基于ANN端元估计的高光谱图像解混算法

针对高光谱图像解混问题进行研究,发现传统解混算法在保持端元数目不变的情况下,得到的解混精度不高。为此,基于人工神经网络（ANN）提出一种估计单像素点中端元数目和类别的解混算法。首先利用人工神经网络对遥感图像中各个像素的端元数目和类别进行估计;之后依据估计结果确定解混算法的目标函数,并引入改进的差分搜索算法对目标函数进行优化求解;最终获取地物丰度和待求参数,实现高光谱图像的解混。仿真数据和真实遥感数据实验表明,与现有的解混算法相比,所提解混算法具有更高的解混性能,更加符合实际场景的情况。     

基于差分搜索的高光谱图像解混算法

针对高光谱图像解混问题进行研究,发现高光谱图像中各个端元的分布不完全独立,不能将盲源分离方法直接应用于高光谱图像解混。为此,提出了一种基于差分搜索的高光谱图像解混算法。该算法根据高光谱图像丰度非负和丰度和为一特性构造相应的约束项,与互信息相结合作为目标函数,利用差分搜索算法对该目标函数进行优化求解来实现高光谱图像解混。仿真数据和实际数据实验表明,该算法能够有效解决高光谱图像解混问题,与已有其它算法相比,提高了图像解混的精度,并且针对不含纯像元的高光谱图像具有很好的解混效果。     

Abundance estimation of crossover double particle swarms optimization for hyperspectral remote sensing imagery

ABSTRACT

Hyperspectral unmixing (HU) is an important technique for extracting materials and their abundance in hyperspectral remote sensing imagery. The presence of nonlinear mixing of light on the ground poses a difficult problem when estimating abundance fractions of all pixels. This problem makes the foundation of algorithms that can adapt all types of nonlinear mixing on the ground more complex and challenged. In this paper, a new bionic intelligent algorithm named crossover double particle swarms optimization (CDPSO) has been presented to estimate abundance for hyperspectral remote sensing imagery. The reconstruction error is used as the objective function for HU based on multilinear mixing model, and the nonlinear unmixing is transformed into an optimization problem. By improving the optimization performance of PSO for HU, we embed two types of new strategies, including double particle swarms crossover and swarm re-initialization, respectively. Our experiments, conducted using both synthetic and real hyperspectral data, demonstrate that the proposed CDPSO algorithm can outperform other state-of-the-art unmixing methods.     

Comparison of hyperspectral imagery with aerial photography and multispectral imagery for mapping broom snakeweed

Broom snakeweed (Gutierrezia sarothrae (Pursh) Britt. & Rusby) is one of the most widespread and abundant rangeland weeds in western North America. The objectives of this study were to evaluate airborne hyperspectral imagery and compare it with aerial colour-infrared (CIR) photography and multispectral digital imagery for mapping broom snakeweed infestations. Airborne hyperspectral imagery along with aerial CIR photographs and digital CIR images was acquired from a rangeland area in south Texas. The hyperspectral imagery was transformed using minimum noise fraction (MNF) and then classified using minimum distance, Mahalanobis distance, maximum likelihood, and spectral angle mapper (SAM) classifiers. The digitized aerial photographs and the digital images were respectively mosaicked as one photographic image and one digital image; these were then classified using the same classifiers. Accuracy assessment showed that the maximum likelihood classifier performed the best for the three types of images. The best overall accuracies for three-class classification maps (snakeweed, mixed woody and mixed herbaceous) were 91.0%, 92.5%, and 95.0%, respectively, for the CIR photographic image, the digital CIR image and the MNF-transformed hyperspectral image. Kappa analysis showed that there were no significant differences in maximum likelihood-based classifications among the three types of images. These results indicate that airborne hyperspectral imagery along with aerial photography and multispectral imagery can be used for monitoring and mapping broom snakeweed infestations on rangelands.     

Commodity cluster-based parallel processing of hyperspectral imagery

The rapid development of space and computer technologies has made possible to store a large amount of remotely sensed image data, collected from heterogeneous sources. In particular, NASA is continuously gathering imagery data with hyperspectral Earth observing sensors such as the Airborne Visible-Infrared Imaging Spectrometer (AVIRIS) or the Hyperion imager aboard Earth Observing-1 (EO-1) spacecraft. The development of fast techniques for transforming the massive amount of collected data into scientific understanding is critical for space-based Earth science and planetary exploration. This paper describes commodity cluster-based parallel data analysis strategies for hyperspectral imagery, a new class of image data that comprises hundreds of spectral bands at different wavelength channels for the same area on the surface of the Earth. An unsupervised technique that integrates the spatial and spectral information in the image data using multi-channel morphological transformations is parallelized and compared to other available parallel algorithms. The code's portability, reusability and scalability are illustrated by using two high-performance parallel computing architectures: a distributed memory, multiple instruction multiple data (MIMD)-style multicomputer at European Center for Parallelism of Barcelona, and a Beowulf cluster at NASA's Goddard Space Flight Center. Experimental results suggest that Beowulf clusters are a source of computational power that is both accessible and applicable to obtaining results in valid response times in information extraction applications from hyperspectral imagery.     

