基于混合熵模型的遥感分类不确定性的多尺度评价方法研究

不确定性是影响遥感图像分类质量的最主要因素,针对在遥感图像分类过程中同时存在随机不确定性和模糊不确定性的特点,提出基于混合熵模型来综合测度这两种不确定性的方法,并建立起多尺度的评价指标.在分析混合熵模型基本原理的基础之上,提出利用特征空间的和模糊分类器的统计数据来建立信息熵、模糊熵以及混合熵的方法.同时,在像元和类别尺度上,分别建立像元混合熵和类别混合熵的指标对分类不确定性进行评价.最后,应用湖北省黄石市的遥感影像对上述评价方法进行验证分析,实验结果表明,混合熵模型能有效地反映分类过程中随机不确定性和模糊不确定性的综合影响,并从不同尺度反映出遥感影像分类的质量问题.     

遥感云分类不确定性的多维混合熵模型评价

针对遥感影像提取信息分类过程中存在随机不确定性和模糊不确定性两种噪声,影响分类结果的准确性问题,该文在多光谱遥感影像处理中,通过对传统的混合熵模型进行多维化改进,提出多维混合熵的不确定性评价模型。采用云算法对遥感影像进行解译分类,获取相应的不确定性模型参数计算出信息熵和模糊熵,从像元和类别两个尺度构建出遥感云分类不确定性的多维混合熵评价模型。结果表明,多维混合熵模型能够充分考虑多光谱遥感数据的多维性,可以从不同尺度对遥感分类的随机不确定性和模糊不确定性进行有效全面地评价。     

模糊对象粗糙表达及其空间关系研究

空间数据的模糊性是当前GIS领域研究的难点之一。现有的GIS数据模型基本上还不具备对模糊性进行描述、处理和分析的功能 ,使得GIS对数据的处理容易造成信息损失。把粗糙集和模糊集对空间数据的模糊性和不确定性处理的优势结合起来表达模糊对象 ,即利用粗糙集处理由于知识的粒度性所造成的模糊性和不精确性 ;利用模糊集处理模糊对象原始数据所具有的不确定性 ,从而扩展了空间数据模型对模糊数据的表达能力 ,并详细研究了模糊面与模糊面以及模糊线与模糊面之间的拓扑关系     

GIS线元的平均熵不确定带

在GIS线元的位置不确定性方面，国内外学者已提出了“e-带”、“e-带”、“g-带”、“H-带”等模型，然而就应用而言，由于“e-带”具有不变带宽，因而应用最为广泛。但是“e-带”的宽度往往难以确定，从而限制了它的使用范围。在“H-带”的基础上，提出了根据线元的平均信息熵确定“e-带”宽度的思想，建立了线元的平均熵不确定带，并以此作为线元位置不确定性的度量。     

熵理论在确定点位不确定性指标上的应用

分析了传统点位不确定性指标的局限性，基于信息论中的联合熵和最大熵定理导出了n维随机点熵不确定指标以及落入其内概率的统一公式；提出了以熵误差椭圆与熵误差椭球作为2维、3维GIS中点元的位置不确定性度量指标。提出的熵指标具有唯一确定、不受置信水平选取的主观性影响等特点，适合于度量未知分布的点位不确定性。     

熵理论在确定点位不确定性指标上的应用

分析了传统点位不确定性指标的局限性,基于信息论中的联合熵和最大熵定理导出了n维随机点熵不确定指标以及落入其内概率的统一公式;提出了以熵误差椭圆与熵误差椭球作为2维、3维GIS中点元的位置不确定性度量指标.提出的熵指标具有唯一确定、不受置信水平选取的主观性影响等特点,适合于度量未知分布的点位不确定性.     

2维线元不确定性εσ模型误差带几何特征的代数研究

对地理信息系统（GIS）中2维线元不确定性εσ模型的误差带的几何特征进行研究。运用函数单调性和极值理论,从理论上证明εσ模型的误差带的几何特征;得到线元误差带不仅具有“两端大、中间小”,而且也可能具有“一端大、一端小”的几何形状;给出了误差带的最小带宽及其位置;论证了误差带的最小误差带宽的位置靠近中误差较小的一端。本研究完善了2维线元不确定性εσ模型,使其更具有严密性,同时为GIS空间数据不确定性的研究提供了新的方法。     

GIS中三维空间直线的误差熵模型

从信息熵的角度提出了三维空间直线的误差熵模型，该模型由以垂直直线的平面误差熵为半径的圆柱体和两端点的误差球组成，是一种完全确定的度量空间线元不确定性的模型。理论分析与实验表明，本文所提出的模型具有较好的效果。     

