HLE-SLAM: SLAM for overexposed construction environment

Robustness to overexposure conditions poses a significant challenge in the visual simultaneous localization and mapping (V-SLAM) of construction robots. Limitations in sensor exposure lead to the loss of information, making it difficult to track enough feature points for construction robots under overexposure conditions, owing to the assumption of constant brightness in SLAM. To recover the missing information in overexposed frames, we propose a High and Low Exposure SLAM (HLE-SLAM) system that uses frame exposure fusion to generate globally well-exposed frames. As indicators, we adopt exposure, contrast and information entropy to select the best part of brightness and information in high and low exposure frames. The Shi-Tomasi and Kanade-Lucas-Tomasi (KLT) sparse optical flow algorithms are used to improve the ability to detect and track feature points in the overexposed environment. Our experimental results on data sets and real environments demonstrate that HLE-SLAM can effectively address the overexposure problem.