地球信息科学学报 ›› 2020, Vol. 22 ›› Issue (4): 857-866.doi: 10.12082/dqxxkx.2020.190747

• • 上一篇    下一篇

面向地理国情监测的变化检测与地表覆盖信息更新方法

杜培军1,2,3,*(), 王欣1,2,3, 蒙亚平1,2,3, 林聪1,2,3, 张鹏1,2,3, 卢刚2,4   

  1. 1. 南京大学地理与海洋科学学院,南京 210023
    2. 南京大学自然资源部国土卫星遥感应用重点实验室,南京 210023
    3. 江苏省地理信息资源开发与利用协同创新中心,南京 210023
    4. 江苏省测绘工程院,南京 210013
  • 收稿日期:2019-12-04 修回日期:2020-01-03 出版日期:2020-04-25 发布日期:2020-06-10
  • 通讯作者: 杜培军 E-mail:peijun@nju.edu.cn
  • 作者简介:杜培军(1975— ),男,山西五台人,教授,博士,主要研究方向为城市环境遥感。
  • 基金资助:
    国家自然科学基金重点项目(41631176)

Effective Change Detection Approaches for Geographic National Condition Monitoring and Land Cover Map Updating

DU Peijun1,2,3,*(), WANG Xin1,2,3, MENG Yaping1,2,3, LIN Cong1,2,3, ZHANG Peng1,2,3, LU Gang2,4   

  1. 1. School of Geography and Ocean Science, Nanjing University, Nanjing 210023,China
    2. Key Laboratory for Land Satellite Remote Sensing Applications of Ministry of Natural Resources, Nanjing University, Nanjing 210023, China
    3. Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing 210023, China
    4. Jiangsu Provincial Surveying and Mapping Engineering Institute, Nanjing 210013, China
  • Received:2019-12-04 Revised:2020-01-03 Online:2020-04-25 Published:2020-06-10
  • Contact: DU Peijun E-mail:peijun@nju.edu.cn
  • Supported by:
    National Natural Science Foundation of China(41631176)

摘要:

常态化地理国情监测能够全面、动态地掌握地理国情信息及其变化,为经济建设和社会发展提供数据基础。地理国情普查成果是按照统一规范标准、经过内业解译和外业核查形成的矢量数据。如何在普查成果的基础上,利用多时相遥感影像实现变化信息提取与更新是地理国情监测的关键。针对地理国情普查成果的特点与监测需求,以多时相遥感影像处理分析为基础,构建了针对地理国情监测的变化检测方法体系,提出了像元—对象结合的多时相影像变化检测、基于对象实体统计分析的变化识别方法,实现了综合地理国情普查成果和遥感影像的地理国情变化检测与数据更新。基于像元—对象结合的多时相影像变化检测首先根据传统的变化矢量分析法提取基于像元的变化检测结果,再以地理国情普查的矢量对象为统计单元计算对象内变化像元的比例,以此判断该矢量对象是否发生了变化,并根据变化像元的比例计算其变化强度。基于对象实体统计分析的变化检测方法直接以地理国情矢量为对象进行特征提取和差异构造,再将差异影像进行阈值分割得到基于地理国情对象的变化检测图。最后,根据变化检测结果,对变化区域进行面向对象分割,并从上一期未变化区域选取训练样本训练分类器模型以得到变化区域的地表覆盖类型,将变化区域与未变化区域结合得到更新后的地理国情矢量图。选取江阴市地理国情普查成果和两期高分辨率遥感影像进行试验,结果表明本文提出的方法在准确提取和解释变化区域的同时,明显提高了变化检测和数据更新的效率,可用于常态化地理国情监测。

关键词: 地理国情监测, 高分辨率遥感, 变化检测, 面向对象图像分析, 地表覆盖, 地理信息更新, 变化矢量分析, 支持向量机

Abstract:

Geographic national condition monitoring can comprehensively and dynamically grasp the changes of national information. They can also provide the data for economic and social development. Geographic national condition census generates the vector data according to unified standards, visual interpretation and field verification. The extraction of change information and updating of land cover maps based on geographic national condition census and multi-temporal remote sensing images is the key to geographic national condition monitoring. According to the characteristics of the geographic national condition census and monitoring demand, a change detection framework for geographic national condition monitoring based on multi-temporal remote sensing images is constructed. Multi-temporal image change detection methods and statistical analysis of object entities are proposed. The proposed method realizes the change detection and updating of geographic national condition census with the combination of the previous census outcomes and bi-temporal remote sensing images. The change detection method based on pixel-object combination first extracts the pixel-based change according to the traditional change vector analysis. Taking the vectors of the geographic national census as the statistical unit, this method then calculates the proportion of the changing pixel within the objects to determine whether they have changed and their change intensities. While the change detection method based on statistical analysis of object entities directly considers the geographic national census vectors as the objects for feature extraction and difference image construction. The resulting difference image is then segmented by an automated threshold to achieve the geographic national condition object-based change detection map. According to the change detection results, the pixels in changed areas are segmented by object-based segmentation, and the training samples are selected from the unchanged areas in the previous temporal image and census map to train the classifier model. Finally, the updated geographical condition vectors are achieved based on the combination of the original unchanged pixels and supervised classification results of the changed areas. The geographic national condition census of Jiangyin County and two high-resolution remote sensing images are used in the experiments. The results demonstrate the effectiveness of the proposed method with accurate results and low cost for change detection and geographic national condition information updating, which provides potential means for geographic national condition monitoring.

Key words: geographic national condition monitoring, high-resolution remote sensing, change detection, object-based image analysis, land cover, geographic information updating, Change Vector Analysis (CVA), Support Vector Machine (SVM)