一种基于局部判别正切空间排列的高光谱遥感影像降维方法

提出一种基于局部判别正切空间排列(local discriminative tangent space alignment,LDTSA)的高光谱影像降维方法。LDTSA源于局部正切空间排列(LTSA)中的排列机制,在一个局域块内利用线性局部正切平面对类内样本的流形结构建模,同时还考虑到类间判别信息以最大化判别边界。利用多幅高光谱数据进行降维和分类试验。结果表明,LDTSA主要有三个优点:①在小样本问题上性能稳定;②在降维过程中保持类别间的判别信息;③有效挖掘数据集的几何流形结构。     

Selection of band combination for IRS data

Selection of band combination for generating a colour composite image using IRS data is discussed from statistical considerations. Most suitable three band combination turns out to be bands 1, 3 and 4. It is also shown that intrinsic dimensionality of IRS data is approximately two.     

Review on graph learning for dimensionality reduction of hyperspectral image

ABSTRACT

Graph learning is an effective manner to analyze the intrinsic properties of data. It has been widely used in the fields of dimensionality reduction and classification for data. In this paper, we focus on the graph learning-based dimensionality reduction for a hyperspectral image. Firstly, we review the development of graph learning and its application in a hyperspectral image. Then, we mainly discuss several representative graph methods including two manifold learning methods, two sparse graph learning methods, and two hypergraph learning methods. For manifold learning, we analyze neighborhood preserving embedding and locality preserving projections which are two classic manifold learning methods and can be transformed into the form of a graph. For sparse graph, we introduce sparsity preserving graph embedding and sparse graph-based discriminant analysis which can adaptively reveal data structure to construct a graph. For hypergraph learning, we review binary hypergraph and discriminant hyper-Laplacian projection which can represent the high-order relationship of data.     

空-谱协同正则化稀疏超图嵌入的高光谱图像分类

传统依据图嵌入的高光谱图像维数约简算法多数仅利用光谱信息表征像元间单一关系,忽视了数据间的多元几何结构。本文提出了一种面向高光谱图像分类的空-谱协同正则化稀疏超图嵌入算法(SSRSHE)。该算法首先利用稀疏表示揭示像元之间的相关性,自适应选择近邻,并构建稀疏本征超图和惩罚超图,以有效表征像元间的复杂多元关系,并进行正则化处理。然后利用遥感图像空间一致性原则,计算局部空间邻域散度来保持样本局部邻域结构,并引入样本总体散度来保持高光谱数据的整体结构。在低维嵌入空间中,尽可能使类内数据聚集、类间数据远离,提取鉴别特征用于分类。在Indian Pines和PaviaU高光谱遥感数据集上试验结果表明,本文算法总体分类精度分别达到86.7%和 92.2%。相比传统光谱维数约简算法,该算法可有效改善高光谱图像地物分类性能。     

Dimensionality reduction of hyperspectral data using spectral fractal feature

A new approach for dimensionality reduction of hyperspectral data has been proposed in this article. The method is based on extraction of fractal-based features from the hyperspectral data. The features have been generated using spectral fractal dimension of the spectral response curves (SRCs) after smoothing, interpolating and segmenting the curves. The new features so generated have then been used to classify hyperspectral data. Comparing the post classification accuracies with some other conventional dimensionality reduction methods, it has been found that the proposed method, with less computational complexity than the conventional methods, is able to provide classification accuracy statistically equivalent to those from conventional methods.     

一种基于多维正交判别子空间投影的人脸识别方法

人脸识别中,传统数据降维方法将人脸图像重排列成向量后进行处理,丢失了数据本身的结构特性,导致识别精度不高。本文发展了一种基于张量的数据降维方法———多维正交判别子空间投影。该算法直接用张量描述人脸,并通过张量到矢量投影(tensortovectorprojection,TVP)将张量数据投影到向量判别子空间。此方法寻找相互正交的投影向量集,使得判别子空间中数据类间离散度最大,同时类内离散度最小;进而利用TVP投影将高维张量数据映射成低维向量数据,在合适的约束条件下,这些降维后的向量特征数据是整个人脸数据中最具代表性的特征数据;最后,使用k最近邻(KNN)分类器将这些特征数据分类。利用经典人脸数据库ORL进行实验,验证了本文方法的有效性。     

Variogram Fractal Dimension Based Features for Hyperspectral Data Dimensionality Reduction

In this paper a new approach for fractal based dimensionality reduction of hyperspectral data has been proposed. The features have been generated by multiplying variogram fractal dimension value with spectral energy. Fractal dimension bears the information related to the shape or characteristic of the spectral response curves and the spectral energy bears the information related to class separation. It has been observed that, the features provide accuracy better than 90 % in distinguishing different land cover classes in an urban area, different vegetation types belonging to an agricultural area as well as various types of minerals belonging to the same parent class. Statistical comparison with some conventional dimensionality reduction methods validates the fact that the proposed method, having less computational burden than the conventional methods, is able to produce classification statistically equivalent to those of the conventional methods.     

