environments, we propose an improved ORB-SLAM method tailored for such challenging conditions. The method employs a joint histogram that integrates pixel intensity values with the mean intensity of their neighborhoods, using a twodimensional cumulative distribution function (CDF) to map intensity values and effectively expand the dynamic range. Additionally, a dynamic gamma correction technique is introduced to further enhance image brightness and optimize detail representation. Experimental evaluations on public datasets demonstrate that the proposed method significantly improves the robustness of ORB-SLAM in low-light and illumination-varying scenarios.