基于对象的高分辨率遥感影像土地利用变化检测技术研究

发布时间：2018-07-25 12:59

【摘要】：随着现代社会经济的高速发展,土地资源的开发速度越来越快,土地利用变化呈现变化周期小、范围广、更新慢的特点。获取数据速度快、覆盖面积广、人工工作量要求小的遥感技术是变化检测的最佳手段。而高分辨率遥感影像一方面提高了人们提取遥感影像信息的精度；另一方面对遥感影像处理技术的要求也更高。优化高分辨率遥感影像的处理技术对土地利用变化检测具有重要意义。本文采用面向对象技术进行土地利用变化检测,针对面向对象影像处理过程中的如何选择最佳分割尺度和如何应用对象特征实现分类的两大问题进行了具体的研究。首先,本文提出了一种最佳分割尺度的评价指数和计算模型,并用实验证明了评价指数和计算模型的有效性。评价指数包括对象内部的同质性指数和对象之间的异质性指数。同质性指数是利用对象的标准差、面积和分割对象总数三个要素进行计算；异质性指数是利用对象边界上的邻域总变分进行计算。最佳分割尺度计算模型是运用数理统计法对实验数据分析得到的非线性回归模型。其次,本文针对影像分割后的分类问题,设计了土地类型特征—规则库,对土地类型的主要对象特征和分类规则进行有效存储和管理,从而为土地分类提供动态的分类规则链接。分类规则将模糊逻辑分类方法和决策树分类方法结合起来,能为地类制定了层次分明的模糊规则。特征优化选择是依据信息论方法统计的特征集信息量和相关性来选择包含信息量最大的特征集。最后,本文综合利用以上方法进行了土地利用变化检测实验,针对典型的土地类型,进行了最佳分割尺度计算、特征优化组合和分类规则制定,并建立了特征—规则库实例。文章将不同时期的影像数据按最佳分割尺度分割、分类规则分类、分类结果对比分析的流程进行操作,最终得到土地利用类型的变化信息。实验结果证明本文方法的有效性和实用性。
[Abstract]:With the rapid development of modern society and economy, the development of land resources is becoming faster and faster. The change of land use is characterized by small period of change, wide range and slow renewal. Remote sensing technology with fast data acquisition, wide coverage and low manual workload is the best method for change detection. On the one hand, the high resolution remote sensing image improves the precision of extracting remote sensing image information; on the other hand, the request of remote sensing image processing technology is also higher. It is very important to optimize the processing technology of high resolution remote sensing image for land use change detection. In this paper, the object oriented technology is used to detect land use change. Two major problems, how to choose the best segmentation scale in the process of object oriented image processing and how to use object features to realize classification, are studied in detail. First of all, this paper presents an evaluation index and a computational model of the optimal segmentation scale, and the validity of the evaluation index and the computational model is proved by experiments. The evaluation index includes the homogeneity index within the object and the heterogeneity index between the objects. The homogeneity index is calculated by using the standard deviation of the object, the area and the total number of partitioned objects, while the heterogeneity index is calculated by using the total variation of the neighborhood on the boundary of the object. The optimal partition scale model is a nonlinear regression model based on mathematical statistics to analyze the experimental data. Secondly, aiming at the classification problem after image segmentation, this paper designs the land type feature-rule base, and stores and manages the main object features and classification rules of land type effectively. Thus provides the dynamic classification rule link for the land classification. By combining the fuzzy logic classification method with the decision tree classification method, the classification rules can make clear fuzzy rules for the ground classification. The optimal selection of features is based on the information quantity and correlation of the feature set counted by the information theory method to select the feature set that contains the largest amount of information. Finally, the paper makes comprehensive use of the above methods to carry on the land use change detection experiment, for the typical land type, carries on the best partition scale calculation, the feature optimized combination and the classification rule formulation, and has established the characteristic rule base example. In this paper, the image data of different periods are segmented according to the best segmentation scale, the classification rules are classified, and the classification results are compared and analyzed. Finally, the change information of land use type is obtained. The experimental results show that the proposed method is effective and practical.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F301;P237

【参考文献】