Multiple Classifier Systems for Hyperspectral Remote Sensing Data Classification

One of the most widely used outputs of remote sensing technology is Hyperspectral image. This large amount of information can increase classification accuracy. But at the same time, conventional classification techniques are facing the problem of statistical estimation in high-dimensional space. Recently in remote sensing, support vector machines (SVMs) have shown very suitable performance in classifying high dimensionality problem. Another strategy that has recently been used in remote sensing is multiple classifier system (MCS). It can also improve classification accuracy by combining different classifier methods or by a diversity of the same classifier. This paper aims to classify a Hyperspectral data using the most common methods of multiple classifier systems i.e. adaboost and bagging and a MCS based on SVM. The data used in the paper is an AVIRIS data with 224 spectral bands. The final results show the high capability of SVMs and MCSs in classifying high dimensionality data.     

Utilization of Resourcesat-1 data for improved crop discrimination

Subsequent to the launch of the state-of-art third generation Indian Remote Sensing satellite, Resourcesat-1, studies have been conducted to understand the capabilities of the on-board sensors for crop discrimination. The paper discusses the unique capabilities of the AWiFS, LISS-III and LISS-IV sensors in terms of their dimensionality, radiometry and spatial resolutions for crop discrimination and monitoring. The studies have indicated better crop discriminability especially using the short wave infrared data in 1.55–1.70 μm data among the spectrally confusing land cover classes, attributed to the relative differences of water contents. 10-bit radiometry of AWiFS data in four bands has been observed to be a better discriminant. Intrafield variability was very well captured by the LISS-IV data revealing the potential of data for applications like precision farming. The studies have revealed that potential of Resourcesat-1 data becoming the workhorse for several agricultural applications.     

Some issues related with sub-pixel classification using HYSI data from IMS-1 satellite

The mixed pixels are treated as noise or uncertainty in class allocation of a pixel and conventional hard classification algorithms may thus produce inaccurate classification outputs. Thus application of sub-pixel or soft classification methods have been adopted for classification of images acquired in complex and uncertain environment. The main objective of this research work has been to study the effect of feature dimensionality using statistical learning classifier — support vector machine (SVM with sigmoid kernel) while using different single and composite operators in fuzzy-based error matrixes generation. In this work mixed pixels have been used at allocation and testing stages and sub-pixel classification outputs have been evaluated using fuzzy-based error matrixes applying single and composite operators for generating matrix. As subpixel accuracy assessment were not available in commercial software, so in-house SMIC (Sub-pixel Multispectral Image Classifier) package has been used. Data used for this research work was from HySI sensor at 506 m spatial resolution from Indian Mini Satellite-1 (IMS-1) satellite launched on April 28, 2008 by Indian Space Research Organisation using Polar Satellite Launch Vehicle (PSLV) C9, acquired on 18^th May 2008 for classification output and IRS-P6, AWIFS data for testing at sub-pixel reference data. The finding of this research illustrate that the uncertainty estimation at accuracy assessment stage can be carried while using single and composite operators and overall maximum accuracy was achieved while using 40 (13 to 52 bands) band data of HySI (IMS-1).     

结合多光谱影像降维与深度学习的城市单木树冠检测

多光谱数据的降维处理对基于深度学习的单木树冠检测研究有重要意义,如何使用合适的降维方法以提高单木检测的精度却少有研究讨论。本文使用无人机搭载多光谱相机进行航拍作业,采集研究区内银杏树种多光谱影像。将原始多光谱影像通过特征波段选择、特征提取、波段组合的方法生成5种不同的数据集用于训练3种经典的深度学习网络FPN-Faster-R-CNN,YOLOv3,Faster R-CNN。其中由波段组合方法得到的近红外、红色、绿色波段组合在不同类型的目标检测网络中都有最好的检测结果,其中FPN-Faster-R-CNN网络对银杏树冠的检测精度最高为88.4%,由OIF指标得到的蓝色、红色、近红外波段组合信息量最高,但在所有网络中的平均检测精度最低,仅为79.3%。实验结果表明：在不同波段降维方法中,若降维后的影像中目标物体的色彩与背景差异较明显,且轮廓清晰,则深度学习网络对树冠的检测可获得较好的结果。而影像自身的信息量则对深度学习网络的树冠检测能力的提升作用有限。本研究中针对多光谱影像的降维方法分析,为基于深度学习的单木树冠检测研究提供了重要的实验参考。     

