天基大气背景测量处理系统总体设计与数据挖掘方法研究

发布时间：2018-01-26 03:21

本文关键词： 定量遥感定量化处理遥感图像数据挖掘灰度共生矩阵支持向量机　出处：《哈尔滨工业大学》2014年硕士论文　论文类型：学位论文

【摘要】：本文从我国某型号红外大气背景遥感卫星对数据定量化处理及后期应用的需求出发，设计、实现、应用了天基大气背景测量数据定量化处理系统；采用数据挖掘的思想与方法实现了大气背景测量数据的应用，在数据质量评价与筛选、图像纹理特征描述、数据挖掘分类技术及机器学习方面开展了深入的研究工作，实现、应用了具有自学习能力的大气背景测量图像分类系统。本文的主要工作及成果如下：在天基大气背景测量处理系统总体设计研究方面，针对遥感数据定量化处理的实际需要，对天基大气背景测量处理系统进行了总体设计，，按功能将系统进一步划分为相互独立的处理模块，并对各模块的实现功能、实现方法及基本处理模型进行了设计。在面向数据挖掘的数据预处理方法研究方面，以红外大气背景辐射亮度图像的误差分析为切入点，对数据的质量进行评价，对误差计算结果进行分析，研究了各误差分量对数据质量的影响，最后综合各种因素确定数据的筛选原则；选用图像的纹理特征对图像中云的分布情况进行描述，采用灰度共生矩阵的方法对图像的纹理特征进行提取描述，通过比较试验确定了灰度共生矩阵中的各项参数，获得了12个纹理特征参数，对纹理特征参数的描述结果进行分析，将其按描述能力初步划分为三类。在基于支持向量机的图像分类方法研究方面，使用非线性可分支持向量机分类器按图像中的云分布情况对数据进行分类，通过试验进一步修正了对纹理特征参数的分类，确定了用于学习和分类的纹理描述参数，实现了较高的分类精度；设计了支持向量机分类器的自学习算法，试验结果表明其分类能力优于传统分类器，克服了无效参数造成分类性能下降的问题。在定量化处理系统及图像分类系统的实现研究方面，完成了天基红外大气背景测量处理系统的建设，系统已投入使用，性能稳定，定量处理精度高；研发的辐射亮度图像分类系统软件，可完成数据筛选、纹理描述及图像分类的数据挖掘分类全过程处理任务。本文的研究工作，可为我国大气背景定量遥感测量后续型号的地面处理系统设计与建设提供理论方法与技术支持。
[Abstract]:This paper designs and implements a space-based atmospheric background measurement data quantification processing system based on the requirement of a certain type of infrared atmospheric background remote sensing satellite and its later application. The idea and method of data mining are used to realize the application of atmospheric background measurement data, in data quality evaluation and screening, image texture feature description. The classification technology of data mining and machine learning have been deeply studied and realized, and the image classification system of atmospheric background measurement with self-learning ability has been implemented. The main work and results of this paper are as follows: In the aspect of the overall design and research of space-based atmospheric background measurement and processing system, the overall design of space-based atmospheric background measurement and processing system is carried out according to the actual needs of quantitative processing of remote sensing data. According to the function, the system is further divided into independent processing modules, and the realization function, realization method and basic processing model of each module are designed. In the research of data preprocessing method for data mining, the error analysis of infrared atmospheric background radiance image is taken as the starting point, the quality of data is evaluated, and the error calculation results are analyzed. The influence of each error component on data quality is studied. Finally, the screening principle of data is determined by synthesizing all kinds of factors. The distribution of cloud in the image is described by the texture feature of the image, and the texture feature of the image is extracted and described by the method of gray level co-occurrence matrix. The parameters in the gray level co-occurrence matrix are determined by comparison experiments, and 12 texture feature parameters are obtained. The description results of the texture feature parameters are analyzed and divided into three categories according to the description ability. In the research of image classification based on support vector machine, the nonlinear separable support vector machine classifier is used to classify the data according to the cloud distribution in the image. Through experiments, the classification of texture feature parameters is further corrected, and the texture description parameters for learning and classification are determined, and the classification accuracy is achieved. The self-learning algorithm of SVM classifier is designed. The experimental results show that the classification ability of SVM classifier is better than that of traditional classifier. In the realization of quantitative processing system and image classification system, the construction of space-based infrared atmospheric background measurement and processing system has been completed, the system has been put into use, the performance of the system is stable, and the quantitative processing accuracy is high. The software of radiance image classification system can complete the whole process of data filtering, texture description and image classification. The research work in this paper can provide theoretical method and technical support for the design and construction of ground processing system for quantitative remote sensing measurement of atmospheric background in China.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP311.13;TP751

【参考文献】