遥感影像分类结果的空间抽样精度检验方法研究

发布时间：2018-01-18 02:30

本文关键词：遥感影像分类结果的空间抽样精度检验方法研究　出处：《上海海洋大学》2016年硕士论文　论文类型：学位论文

【摘要】：高分四号的成功发射(2016年),实现了遥感卫星高稳定一体化结构,观测技术的发展使遥感影像的分辨率越来越高,遥感影像分类结果作为数据产品,其数据量和更新速率极速增加,故此,对遥感影像分类结果精度的快速检验成为研究热点之一。抽样方法作为一种简单、高效的调查手段已经被运用到遥感影像分类结果的精度检验中,而传统抽样方式下的遥感影像分类结果的精度检验方法受样本量及空间布样的影响,存在效率低和信息冗余等问题。如何针对遥感影像信息量大、多时相、多光谱等特点,研究适合于遥感影像分类结果的快速检验方法是遥感数据投入使用的重要保障。本文旨在研究一种高效的遥感影像分类结果精度检验方法,将空间抽样理论应用于遥感影像分类结果的精度检验中,首先,将遥感影像像元进行空间相关性分析,通过基于Moran’s I的空间布样模型对影像中遥感数据像元之间的距离与相关性之间的关系进行量化,并根据量化结果分析遥感数据的聚类程度,了解遥感影像中各土地类型空间分布情况;其次,根据基于Moran’s I的空间布样算法进行计算分析确定空间布样方案,解决空间布样中遥感影像数据样本点分布的难题;通过将基于Moran’s I的空间布样方法与基于误差矩阵的精度检验方法相结合,对遥感影像分类结果进行精度检验计算,解决遥感影像分类结果精度检验结果准确性低的难题;最后利用某区域真实遥感数据验证该方法的合理性以及其可行性。本文的主要内容如下:(1)介绍了遥感影像分类结果的空间抽样精度检验方法研究背景和意义分析了遥感影像分类结果精度检验的必要性;综述了遥感影像分类结果精度检验的国内外研究现状;归纳了空间抽样理论应用于遥感影像分类结果精度检验的优劣性,以及精度检验方案研究中存在的难点和不足之处。(2)解释了遥感影像分类精度检验中的理论基础及背景知识通过分析遥感影像数据的特征,引出了在遥感影像分类结果精度检验中利用空间布样方法的必要性,并对本文所涉及的空间抽样理论、精度检验方法相关背景知识和理论基础进行了解释和定义。(3)设计了基于moran’si的遥感影像分类结果精度检验模型基于moran’si实现遥感数据像元之间距离与相关性的定量化分析,确定用于遥感影像分类的抽样精度检验的最佳样本量和优化布样方式;通过误差矩阵等,对精度检验方法的效率进行评价。(4)实证分析利用上海和武汉某地区遥感影像进行实验,通过传统抽样方法与设计方法的比较,验证了基于moran’si空间布样的遥感影像分类结果精度检验方法的可行性和合理性。实验结果表明:基于moran’si空间布样的遥感影像分类精度检验结果与全精度检验结果相近,该方法不仅保证遥感影像分类结果精度检验的准确性,而且有效地减少样本冗余,提高检验的时效性。
[Abstract]:High four (2016), the successful launch of the satellite remote sensing high stable integration structure, development of observation technology makes more and more high resolution remote sensing image, remote sensing image classification results as data products, the amount of data and update rate of speed increase, therefore, rapid test on the accuracy of the classification results of remote sensing images has become a research hotspot sampling methods. As a simple and efficient means of investigation, precision test has been applied to remote sensing image classification results, and the method for testing the precision of remote sensing image classification results of traditional sampling methods under the influence of sample size and spatial pattern, which has low efficiency and information redundancy and other issues. According to a large amount of information of remote sensing images, multitemporal, multispectral characteristics, fast detection method for remote sensing image classification results is an important guarantee for remote sensing data. The purpose of this paper is put into use In precision test method research an efficient remote sensing image classification results, the accuracy test of spatial sampling theory is applied to the remote sensing image classification results, firstly, the remote sensing image pixel spatial correlation analysis, the spatial pattern of I model based on Moran 's to image the relation between distance and correlation between pixels in remote sensing data to quantify the degree of clustering, and according to the remote sensing data of quantitative analysis results, understand the spatial distribution of various land types in the remote sensing image; secondly, according to the Moran' s I space pattern algorithm based on the calculation and analysis to determine the solution space pattern, solve the problem of the sample point distribution pattern in space remote sensing data by spatial sampling method; Moran s based on I and precision test error matrix method based on the combination of remote sensing image classification results to test the precision of calculation, solution No problem of remote sensing image classification accuracy test results of low accuracy; finally, verify the rationality of the method and the feasibility of using a regional real remote sensing data. The main contents of this paper are as follows: (1) introduced the classification results of remote sensing image inspection method of spatial sampling precision method research background and significance analyzes the necessity of checking the precision of the classification results remote sensing image; summarizes the classification results of remote sensing image precision testing research status at home and abroad; summarizes the pros and cons of spatial sampling theory is applied to remote sensing image classification accuracy test, and accuracy test scheme research of difficulties and deficiencies. (2) explained the characteristics of remote sensing image data through the theoretical basis and background knowledge the precision of remote sensing image classification test, leads to the necessity in the spatial sampling method using precision inspection in remote sensing image classification results Spatial sampling theory, and involved in this paper, the definition and interpretation of the method of precision inspection of related background knowledge and theoretical basis. (3) the design of the model of remote sensing image classification accuracy test results of Moran 'Si to realize the quantification of remote sensing data between the pixel distance and correlation analysis of Moran Si based on the best sample for sure the sampling precision testing and optimization of sampling mode of remote sensing image classification; through the error matrix, to evaluate the efficiency of the inspection method of accuracy. (4) the empirical analysis experiments using the remote sensing images of Shanghai and Wuhan area, by comparing the traditional sampling method and design method, verify the feasibility accuracy test method of remote sensing image classification results Moran' Si spatial pattern and rationality. Based on experimental results show that: Based on the results of the remote sensing image classification accuracy test of Moran Si 'spatial pattern and The results of the complete accuracy test are similar. This method not only guarantees the accuracy of the accuracy of the classification results of remote sensing images, but also effectively reduces sample redundancy and improves the timeliness of the inspection.

【学位授予单位】：上海海洋大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP751

【相似文献】