发表机构
IBM Research; The University of Alabama in Huntsville; University Space Research Association (USRA); NASA Goddard Space Flight Center; SETI Institute; NASA Ames Research Center; University of Maryland, Baltimore County (UMBC); Howard University（IBM研究院; 阿拉巴马大学亨茨维尔分校; 大学空间研究协会; 美国国家航空航天局戈达德太空飞行中心; 搜寻地外文明研究所; 美国国家航空航天局艾姆斯研究中心; 马里兰大学巴尔的摩县分校; 霍华德大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出一种月球遥感多模态基础模型，在SomBench上预训练，通过多尺度联合训练和几何上下文，在陨石坑检测、IMP分割和极地冰回归等任务上匹配或超越ImageNet基线，并展示LoRA的高效性。
AI 中文摘要
我们提出了一种用于月球遥感的多模态基础模型，该模型在SomBench上从头开始预训练。SomBench是一个按地理划分的数据集，包含近两百万个配准的瓦片束，覆盖11种模态，空间尺度为两种（1米/像素和100米/像素）。该模型改编了TerraMind掩码令牌架构，并增加了两个针对月球的扩展：采集几何信息作为显式上下文提供，米级和百米级瓦片联合训练，使得单一权重集覆盖两种分辨率。FlexiViT补丁嵌入允许在不重新训练的情况下适应不同的补丁大小，而按模态的输入支持灵活的多模态微调。定性生成实验表明，该模型学习了有意义的跨模态对应关系，包括从高程得到的地形衍生量，以及从几何得到的与光照一致的反射率。我们在四个基准上进行了评估：WAC和NAC尺度的陨石坑检测、不规则月海斑（IMP）分割以及极地冰前景回归。在所有任务中，预训练模型匹配或优于ImageNet预训练基线以及架构相同的随机初始化对照。在多模态冰前景回归中，预训练变体取得了最佳结果，而随机初始化模型优于大多数基线，这表明性能提升既来自架构也来自预训练。在WAC陨石坑检测中，标签效率显著：在50%数据上训练的预训练模型超过了在完整数据集上训练的最强ImageNet基线。在适应策略中，LoRA在陨石坑检测和IMP分割上匹配或超过全量微调，同时使用的可训练参数少得多，而全量微调在冰前景回归中表现最佳。我们发布了预训练检查点、基准数据集和微调代码，以支持可复现的月球AI研究。
英文摘要
We present a multimodal foundation model for lunar remote sensing, pretrained from scratch on SomBench, a geographically partitioned corpus of nearly two million co-registered tile bundles spanning 11 modalities at two spatial scales (1 m/pixel and 100 m/pixel). The model adapts the TerraMind masked-token architecture with two lunar-specific extensions: acquisition geometry is provided as explicit context, and meter- and hundred-meter-scale tiles are trained jointly so that a single set of weights covers both resolutions. FlexiViT patch embeddings allow adaptation to different patch sizes without retraining, while modality-wise inputs enable flexible multimodal fine-tuning. Qualitative generation experiments suggest the model learns meaningful cross-modal correspondences, including terrain derivatives from elevation and illumination-consistent reflectance from geometry. We evaluate on four benchmarks: crater detection at WAC and NAC scales, irregular mare patch (IMP) segmentation, and polar ice prospectivity regression. Across tasks, the pretrained model matches or outperforms ImageNet-pretrained baselines and an architecturally identical random-init control. On multimodal ice prospectivity regression, pretrained variants achieve the best results, while the random-init model outperforms most baselines, suggesting gains arise from both the architecture and pretraining. Label efficiency is notable for WAC crater detection, where the pretrained model trained on 50% of the data exceeds the strongest ImageNet baseline trained on the full dataset. Among adaptation strategies, LoRA matches or surpasses full fine-tuning on crater detection and IMP segmentation while using far fewer trainable parameters, whereas full fine-tuning performs best for ice prospectivity regression. We release the pretrained checkpoint, benchmark datasets, and fine-tuning code to support reproducible lunar AI research.
