Efficient classification of the hyperspectral images using deep learning

Multimedia Tools and Applications - Classification techniques applicable to the hyperspectral images do not extract deep features from the hyperspectral image efficiently. In this work, a deep...     

Wildfire temperature and land cover modeling using hyperspectral data

Wildfire temperature retrieval commonly uses measured radiance from a middle infrared channel and a thermal infrared channel to separate fire emitted radiance from the background emitted radiance. Emitted radiance at shorter wavelengths, including the shortwave infrared, is measurable for objects above a temperature of 500 K. The spectral shape and radiance of thermal emission within the shortwave infrared can be used to retrieve fire temperature. Airborne Visible Infrared Imaging Spectrometer (AVIRIS) data were used to estimate fire properties and background properties for the 2003 Simi Fire in Southern California, USA. A spectral library of emitted radiance endmembers corresponding to a temperature range of 500-1500 K was created using the MODTRAN radiative transfer model. A second spectral library of reflected solar radiance endmembers, corresponding to four vegetation types and two non-vegetated surfaces, was created using image spectra selected by minimum endmember average root mean square error (RMSE). The best fit combination of an emitted radiance endmember and a reflected solar radiance endmember was found for each spectrum in the AVIRIS scene. Spectra were subset to reduce the effects of variable column water vapor and smoke contamination over the fire. The best fit models were used to produce maps of fire temperature, fire fractional area, background land cover, land cover fraction, and RMSE. The highest fire temperatures were found along the fire front, and lower fire temperatures were found behind the fire front. Saturation of shortwave infrared channels limited modeling of the highest fire temperatures. Spectral similarity of land cover endmembers and smoke impacted the accuracy of modeled land cover. Sensitivity analysis of modeled fire temperatures revealed that the range of temperatures modeled within 5% of minimum RMSE was smallest between 750 and 950 K. Hyperspectral modeling of wildfire temperature and fuels has potential application for fire monitoring and modeling.     

Accuracy of land cover area estimated from coarse spatial resolution images using an unmixing method

Many techniques intended to estimate land coverage of multiple categories occupied within each pixel from such coarse resolution data have been proposed. However, in traditional unmixing studies with coarse resolution imagery such as Advanced Very High Resolution Radiometer (AVHRR) data, it is assumed that only a few endmembers exist throughout an entire image. Therefore, it is essential to evaluate how well an unmixing method would work for various categories within pixels of coarse resolution images. In this study, the land coverage of eight classes in National Oceanic and Atmospheric Administration (NOAA) AVHRR imagery by using finer resolution Landsat Thematic Mapper (TM) imagery was estimated, and the accuracy of these estimated classes was evaluated. The results suggest that this method may be generally useful for comparing multi-spectral images in space and time.     

A new image classification approach for mapping coral density in State of Kuwait using high spatial resolution satellite images

Remote sensing technology can be a valuable tool for mapping coral reef ecosystems. However, the resolution capabilities of remote sensors, the diversity and complexity of coral reef ecosystems, and the low reflectivity of marine environments increase the difficulties in identifying and classifying their features. This research study explores the capability of high spatial resolution (WorldView-2 (WV-2) and Pleiades-1B) and low spatial resolution (Land Remote-Sensing Satellite (Landsat 8)) multispectral (MS) satellite sensors in quantitatively mapping coral density. The Kubbar coral reef ecosystem, located in Kuwait’s southern waters, was selected as the research site. The MS imagery of WV-2, Pleiades-1B and Landsat 8 were, after geometric and radiometric assessment and corrections, subjected to new image classification approach using a Multiple Linear Regression (MLR) analysis. The new approach of MLR coral density analysis used the dependent variable of coral density percentage from ground truth and independent variables of spectral reflectance from selected imagery, depth (as estimated from a surface derived from bathymetric charts) and distance to land or reef unit centre. Accuracy assessment using independent ground truth was performed for the selected approach and satellite sensors to determine the quality of the information derived from image classification processes. The results showed that coral density maps developed using the MLR coral density model proved to have some level of reliability (radiometrically corrected WV-2 image (the coefficient determination denoted as R-squared (R²) = 0.5, Root-Mean-Square Error (RMSE) = 10) and radiometrically corrected Pleiades-1B image (R² = 0.8, RMSE = 10)). This study suggested using high spectral resolution data and including additional factors (variables) (e.g. water turbidity, temperature and salinity) could contribute to improving the accuracy of coral density maps produced by application of the MLR model; however, all of these would add cost and effort to the mapping process. The outcomes of this research study provide coral reef ecosystem researchers, managers, and decision makers a tool to determine and map coral reef density in more detail than in the past. It will help quantify coral density at particular points in time leading to estimates of change, and allow coral reef ecologists to identify the current coral reef habitat health status, distribution and extent.     

