Abstract
This paper proposes a visible light and active light fusion obstacle fingerprinting technique for the obstacle recognition challenge under the complex light environment on the lunar surface. The method obtains the lunar surface terrain data through visible light images and combines the active light field of view compensation information of the lunar rover to complete the path recognition of the moon. Obstacle features are extracted using Otsu algorithm, morphological processing, and a multispectral fingerprinting model is introduced to achieve high-precision obstacle classification. The experiments are conducted in a simulated lunar surface scene, and the results show that the method achieves 95.7% obstacle recognition accuracy under low-light conditions, and the misclassification rate is reduced to 4.3%, which significantly improves the perception reliability of the unmanned rover in extreme environments. This study provides a new solution that can be referred to for autonomous obstacle avoidance systems for deep space exploration.