Land cover mapping at Alkali Flat and Lake Lucero,White Sands,New Mexico,USA using multi-temporal and multi-spectral remote sensing data

The goal of this research is to map land cover patterns and to detect changes that occurred at Alkali Flat and Lake Lucero, White Sands using multispectral Landsat 7 Enhanced Thematic Mapper Plus (ETM+), Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), Advanced Land Imager (ALI), and hyperspectral Hyperion and Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) data. The other objectives of this study were: (1) to evaluate the information dimensionality limits of Landsat 7 ETM+, ASTER, ALI, Hyperion, and AVIRIS data with respect to signal-to-noise and spectral resolution, (2) to determine the spatial distribution and fractional abundances of land cover endmembers, and (3) to check ground correspondence with satellite data. A better understanding of the spatial and spectral resolution of these sensors, optimum spectral bands and their information contents, appropriate image processing methods, spectral signatures of land cover classes, and atmospheric effects are needed to our ability to detect and map minerals from space. Image spectra were validated using samples collected from various localities across Alkali Flat and Lake Lucero. These samples were measured in the laboratory using VNIR–SWIR (0.4–2.5 μm) spectra and X-ray Diffraction (XRD) method. Dry gypsum deposits, wet gypsum deposits, standing water, green vegetation, and clastic alluvial sediments dominated by mixtures of ferric iron (ferricrete) and calcite were identified in the study area using Minimum Noise Fraction (MNF), Pixel Purity Index (PPI), and n-D Visualization. The results of MNF confirm that AVIRIS and Hyperion data have higher information dimensionality thresholds exceeding the number of available bands of Landsat 7 ETM+, ASTER, and ALI data. ASTER and ALI data can be a reasonable alternative to AVIRIS and Hyperion data for the purpose of monitoring land cover, hydrology and sedimentation in the basin. The spectral unmixing analysis and dimensionality eigen analysis between the various datasets helped to uncover the most optimum spatial–spectral–temporal and radiometric-resolution sensor characteristics for remote sensing based on monitoring of seasonal land cover, surface water, groundwater, and alluvial sediment input changes within the basin. The results demonstrated good agreement between ground truth data and XRD analysis of samples, and the results of Matched Filtering (MF) mapping method.     

遥感数据自动分类中应用辅助数据的原理

本文从目视解译信息空间的模拟、模式维数对模式样本集线性可分性的影响、模式空间各模式类别可分性测度OIF指数及迹准则判断原理等模式识别原理出发,初步讨论了在遥感影象自动分类中利用辅助数据的理论依据,并以藏北西扎地区草资源调查中利用辅助数据进行自动分类的实例说明辅助数据提高自动分类精度的有效性。     

加权空-谱联合保持嵌入的高光谱遥感影像降维方法

高光谱遥感影像数据量大、波段数多,容易导致“维数灾难”。传统流形学习方法一般仅考虑其光谱特征,忽略了空间信息。为此提出一种非监督的基于加权空-谱联合保持嵌入(WSCPE)的维数约简算法。首先采用加权均值滤波(WMF)方法对高光谱影像进行滤波,以消除噪点和背景点的干扰。然后根据遥感影像地物分布的空间一致性,通过采用加权空-谱联合距离(WSCD)来融合像素点的光谱信息和空间信息,有效选取各像素点的空-谱近邻,并根据像素点与其空-谱近邻点之间的坐标距离来有区别的利用其近邻点进行流形重构,提取低维鉴别特征进行地物分类。在PaviaU和Indian Pines数据集上的分类结果表明,总体分类精度分别达到了98.89%和95.47%。该方法在反映影像内部流形结构的同时,有效融合了影像的空间-光谱信息,故能提高影像特征的鉴别性,并提升分类性能。     

Fractal-based dimensionality reduction of hyperspectral images

The spectral reflectance of any pixel in a remote sensing image depends on the characteristics of the particular land cover (LC) present in the Instantaneous Field of View (IFOV) of the sensor. The fractal dimension of the spectral reflectance curve (SRC) of any pixel can thus be visualized as a representation of the characteristics of the LC. Based on this, a fractalbased method for reduction of the dimensionality of Hyperspectral (HS) images has been investigated. The fractal dimension (FD) of SRC has been calculated by adopting a method based on Hausdorff metric that reduces the dimensionality from N HS bands to a single feature incorporating the characteristics associated with each of the bands. Further, it has been established that FD values can be used as a feature to identify anomaly in water cover.     