Aspects of operational atmospheric correction of hyperspectral imagery

The large number of spectral bands of hyperspectral instruments and the time required for the calculation of atmospheric look-up tables and the reflectance image cube pose very challenging requirements on an operational processing facility. This contribution presents some aspects and suggestions to reduce the processing time. Essential components are a precalculated database with a reduced number of spectral bands, an interactive phase to determine the appropriate atmospheric parameters, and a choice between medium and high accuracy levels for the atmospheric correction. The medium accuracy levels work with look-up tables for a reduced number of spectral bands employing interpolation for the channels omitted in the look-up tables. The high accuracy level uses tables for all channels and includes the scan angle dependence of the atmospheric radiance and transmittance functions. These ideas were successfully implemented and tested during several airborne hyperspectral campaigns resulting in an estimated time saving of a factor 3-7. The deviations of field measured reflectance spectra and spectra retrieved from airborne HyMap imagery are in the range of 2-3% or better.     

基于目标优化的高光谱图像亚像元定位

目的 高光谱图像混合像元的普遍存在使得传统的分类技术难以准确确定地物空间分布,亚像元定位技术是解决该问题的有效手段。针对连通区域存在孤立点或孤立两点等特例时,通过链码长度求周长最小无法保证最优结果及优化过程计算量大的问题,提出了一种改进的高光谱图像亚像元定位方法。方法 以光谱解混结合二进制粒子群优化构建算法框架,根据光谱解混结果近似估计每个像元对应的亚像元组成,通过分析连通区域存在特例时基于链码长度求周长最小无法保证结果最优的原因,提出修改孤立区域的周长并考虑连通区域个数构造代价函数,最后利用二进制粒子群优化实现亚像元定位。为了减少算法的时间复杂度,根据地物空间分布特点,采用局部分析代替全局分析,提出了新的迭代优化策略。结果 相比直接基于链码长度求周长最小的优化结果,基于改进的目标函数优化后,大部分区域边界更明显,并且没有孤立1点和孤立两点的区域,识别率可以提高2%以上,Kappa系数增加0.05以上,新的优化策略可以使算法运算时间减少近一半。结论 实验结果表明,本文方法能有效提高亚像元定位精度,同时降低时间复杂度。因为高光谱图像中均匀混合区域不同地物的分布空间相关性不强,因此本文方法适用于非均匀混合的高光谱图像的亚像元定位。     

Fast real-time onboard processing of hyperspectral imagery for detection and classification

Remotely sensed hyperspectral imagery has many important applications since its high-spectral resolution enables more accurate object detection and classification. To support immediate decision-making in critical circumstances, real-time onboard implementation is greatly desired. This paper investigates real-time implementation of several popular detection and classification algorithms for image data with different formats. An effective approach to speeding up real-time implementation is proposed by using a small portion of pixels in the evaluation of data statistics. An empirical rule of an appropriate percentage of pixels to be used is investigated, which results in reduced computational complexity and simplified hardware implementation. An overall system architecture is also provided.

Integrating waveform lidar with hyperspectral imagery for inventory of a northern temperate forest

