基于激光测高数据的月表撞击坑自动检测方法  

Automatic Crater Detection with Laser Altimetric Data

作　　者：李卉[1] 钟成[2] 

机构地区：[1]中国地质大学行星科学研究所,湖北武汉430074 [2]中国地质大学教育部长江三峡库区地质灾害研究中心,湖北武汉430074

出　　处：《地球科学（中国地质大学学报）》2013年第S1期161-166,共6页Earth Science-Journal of China University of Geosciences

基　　金：国家重点基础研究发展计划(No.2011CB707001);国家自然科学基金(Nos.41102209;41102210);中国博士后科学基金(No.20110491232);湖北省自然科学基金(No.2010CDB04105);武汉市科技攻关计划(No.201110821237);武汉市青年科技晨光计划(No.201150431074);中央高校基本科研业务费专项资金(Nos.CUG100705;CUG100313)

摘　　要：撞击坑是月球表面最醒目和最重要的地质构造标志.基于遥感影像的撞击坑检测方法受到数据本身的局限,遗漏检测的情况比较严重.激光测高数据可以快速构建全球的数字高程分析(DEM),为撞击坑检测提供丰富的几何形态信息.利用激光测高数据,提出了兼顾准确性和完整性的撞击坑自动检测方法.首先采用松散耦合的形态约束检测候选撞击坑,再利用机器学习的相关理论来提取正确的撞击坑.其中,提出基于可变尺度的撞击坑检测算法,以有效解决叠置撞击坑的检测问题;提出基于决策树的机器学习方法,以解决从候选坑中有效提取正确撞击坑的问题.实验证明了整个流程和关键技术的可行性和有效性.通过研究,实现高效、准确的月表撞击坑检测,充分挖掘探月卫星提供激光测高数据,向后续月球地质地貌研究提供具有准确性和丰富形态信息的基础数据.Impact craters are the most evident geological signs and the most important geological tectonics on the lunar surface.The methods based on remote sensing image for detecting craters are constrained by the limitations of the data,resulting in serious omission.It is easily to reconstruct the global digital elevation model(DEM)with laser altimetry data,which provides sufficient spatial information for crater detection.An automatic method of crater detection based on laser altimetry data is presented in this paper,to improve both the accuracy and completeness.Firstly,a loose coupling shape based rule is employed to extract the candidate crater;then machine learning method is used to find the correct crater.Among them,a variable scale is utilized to effectively detect small craters and nest craters;decision tree based machine learning method is employed to distinguish the correct crater from candidates.Experiments prove the feasibility and effectiveness of the whole process and the key technologies emplored.This study ensures efficient and accurate craters detection and access to more accurate spatial data for further lunar geological exploration.

关 键 词：撞击坑 激光测高数据 机器学习 决策树 地质构造 

分 类 号：P184.5[天文地球—天文学]