地球信息科学学报 ›› 2016, Vol. 18 ›› Issue (3): 396-405.doi: 10.3724/SP.J.1047.2016.00396

• 地球信息科学理论与方法 • 上一篇    下一篇

基于行为聚类算法的土地利用聚类适宜性分析研究

孙云华1,2(), 郭涛3,*(), 崔希民1, 崔伟宏2   

  1. 1. 中国矿业大学(北京)地球科学与测绘工程学院,北京 100083
    2. 中国科学院遥感与数字地球研究所,北京 100101
    3. 湖南科技大学计算机科学与工程学院,湘潭 411201
  • 收稿日期:2015-05-30 修回日期:2015-07-07 出版日期:2016-03-10 发布日期:2016-03-10
  • 通讯作者: 郭涛 E-mail:yunhua07@163.com;guotao0628@outlook.com
  • 作者简介:

    作者简介:孙云华(1985-),女,山东临清人,博士生,主要从事土地利用变化研究.E-mail:yunhua07@163.com

  • 基金资助:
    国家自然科学基金项目(51474217,71150001)

Suitability Analysis on Behavior-based Aggregation of Land Use Classification in Yunnan Province

SUN Yunhua1,2(), GUO Tao3,*(), CUI Ximin1, CUI Weihong2   

  1. 1. College of Geoscience and Surveying Engineering, China University of Mining & Technology, Beijing 100083, China
    2. Institute of Remote Sensing and Digital Earth, CAS, Beijing 100101, China
    3. School of Computer Science and Technology, Hunan University of Science and Technology, Xiangtan 411201, China
  • Received:2015-05-30 Revised:2015-07-07 Online:2016-03-10 Published:2016-03-10
  • Contact: GUO Tao E-mail:yunhua07@163.com;guotao0628@outlook.com

摘要:

为了在较少地类的基础上,深入研究土地利用变化过程,可把较多的地类合并成较少的具有重要变化特征的地类.本文运用基于行为聚类的方法,将净增加和净增加的地类合并或净减少和净减少的地类合并,但不能将净增加和净减少的地类合并.主要分为3个阶段实现:第1个阶段是完全不变阶段,聚类结果是形成综合的完全不变地类;第2个阶段是完全减少和完全增加阶段,该阶段将完全减少地类和完全增加地类分别进行合并;第3个阶段是转变阶段,该阶段计算了每一对有相同净变化方向地类的相互转变信息.基于行为聚类的算法在每一步的计算过程中都会保留净变化面积为常数,转变变化面积保留最大化.本文对3个聚类原则进行了数学证明,定义了6种聚类行为,以云南省土地利用分类体系为例,从面积变化和强度变化2个角度阐述了行为聚类方法的可行性和优势性.最后,与刘纪远等建立的土地利用遥感监测分类系统进行比较,结果表明:行为聚类算法聚类到9种类型时,类型总变化面积达到5.10%,比原始类型的总变化面积下降了0.06%;而基于遥感监测分类系统的6个一级分类将总变化减少至4.7%,与聚类算法比较,总变化面积减少了0.4%.实践证明,行为聚类的算法可更好地保留土地利用动态变化信息,证明了该聚类方法的有效性.

关键词: 行为聚类, 土地利用, 分类系统, 净变化, 转变变化, 云南省

Abstract:

Comparison of the same study area between two time points on the same categorical variable can reveal changes occurred among categories over time, such as transitions among land categories. Therefore, it is an effective method to aggregate a large number of categories into a smaller number of broader categories in order to simplify interpretation and give insights concerning categorical change over time. We use an algorithm to aggregate categories in a sequence of steps based on the categories' behaviors in terms of gross losses and gross gains. The behavior-based algorithm aggregates net gaining categories with net gaining categories, and aggregates net losing categories with net losing categories, but doesn't aggregates a net gaining category with a net losing category. The algorithm's steps are summarized into three phases. The first phase is the Exclusive Zero phase, when our algorithm aggregates pairs of Exclusive Zero categories until all Exclusive Zero categories are aggregated into one comprehensive Exclusive Zeroes category, which remains as one of the three categories at the end of the algorithm. The second phase is the Exclusive Loser and Gainer phase, when the algorithm aggregates pairs of Exclusive Gainer categories and pairs of Exclusive Loser categories. The third phase is the swapping phase. In this final phase, our algorithm computes the transition sum for every possible pair of categories that have the same direction of net change. Moreover, the behavior-based algorithm at each step in the sequence keeps the values of net change and maximizes swap change. This article introduces three mathematical principles and defines six aggregation behaviors. In the last section of this article, we present a case study, in which the data are obtained in Yunnan Province between 1990 and 2010 for 20 land categories, in order to prove the advantage and feasibility of this algorithm in terms of area change and intensity change. The results show that the behavior-based algorithm produces a set of 9 categories that retains almost the original amount of change, while giving a total area change of 5.10%, which is only 0.06% lower than the original change. In contrast, the common used classification system of 6 categories produces a total area change of 4.7%, which is 0.4% lower than the results of our algorithm. It has been found that the behavior-based algorithm is an effective method to retain land use dynamic change information.

Key words: behavior-based aggregation, land use, classification, net change, swap change, Yunnan Province