Scrutinising MODIS and GIMMS Vegetation Indices for Extracting Growth Rhythm of Natural Vegetation in India

Satellite derived vegetation vigour has been successfully used for various environmental modeling since 1972. However, extraction of reliable annual growth information about natural vegetation (i.e., phenology) has been of recent interest due to their important role in many global models and free availability of time-series satellite data. In this study, usability of Moderate Resolution Imaging Spectro-radiometer (MODIS) and Global Inventory Modelling and Mapping Studies (GIMMS) based products in extracting phenology information about evergreen, semi-evergreen, moist deciduous and dry deciduous vegetation in India was explored. The MODIS NDVI and EVI time-series data (MOD13C1: 5.6 km spatial resolution with 16 day temporal resolution—2001 to 2010) and GIMMS NDVI time-series data(8 km spatial resolution with 15 day temporal resolution—2000 to 2006) were used. These three differently derived vegetation indices were analysed to extract and understand the vegetative growth rhythm over different regions of India. Algorithm was developed to derive onset of greenness and end of senescence automatically. The comparative analysis about differences in the results from these products was carried out. Due to dominant noise in the values of NDVI from GIMMS and MODIS during monsoon period the phenology rhythm were wrongly depicted, especially for evergreen and semi-evergreen vegetation in India. Hence, care is needed before using these data sets for understanding vegetative dynamics, biomass cestimation and carbon studies. MODIS EVI based results were truthful and comparable to ground reality. The study reveals spatio-temporal patterns of phenology, rate of greening, rate of senescence, and differences in results from these three products.     

Exploring the effectiveness of location‐based social media in modeling user activity space: A case study of Weibo

Location‐based social media (LBSM) has been widely utilized to supplement traditional survey methods in modeling human activity patterns. However, there has not been sufficient study to assess the reliability of these data in deriving human movement. This research aims to evaluate how data collection duration and sample sizes affect the reliability of LBSM data in activity modeling based on two indicators: radius of gyration (ROG) and entropy. We use a linear regression model with logarithmic transformation to approximate how the magnitude of each indicator changes with different data collection durations—from 1 to 12 months. The results indicate that both ROG and entropy increase when the amount of data increases. However, the rate of increase slows down and approaches zero eventually. We also approximated the limit values and verified that with 12‐month data, we are at approximately >95% magnitude of the limit values for both indicators in all three cities. The clustering analysis also demonstrated that there are outlier users who exhibit distinct patterns. This case study focuses on three Chinese cities (Beijing, Shanghai, and Guangzhou) and provides a useful reference to explore the balance point between data effectiveness and an appropriate sample size from LBSM data.     

Advances in techniques to formulate the watertight concept for cadastre

The world’s rising urban density expansion has resulted in a proliferation of attempts to efficiently use space and a higher level of spatial complexity in metropolitan areas. 3D geospatial data models are increasingly being embraced to facilitate communicating the spatial dimensions of complex built environments in different applications. For example, the use of 3D models in land administration systems has been recognized as a good approach for communicating the spatial complexity of legal spaces within multi‐storey buildings. The spatial extent of legal space—to which rights, restrictions and responsibilities relate in a 3D digital cadastre—needs to be accurately defined and geometrically closed; watertight. Therefore, this study aims to address the challenges regarding checking the closure of diverse 3D legal spaces and engage several techniques to formulate the watertight concept for cadastre. The research’s methodology is built on a 3D polyhedral surface using a half‐edge data structure. A primitive check is employed to assess the spatial consistency of lower‐dimensional primitives of 3D objects. Subsequently, advanced closure checks ensure the closure of volumetric legal spaces represented by 2‐manifold and non‐2‐manifold data models. The article concludes that, by adopting the proposed approaches, the internal spatial consistency of legal spaces in urban land administration will be certified.     

高光谱数据的降维及Tabu搜索算法的应用

在高光谱影像的分类过程中,如何有效地降低特征空间的维数,又能保证原始数据所包含的丰富地物信息是一项十分重要而繁琐的工作.深入分析了这种降维的必要性,并针对当前常用的降维方法存在的问题,提出了运用Tabu搜索算法获取对分类最为有利的特征波段的思想.考虑到高光谱数据的特点,指出了算法运行中应该注意的若干关键参数设置问题.实验表明,Tabu搜索算法在求解质量和执行效率方面都有着良好的表现,可以用于高光谱数据的降维处理.     

Some issues in contextual fuzzy <Emphasis Type="Italic">c</Emphasis>-means classification of remotely sensed data for land cover mapping

Earlier for the hard classification techniques contextual information was used to improve classification accuracy. While modelling the spatial contextual information for hard classifiers using Markov Random Field it has been found that Metropolis algorithm is easier to program and it performs better in comparison to the Gibbs sampler. In the present study it has been found that incase of soft contextual classification Metropolis algorithm fails to sample from a random field efficiently and from the analysis it was found that Metropolis algorithm is not suitable for soft contextual classification due to the high dimensionality of the soft outputs.