利用多源遥感数据的月表环形构造提取及其地层年龄估算

Under the condition of lacking lunar regolith samples, statistical methods of crater size-frequency distribution are used widely to estimate the geological age of lunar surface blocks. Due to the incomplete crater statistics data (lacking degraded craters which are difficult to be detected and identified), the estimated retention ages of those geological units are unreliable and have no cross-validation methods. Thus, the extraction and renewal of degraded craters for optimizing crater statistical data, and introduction of new statistical evidence for the cross-validation of the crater size-frequency based geologic age, become research focus. Multivariate data, including hyperspectral images, CCD images, DEM and the other data, are integrated in this proposal to solve the extraction of degraded circle structure (especially the craters) and the geological age estimation of lunar surface blocks. There are three main parts as follows: the automatic extraction and identification of degraded craters using deep learning; the key group degradation statistical factors mining based on the massive crater statistical spatial morphological data; improving the crater-size-frequency-distribution based geologic age estimation and providing the cross-validation of the geologic unit age under the condition of using crater group degradation statistical data. The study of this project will promote the development of planetary geological age theory and will be helpful to bring some significant science achievements after applying this method into those future Chinese deep space missions, such as the first landing on back of the moon, and the first landing on mars.

在缺少月球样品的情况下，基于撞击坑统计信息的大小-频率定年方法广泛用于月表地层年龄的确定，然而由于部分撞击坑退化严重难以提取和识别，使得撞击坑统计数据不完整，导致地层年龄确定存在不确定性，且定年结果目前缺乏有效验证的方法，因此如何提取并复原退化撞击坑以完善撞击坑统计数据，如何引入新的统计规律验证定年结果，是当前国内外月表地质定年的研究重点。本课题将融合高光谱、高分辨率CCD影像等数据，对退化环形构造（侧重于撞击坑）准确提取与月表地层定年问题进行研究，主要研究内容包括：基于深度学习的退化撞击坑的自动提取与识别；融合大规模撞击坑空间形态统计数据的群体退化关键统计因素挖掘；顾及撞击坑群体退化影响的大小-频率统计定年结果精化与交叉验证。项目的研究有助于推动地外星体估算地层地质年龄理论的发展，尤其是有助于未来我国嫦娥四号首次月球背面着陆和首次火星着陆任务在行星地层定年方面产出具有重大意义的科学成果。

在缺少月球样品的情况下，基于撞击坑统计信息的大小-频率定年方法广泛用于月表地层年龄的确定，然而由于部分撞击坑退化严重难以提取和识别，使得撞击坑统计数据不完整，导致地层年龄确定存在不确定性，且定年结果目前缺乏有效验证的方法，因此如何提取并复原退化撞击坑以完善撞击坑统计数据，如何引入新的统计规律验证定年结果，是当前国内外月表地质定年的研究重点。本项目针对退化撞击坑提取的关键性多源数据融合，提出了以撞击坑为基础的光学影像、光谱数据精确配准方法，实现了不同分辨率影像数据与光谱数据的精确地理标定；针对退化撞击坑提取，提出了基于Centernet*的退化撞击坑提取方法，实现了退化撞击坑的稳健识别与提取，并利用铁元素含量区分了主撞击坑与次生撞击坑；以嫦娥二号DEM为基础实现了撞击坑的多参数自动提取，并就其群体统计特征进行了聚类统计分析；开展了虹湾地区地质演化历史、哥白尼-H掩埋撞击坑确定、嫦娥四号着陆区石块溅射来源、嫦娥五号着陆区岩浆来源等拓展研究。项目研究结果拓展了嫦娥系列影像的应用领域，有助于提高地外星体撞击坑定年法的合理性与准确性可为我国探月任务的月球表面地质单元划分与遥感定年、区域地质演化、着陆区选址等提供有力支撑。