Review of visual SLAM technology in complex illumination environments

With the rapid development of autonomous robots, autonomous vehicles, and other intelligent devices, the demand for Simultaneous Localization and Mapping (SLAM) technology that can operate stably in complex lighting environments such as low light, significant lighting changes, smoke interference, and extreme weather is becoming increasingly urgent. The traditional SLAM scheme based on visible light cameras can operate stably under good lighting conditions, but it is prone to accuracy degradation and insufficient robustness in complex illumination environments, which limits its wide application in practical scenarios. Therefore, this article conducts investigations on visible light, infrared, and event cameras. Firstly, the principle of visual SLAM technology is systematically explained, and the classic algorithms of each module of visual SLAM are discussed. The advantages and disadvantages of relevant algorithms are analyzed and compared. A comparative analysis was conducted on the main research methods and application effects of visual SLAM technology in complex environments in recent years. The processing methods of vision sensors in complex illumination environments were analyzed from the perspectives of image enhancement and multi-sensor fusion. The fusion scheme of infrared inertial SLAM and multimodal visual SLAM was analyzed in detail, among which the fusion scheme of infrared cameras and LiDAR has great potential. Finally, multi-sensor datasets containing complex illumination environments were introduced, and the future development trends and challenges of visual SLAM technology were discussed from three aspects: multi-sensor collaboration, algorithm hardware integration, and semantic information integration.