Global land cover classification at 1 km spatial resolution using a classification tree approach

This paper on reports the production of a 1 km spatial resolution land cover classification using data for 1992-1993 from the Advanced Very High Resolution Radiometer (AVHRR). This map will be included as an at-launch product of the Moderate Resolution Imaging Spectroradiometer (MODIS) to serve as an input for several algorithms requiring knowledge of land cover type. The methodology was derived from a similar effort to create a product at 8 km spatial resolution, where high resolution data sets were interpreted in order to derive a coarse-resolution training data set. A set of 37 294 x 1 km pixels was used within a hierarchical tree structure to classify the AVHRR data into 12 classes. The approach taken involved a hierarchy of pair-wise class trees where a logic based on vegetation form was applied until all classes were depicted. Multitemporal AVHRR metrics were used to predict class memberships. Minimum annual red reflectance, peak annual Normalized Difference Vegetation Index (NDVI), and minimum channel three brightness temperature were among the most used metrics. Depictions of forests and woodlands, and areas of mechanized agriculture are in general agreement with other sources of information, while classes such as low biomass agriculture and high-latitude broadleaf forest are not. Comparisons of the final product with regional digital land cover maps derived from high-resolution remotely sensed data reveal general agreement, except for apparently poor depictions of temperate pastures within areas of agriculture. Distinguishing between forest and non-forest was achieved with agreements ranging from 81 to 92% for these regional subsets. The agreements for all classes varied from an average of 65% when viewing all pixels to an average of 82% when viewing only those 1 km pixels consisting of greater than 90% one class within the high-resolution data sets.     

A hybrid approach to urban land use/cover mapping using Landsat 7 Enhanced Thematic Mapper Plus (ETM+) images

A hybrid method that incorporates the advantages of supervised and unsupervised approaches as well as hard and soft classifications was proposed for mapping the land use/cover of the Atlanta metropolitan area using Landsat 7 Enhanced Thematic Mapper Plus (ETM+) data. The unsupervised ISODATA clustering method was initially used to segment the image into a large number of clusters of pixels. With reference to ground data based on 1?:?40?000 colour infrared aerial photographs in the form of Digital Orthophoto Quarter Quadrangle (DOQQ), homogeneous clusters were labelled. Clusters that could not be labelled because of mixed pixels were clipped out and subjected to a supervised fuzzy classification. A final land use/cover map was obtained by a union overlay of the two partial land use/cover maps. This map was evaluated by comparing with maps produced using unsupervised ISODATA clustering, supervised fuzzy and supervised maximum likelihood classification methods. It was found that the hybrid approach was slightly better than the unsupervised ISODATA clustering in land use/cover classification accuracy, most probably because of the supervised fuzzy classification, which effectively dealt with the mixed pixel problem in the low-density urban use category of land use/cover. It was suggested that this hybrid approach can be economically implemented in a standard image processing software package to produce land use/cover maps with higher accuracy from satellite images of moderate spatial resolution in a complex urban environment, where both discrete and continuous land cover elements occur side by side.     

A multiple dataset approach for 30-m resolution land cover mapping: a case study of continental Africa

Recent developments in global land-cover mapping have focused on spatial resolution improvement with more heterogeneous features to integrate spatial, spectral and temporal information. In this study, hundreds of features derived from four seasonal Landsat 8 OLI (Operational Land Imager) spectral bands, Moderate Resolution Imaging Spectroradiometer (MODIS) time series vegetation index (VI) data, night-time light (NTL), digital elevation models (DEM) and climatic variables were used for land cover mapping with a target 30-m resolution for the whole African continent. In total, 49,007 training samples (from 11,231 locations) and 23,803 validation samples (from 5,414 locations) interpreted from seasonal Landsat, MODIS Normalized Difference Vegetation Index (NDVI) time series and high-resolution images in Google Earth were used for classifier training (Random Forest) and map validation. Overall accuracy was 76% at 30-m spatial resolution, which is better than previous land cover mapping for the African continent. Besides, accuracies for cropland were improved dramatically by more than 10%. Our method also addressed many remaining issues for 30-m mapping (e.g. boundary effects and declines in resolution). This framework is promising for automatic and efficient global land cover mapping resulting in better visual effects and classification accuracy.     

