MarsRecon：用于火星的自监督与多模态表面表征
MarsRecon: Self-Supervised and Multimodal Surface Representations for Mars
- University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
机构由 AI 辅助整理，请以论文原文为准。
中文总结 AI 辅助
MarsRecon提出地理感知流程，利用HiRISE影像训练掩码自编码器，并融合文本坐标上下文，实现火星表面多模态表征与检索，降低重建损失并取得显著检索性能。
中文摘要 AI 辅助
高分辨率轨道影像提供了丰富的火星表面记录，但稀疏的地质标签限制了监督表征学习。我们提出了MarsRecon，一个地理空间感知的流程，用于从奥林帕斯山的HiRISE观测中学习视觉和多模态表征。该流程校准NASA行星数据系统产品，提取有效的地理参考图像块，并在未标记的影像上训练掩码自编码器。增加输入分辨率和过滤无效令牌将主要Stage A模型系列中的留出重建损失从0.1751降低到0.1342。然后我们冻结视觉编码器，并将其特征与观测文本、坐标和局部-全局图像上下文对齐。当前最强的局部主导模型在留出测试集上实现了图像到文本的recall@10为0.3787，文本到图像的recall@10为0.9161，以及局部到全局的recall@10为0.4350。这些结果建立了一个可行的火星特定预训练和检索流程；需要进一步的裁剪重叠控制和下游地质评估来评估其嵌入的更广泛实用性。
英文摘要
High-resolution orbital imagery offers a rich record of the Martian surface, but sparse geological labels limit supervised representation learning. We present MarsRecon, a geospatially aware pipeline for learning visual and multimodal representations from HiRISE observations of Olympus Mons. The pipeline calibrates NASA Planetary Data System products, extracts valid georeferenced patches, and trains a masked autoencoder on unlabeled imagery. Increasing input resolution and filtering invalid tokens reduced held-out reconstruction loss from 0.1751 to 0.1342 in the principal Stage A model series. We then freeze the visual encoder and align its features with observation text, coordinates, and local--global image context. The strongest current local-primary model achieves image-to-text recall@10 of 0.3787, text-to-image recall@10 of 0.9161, and local-to-global recall@10 of 0.4350 on the held-out test split. These results establish a working Mars-specific pretraining and retrieval pipeline; further crop-overlap controls and downstream geological evaluations are needed to assess the broader utility of its embeddings.
