Abstract
On December 8, 2018, China launched the Chang’E-4 lunar probe and implemented soft landing and patrol exploration on the lunar farside for the first time. Fast and high-precision positioning of the landing point is a critical step of the mission, and also an important prerequisite for surface operations of the lander and rover. Based on the high-precision image matching and geometric transformation methods, and considering the engineering requirements, the landing point location of the Chang’E-4 lander was initially determined using the near real-time high compression ratio descent sequence images. Then, the position was refined using the replayed low compression ratio descent images, and the landing point position was calculated to be (177.5884°E, 45.4565°S). The landing point localization method and result have been successfully applied to actual engineering tasks for the first time in China, thus providing important support to the topographic analysis of the landing area and mission planning of the follow-up teleoperations.
Introduction
Landing on the lunar farside presents unique challenges, primarily the lack of direct line-of-sight communication with Earth. This necessitates the use of the "Queqiao" relay satellite. During the critical "powered descent" phase, traditional radio-based positioning methods like Very Long Baseline Interferometry (VLBI) are ineffective. Furthermore, the limited bandwidth of the relay link requires heavy data compression (up to 1:64) for real-time transmission, which introduces significant "blocking artifacts" that hinder standard computer vision algorithms. Achieving fast, high-precision localization under these constraints is essential for the subsequent surface operations of the Yutu-2 rover.
Methodology
The researchers implemented a "coarse-to-fine" vision-based localization pipeline designed to handle varying data quality and time constraints.
Image Enhancement and Feature Matching
To mitigate the effects of high compression in real-time images, a specialized de-blocking algorithm was developed. This process uses adaptive smoothing and Sigma filters to preserve crater edges while ignoring the lander's structural components. The Scale-Invariant Feature Transform (SIFT) algorithm was then employed to match features across consecutive frames in the descent sequence, allowing the system to track the landing point relative to the lunar surface.
Two-Stage Localization Strategy
The mission utilized a dual-stage approach:
- Initial Coarse Localization: Within 20 minutes of landing, 1:64 compressed real-time images were processed to provide a preliminary location.
- Refinement Stage: Once the lander was safely on the surface, high-quality images (1:8 compression) were replayed from onboard memory. These images allowed for high-precision matching against reference maps to finalize the coordinates.
Reference Mapping
The descent images were mapped to high-resolution reference data, including a 7m/pixel Digital Orthophoto Map (DOM) from the Chang’E-2 mission and a custom 1m/pixel base map created from Lunar Reconnaissance Orbiter (LRO) Narrow Angle Camera imagery.
Results and Discussion
The Chang’E-4 landing point was precisely determined to be 177.5884°E, 45.4565°S within the Von Kármán crater. The initial localization was achieved rapidly, providing critical data for early mission planning. The refined localization achieved sub-pixel matching accuracy, with an estimated absolute positioning error of approximately 20.1 meters. This error margin is primarily attributed to the inherent orbit determination accuracy of the LRO satellite rather than the vision algorithm itself.
The rugged nature of the lunar farside, while challenging for landing, provided a wealth of distinct impact craters that served as excellent geometric anchors. The hybrid approach—combining automated sequence matching with human-in-the-loop verification for the final map anchoring—ensured the high reliability required for mission-critical tasks.
Conclusion
The vision-based localization method successfully supported the historic Chang’E-4 mission. While highly effective, the method's accuracy remains dependent on the availability of high-resolution satellite imagery like that from the LRO. Future work will focus on integrating more diverse remote sensing data sources and increasing the automation of the map-matching process to support autonomous landings on other planetary bodies.
Summary updated