It has been suggested that attempts to use remote sensing to map the spatial and structural patterns of individual tree species abundances in heterogeneous forests, such as those found in northeastern North America, may benefit from the integration of hyperspectral or multi-spectral information with other active sensor data such as lidar. Towards this end, we describe the integrated and individual capabilities of waveform lidar and hyperspectral data to estimate three common forest measurements - basal area (BA), above-ground biomass (AGBM) and quadratic mean stem diameter (QMSD) - in a northern temperate mixed conifer and deciduous forest. The use of this data to discriminate distribution and abundance patterns of five common and often, dominant tree species was also explored. Waveform lidar imagery was acquired in July 2003 over the 1000 ha. Bartlett Experimental Forest (BEF) in central New Hampshire (USA) using NASA's airborne Laser Vegetation Imaging Sensor (LVIS). High spectral resolution imagery was likewise acquired in August 2003 using NASA's Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). Field data (2001-2003) from over 400 US Forest Service Northern Research Station (USFS NRS) plots were used to determine actual site conditions.Results suggest that the integrated data sets of hyperspectral and waveform lidar provide improved outcomes over use of either data set alone in evaluating common forest metrics. Across all forest conditions, 8-9% more of the variation in AGBM, BA, and QMSD was explained by use of the integrated sensor data in comparison to either AVIRIS or LVIS metrics applied singly, with estimated error 5-8% lower for these variables. Notably, in an analysis using integrated data limited to unmanaged forest tracts, AGBM coefficients of determination improved by 25% or more, while corresponding error levels decreased by over 25%. When data were restricted based on the presence of individual tree species within plots, AVIRIS data alone best predicted species-specific patterns of abundance as determined by species fraction of biomass. Nonetheless, use of LVIS and AVIRIS data - in tandem - produced complementary maps of estimated abundance and structure for individual tree species, providing a promising adjunct to traditional forest inventory and conservation biology planning efforts.     

Retrieval of microphysical and optical properties in aerosol plumes with hyperspectral imagery: L-APOM method

This paper presents the retrieval method L-APOM which aims at characterizing the microphysical and optical properties of aerosol plumes from hyperspectral images with high spatial resolution. The inversion process is divided into three steps: estimation of the ground reflectance below the plume, characterization of the standard atmosphere (gases and background aerosols) and estimation of the plume aerosols properties. As using spectral information only is not sufficient to insure uniqueness of solutions, original constraints are added by assuming slow spatial variations of particles properties within the plume. The whole inversion process is validated on a large set of simulated images and reveals to remain accurate even in the worst cases of noise: relative estimation errors of aerosol properties remain between 10% and 20% in most cases. L-APOM is applied on a real AVIRIS hyperspectral image of a biomass burning plume for which in situ measurements are available. Retrieved properties appear globally consistent with measurements.     

Mapping of continuous floristic gradients in grasslands using hyperspectral imagery

Transitions between plant species assemblages are often continuous with the form of the transition dependent on the ‘slope’ of environmental gradients and on the style of self-organization in vegetation. Image segmentation can present misleading or even erroneous results if applied to continuous spatial changes in vegetation. Even methods that allow for multiple-class memberships of pixels presuppose the existence of ideal types of species assemblages that constitute mixtures—an assumption that does not fit the case of continua where any section of a gradient is as ‘pure’ as any other section like in modulations of grassland species composition.Thus, we attempted to spatially model floristic gradients in Bavarian meadows by extrapolating axes of an unconstrained ordination of species data. The models were based on high-resolution hyperspectral airborne imagery. We further modelled the distribution of plant functional response types (Ellenberg indicator values) and the cover values of selected species. The models were made with partial least squares (PLS) regression analyses. The realistic utility of the regression models was evaluated by full leave-one-out cross-validation.The modelled floristic gradients showed a considerable agreement with ground-based observations of floristic gradients (R²=0.71 and 0.66 for the first two axes of ordination). Apart from mapping the most important continuous floristic differences, we mapped gradients in the appearance of plant functional response groups as represented by averaged Ellenberg indicator values for soil pH (R²=0.76), water supply (R²=0.66) and nutrient supply (R²=0.75), while models for the cover of single species were weak.Compared to many other vegetation attributes, plant species composition is difficult to detect with remote sensing techniques. This is partly caused by a lack of compatibility between methods of vegetation ecology and remote sensing. We believe that the present study has the potential to increase compatibility as neither spectral nor vegetation information gets lost by a classifying step.     

A robust and fast geometry based unmixing algorithm for hyperspectral imagery

A geometry based unmixing method is proposed in this paper. A sequential algorithm is used to find a convex cone under the maximum angle criterion. Initialization of this algorithm was improved by an algebra method and its speed was improved with sequential angle subspace updating strategy. In order to demonstrate the performance of the proposed unmixing method, two other unmixing methods, convex cone analysis (CCA) and sequential maximum angle convex cone SMACC, are used for comparison. The experimental results indicate that the proposed method is more robust and faster.     