GIS中线元的误差熵带研究

基于现有的线元位置不确定性模型大多与置信水平的选取有关,而置信水平的选取带有一定程度的主观性,因而不能惟一确定,引入信息熵理论,提出了线元的误差熵带模型,并将它与“E－带”进行了比较,计算了落入其内的概率。该模型根据联合熵惟一确定,与置信水平的选取无关。     

GIS空间数据几何变换的不确定性传播

针对点位误差、线元在对称、旋转变换过程中不确定性的传播规律,该文采用误差椭圆和εm模型描述GIS空间数据对象模型中最常用的点、线元素的误差域,讨论几何变换(如对称、旋转变换)过程中不确定性的积累和传播规律,结合区间算法给出了变换后误差域模型。基于区间运算的INTLAB模拟结果验证了该模型的有效性和实用性。研究结果对进一步探讨其他几何变换过程中不确定性的积累传播规律具有一定参考。     

General entropy model of line segments uncertainty in GIS

Spatial data uncertainty can directly affect the quality of digital products and GIS-based decision making. On the basis of the characteristics of randomicity of positional data and fuzziness of attribute data, taking entropy as a measure, the stochastic entropy model of positional data uncertainty and fuzzy entropy model of attribute data uncertainty are proposed. As both randomicity and fuzziness usually simultaneously exist in linear segments, their omnibus effects are also investigated and quantified. A novel uncertainty measure, general entropy, is presented. The general entropy can be used as a uniform measure to quantify the total uncertainty caused by stochastic uncertainty and fuzzy uncertainty in GIS.     

General Entropy Model of Line Segments Uncerta,nty in GIS

Spatial data uncertainty can directly affect the quality of digital products and GIS-based decision making. On the basis of the characteristics of randomicity of positional data and fuzziness of attribute data, taking entropy as a measure, the stochastic entropy model of positional data uncertainty and fuzzy entropy model of attribute data uncertainty are proposed. As both randomicity and fuzziness usually simultaneously exist in linear segments, their omnibus effects are also investigated and quantified. A novel uncertainty measure, general entropy, is presented. The general entropy can be used as a uniform measure to quantify the total uncertainty caused by stochastic uncertainty and fuzzy uncertainty in GIS.     

General Entropy Model of Line Segments Uncertainty in GIS

Spatial data uncertainty can directly affect the quality of digital products and GIS-based decision making. On the basis of the characteristics of randomicity of positional data and fuzziness of attribute data, taking entropy as a measure, the stochastic entropy model of positional data uncertainty and fuzzy entropy model of attribute data uncertainty are proposed. As both randomic-ity and fuzziness usually simultaneously exist in linear segments, their omnibus effects are also investigated and quantified. A novel uncertainty measure, general entropy, is presented. The general entropy can be used as a uniform measure to quantify the total un-certainty caused by stochastic uncertainty and fuzzy uncertainty in GIS.     

Entropy error model of planar geometry features in GIS

Positional error of line segments is usually described by using “g-band”, however, its band width is in relation to the confidence level choice. In fact, given different confidence levels, a series of concentric bands can be obtained. To overcome the effect of confidence level on the error indicator, by introducing the union entropy theory, we propose an entropy error ellipse index of point, then extend it to line segment and polygon, and establish an entropy error band of line segment and an entropy error donut of polygon. The research shows that the entropy error index can be determined uniquely and is not influenced by confidence level, and that they are suitable for positional uncertainty of planar geometry features.     

Entropy error model of planar geometry features in GIS

Positional error of line segments is usually described byusing “g-band”,however,its band width is in relation to the confidence level choice.In fact,given different confidence levels,a series of concentric bands can be obtained.To overcome the effect of confidence level on the error indicator,by introducing the union entropy theory,we propose an entropy error ellipse index of point,then extend it to line segment and polygon.and establish an entropy error band of line segment and an entropy error do-nut of polygon.The research shows that the entropy error index can be determined uniquely and is not influenced by confidence level,and that they are suitable for positional uncertainty of planar geometry features.     

考虑参数误差的3维空间直线不确定性模型

由于线元上任一点坐标的误差不仅受端点误差的影响,还会受到长度误差的影响,故不确定性模型要考虑各种影响位置精度的参数误差,对3维空间直线不确定性模型作了进一步研究。不但考虑了端点误差的影响,还顾及了长度误差的影响,使模型在理论上更为严密。理论和实验研究表明,长度误差影响了直线方向的精度。     

考虑参数误差的3维空间直线不确定性模型

由于线元上任一点坐标的误差不仅受端点误差的影响,还会受到长度误差的影响,故不确定性模型要考虑各种影响位置精度的参数误差,对3维空间直线不确定性模型作了进一步研究.不但考虑了端点误差的影响,还顾及了长度误差的影响,使模型在理论上更为严密.理论和实验研究表明,长度误差影响了直线方向的精度.