Efficient lunar lander detection using a lightweight saliency map for small planetary rovers

Planetary exploration demands effective image processing techniques for detecting anomalous objects and points of interest. However, technical constraints, such as limited spaceborne CPU performance and the growing need for compact, cost-effective robots, pose significant challenges. To address these issues, we propose a lightweight saliency map-based object detection method that enables efficient lunar lander detection by low-cost robots on the lunar surface. Our approach streamlines saliency maps by reducing the number of feature maps, and layers of an image pyramid that applies Gaussian filtering only to the first layer, resulting in about 97% reducing computational cost compared to conventional methods, while maintaining robustness against noise. By leveraging the distinctive edge features of the surface of lunar landers, our method achieves an accuracy of 95%. The proposed lightweight saliency map method holds potential for enhancing the performance of small, low-cost robots in planetary exploration, paving the way for more efficient and economical space missions.