谐波分析光谱角制图高光谱影像分类

目的 针对光谱角制图(SAM)分类算法对高光谱像元光谱曲线的局部特征和其辐射强度不敏感,而且易受噪声和维数灾难影响,致使分类效率低和精度较差等缺陷,将谐波分析(HA)技术引入到SAM高光谱影像分类中,提出一种基于谐波分析的光谱角制图(HA-SAM)高光谱影像分类算法.方法 利用HA技术将高光谱影像从光谱维变换到能量谱特征维空间,并提取低次谐波分量及特征系数(谐波余项、相位和振幅),用特征系数组成的向量代替光谱向量,对高光谱影像进行SAM分类.结果 将SAM和HA-SAM同时应用于EO-1卫星的Hyperion高光谱影像分类,通过对比和分析,验证了HA-SAM的优越性,再选择AVIRIS(airborne visible infrared imaging spectrometer)高光谱影像对HA-SAM进行验证,结果表明该算法具有较强的普适性.结论 HA-SAM提高了传统SAM高光谱影像分类的效率和精度,而且适用性较强具有良好的应用前景.     

Postfire soil burn severity mapping with hyperspectral image unmixing

Burn severity is mapped after wildfires to evaluate immediate and long-term fire effects on the landscape. Remotely sensed hyperspectral imagery has the potential to provide important information about fine-scale ground cover components that are indicative of burn severity after large wildland fires. Airborne hyperspectral imagery and ground data were collected after the 2002 Hayman Fire in Colorado to assess the application of high resolution imagery for burn severity mapping and to compare it to standard burn severity mapping methods. Mixture Tuned Matched Filtering (MTMF), a partial spectral unmixing algorithm, was used to identify the spectral abundance of ash, soil, and scorched and green vegetation in the burned area. The overall performance of the MTMF for predicting the ground cover components was satisfactory (r² = 0.21 to 0.48) based on a comparison to fractional ash, soil, and vegetation cover measured on ground validation plots. The relationship between Landsat-derived differenced Normalized Burn Ratio (dNBR) values and the ground data was also evaluated (r² = 0.20 to 0.58) and found to be comparable to the MTMF. However, the quantitative information provided by the fine-scale hyperspectral imagery makes it possible to more accurately assess the effects of the fire on the soil surface by identifying discrete ground cover characteristics. These surface effects, especially soil and ash cover and the lack of any remaining vegetative cover, directly relate to potential postfire watershed response processes.     

Band elimination of hyperspectral imagery using partitioned band image correlation and capacitory discrimination

Due to the very large number of bands in hyperspectral imagery, two major problems which arise during classification are the ‘curse of dimensionality’ and computational complexity. To overcome these, dimensionality reduction is an important task for hyperspectral image analysis. An unsupervised band elimination method is proposed which iteratively eliminates one band from the pair of most correlated neighbouring bands depending on the discriminating capability of the bands. Correlation between neighbouring bands is calculated over partitioned band images. Capacitory discrimination is used to measure the discrimination capability of a band image. Finally, four evaluation measures, namely classification accuracy, kappa coefficient, class separability, and entropy are calculated over the selected bands to measure the efficiency of the proposed method. The proposed unsupervised band elimination technique is compared to three popular state-of-the-art approaches, both qualitatively and quantitatively, and shows promising results compared to them.     

A new approach for mapping of Biological Soil Crusts in semidesert areas with hyperspectral imagery

Biological Soil Crusts (BSCs), consisting of cyanobacteria, algae, microfungi, lichens and bryophytes in varying proportions, live within or immediately on top of the uppermost millimeters of soil, where they form a more or less firm aggregation of soil particles and organisms. They mainly occur in soils of arid and semi-arid regions, which cover more than 35% of the earth's land surface and are assumed to play a major role as primary producers, C- and N-sinks and soil stabilizers.

In order to establish a methodology for mapping of BSCs, their spectral characteristics with respect to different crust types were analyzed. The resulting reflectance spectra of different crust types had a shallow absorption feature centered around 680 nm in common, in which they differed from the spectra of bare soil.

In October 2004, hyperspectral CASI data with a spatial resolution of 1 m were recorded in conjunction with field spectroscopic measurements in the Succulent Karoo, South Africa. Available spectral indices for Biological Soil Crusts were tested on the image but did not produce satisfying classifications. Therefore, an alternative approach was established based on spectral field data, field observations and the hyperspectral dataset. The newly developed Continuum Removal Crust Identification Algorithm (CRCIA) is based on small and narrow spectral characteristics, that were extracted by continuum removal and subsequently expressed as a set of logical conditions. Using this method, 16.2% of the area which covers 12 km² of gently sloping hills with some granite outcrops were classified as BSCs. Validation of the classification resulted in a Kappa index of 0.831.

In a next step, the methodology will be tested with regard to scale-dependent effects and applied to images covering areas with additional types of BSCs and soil to develop a robust and generally applicable method.     

Surface soil moisture estimation over dense crop using Envisat ASAR and Landsat TM imagery: an approach

This study focuses on developing a new method of surface soil moisture estimation over wheat fields using Environmental Satellite Advanced Synthetic Aperture Radar (Envisat ASAR) and Landsat Thematic Mapper (TM) data. The Michigan Microwave Canopy Scattering (MIMICS) model was used to simulate wheat canopy backscattering coefficients from experiment plots at incidence angles of 23° (IS2) and 43.9° (IS7). Based on simulated data, the scattering characteristics of a wheat canopy were first investigated in order to derive an optimal configuration of polarization (HH) and incidence angle (IS2) for soil moisture estimation. Then a parametric model was developed to describe wheat canopy backscattering at the optimal configuration. In addition, direct backscattering and two-way transmissivity of wheat crowns were derived from the TM normalized difference vegetation index (NDVI); direct ground backscattering was derived from surface soil moisture and TM NDVI; and backscattering from double scattering was derived from total backscattering. A semi-empirical model for soil moisture estimation was derived from the parametric model. Coefficients in the semi-empirical model were obtained using a calibration approach based on measured soil moisture, ASAR, and TM data. A validation of the model was performed over the experimental area. In this study, the root mean square error (RMSE) for the estimated soil moisture was 0.041 m³ m^?3, and the correlation coefficient between the measured and estimated soil moisture was 0.84. The experimental results indicate that the semi-empirical model could improve soil moisture estimation compared to an empirical model.     

Comparison of spectral and spatial windows for local anomaly detection in hyperspectral imagery

A new way of implementing two local anomaly detectors in a hyperspectral image is presented in this study. Generally, most local anomaly detector implementations are carried out on the spatial windows of images, because the local area of the image scene is more suitable for a single statistical model than for global data. These detectors are applied by using linear projections. However, these detectors are quite improper if the hyperspectral dataset is adopted as the nonlinear manifolds in spectral space. As multivariate data, the hyperspectral image datasets can be considered to be low-dimensional manifolds embedded in the high-dimensional spectral space. In real environments, the nonlinear spectral mixture occurs more frequently, and these manifolds could be nonlinear. In this case, traditional local anomaly detectors are based on linear projections and cannot distinguish weak anomalies from background data. In this article, local linear manifold learning concepts have been adopted, and anomaly detection algorithms have used spectral space windows with respect to the linear projection. Output performance is determined by comparison between the proposed detectors and the classic spatial local detectors accompanied by the hyperspectral remote-sensing images. The result demonstrates that the effectiveness of the proposed algorithms is promising to improve detection of weak anomalies and to decrease false alarms.     

Detection and classification of invasive saltcedar through high spatial resolution airborne hyperspectral imagery

We evaluated the performance of airborne HyperSpecTIR (HST) images for detecting and classifying the invasive riparian vegetation saltcedar along the Muddy River in Clark County, Nevada. HyperSpecTIR image reflectance spectra (227 bands, 450–2450 nm) were acquired for the following four vegetation covers: invasive saltcedar, native honey mesquite, grassland patches and crops. We compared five feature reduction approaches: band selection based on Jeffreys–Matusita distance, principal component analysis (PCA), minimum noise fraction (MNF), segmented principal component transform (SPCT) and segmented minimum noise fraction (SMNF). In addition, maximum likelihood (ML) and two spectral angle mapper (SAM) classifiers were applied to all extracted bands or features. Classification accuracies were compared among all classification approaches. Although the overall accuracy of maximal likelihood classifiers generally surpassed that of SAM classifiers, the highest overall accuracy was achieved by a SMNF-SAM combination with adjusted angular thresholds for classes. We concluded that high spectral and spatial resolution imagery can be used to detect and classify invasive saltcedar in this arid area.     

Assessment of noise reduction of hyperspectral imagery using a target detection application

This article presents an evaluation of a previously proposed noise reduction technique for hyperspectral imagery with regard to its use in remote sensing applications. Target detection from hyperspectral imagery was selected as an example for the evaluation. A hyperspectral datacube acquired using the airborne Shortwave Infrared Full Spectrum Imager (SFSI)-II with man-made targets deployed in the scene of the datacube was tested. In addition to an evaluation using the receiver operating characteristic (ROC) curve approach, we used a spectral unmixing technique to generate the fraction images of the target materials, measured the area of the targets derived from the datacube before and after applying the noise reduction technology, and then compared the derived target areas to the real targets to assess the detectability of the targets. The area ratio between a derived target and the real target was used as the criterion in the evaluation. The evaluation results show that the noise reduction technique can help to better serve remote sensing applications. The small targets that cannot be detected from the original datacube were detected after the noise reduction using the technology.