Landsat-comparable land cover maps using ASTER and SPOT images: a case study for large-area mapping programmes

The long-term record of global Landsat data is an important resource for studying Earth's system. Given the identified gaps in Landsat data and the undetermined future status of Landsat data availability, alternatives to Landsat imagery need to be tested in an operational environment. In this study, forest land cover and crown closure maps generated from Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and System Pour 1'Observation de la Terre (SPOT) data were compared to Landsat-based map products currently in use by the grizzly bear habitat-mapping program. Overall accuracies greater than 85% were obtained for both ASTER- and SPOT-based land cover maps. The ASTER and SPOT classification accuracies were higher than that achieved by Landsat. Crown closure maps derived from ASTER and SPOT data show a small increase in accuracy when compared to the Landsat products. Overall, these results demonstrate that ASTER and SPOT could provide alternative data sources for producing maps in the event of a gap in the Landsat data.     

Hierarchical mapping of Northern Eurasian land cover using MODIS data

The Northern Eurasian land mass encompasses a diverse array of land cover types including tundra, boreal forest, wetlands, semi-arid steppe, and agricultural land use. Despite the well-established importance of Northern Eurasia in the global carbon and climate system, the distribution and properties of land cover in this region are not well characterized. To address this knowledge and data gap, a hierarchical mapping approach was developed that encompasses the study area for the Northern Eurasia Earth System Partnership Initiative (NEESPI). The Northern Eurasia Land Cover (NELC) database developed in this study follows the FAO-Land Cover Classification System and provides nested groupings of land cover characteristics, with separate layers for land use, wetlands, and tundra. The database implementation is substantially different from other large-scale land cover datasets that provide maps based on a single set of discrete classes. By providing a database consisting of nested maps and complementary layers, the NELC database provides a flexible framework that allows users to tailor maps to suit their needs. The methods used to create the database combine empirically derived climate–vegetation relationships with results from supervised classifications based on Moderate Resolution Imaging Spectroradiometer (MODIS) data. The hierarchical approach provides an effective framework for integrating climate–vegetation relationships with remote sensing-based classifications, and also allows sources of error to be characterized and attributed to specific levels in the hierarchy. The cross-validated accuracy was 73% for the land cover map and 73% and 91% for the agriculture and wetland classifications, respectively. These results support the use of hierarchical classification and climate–vegetation relationships for mapping land cover at continental scales.     

Unsupervised linear contact distributions segmentation algorithm for land cover high resolution panchromatic images

Multimedia Tools and Applications - Automatic segmentation of land use and land cover from high resolution remote sensing imagery has been an essential research area in image processing for the...     

Contextual correction: techniques for improving land cover mapping from remotely sensed images

Large area land cover mapping is an important application of remote sensing. A digital land cover map of Great Britain has recently been compiled by supervised classification of Landsat Thematic Mapper data. The work has involved development of a range of post classification procedures to correct contextual errors associated with the use of spectral classification algorithms. This paper describes these procedures and examines their effects upon the map product including a comparison with field survey data.     

Super resolution reconstruction algorithm of video image based on deep self encoding learning

Multimedia Tools and Applications - Super resolution reconstruction of video image is a research hotspot in the field of image processing, and it is widely used in video surveillance, image...     

Structured approach to land cover mapping or the Atlantic zone of Costa Rica using single date TM data

Abstract

A structured approach to land cover mapping, involving different stages of field work and processing of remote sensing data, is presented. The Processing of TM data of one acquisition date was done by Analysing the Digital data-Structure (PAD) to produce optimum imagery for land-cover mapping of the Atlantic zone in Costa Rica.

Three stages were evaluated:

1. Image-processing in the pre-fieldwork stage to obtain insight into the overall variation of the scene.

2. Small scale reconnaissance fieldwork, and processing thereafter, directed towards the production of thematic imagery guided by the properties of objects and features.

3. Medium scale reconnaissance fieldwork and classification.

With this method we made use of statistical data, such as standard deviations and correlation coefficients, graphic presentations of mean values for the evaluation of ratios as well as variance percentages expressed by principal components. The selection of training fields for statistic calculation was considered to be essential for the final result.     

Advances in classification for land cover mapping using SPOT HRV imagery

Abstract

High-resolution data from the HRV (High Resolution Visible) sensors onboard the SPOT-1 satellite have been utilized for mapping semi-natural and agricultural land cover using automated digital image classification algorithms. Two methods for improving classification performance are discussed. The first technique involves the use of digital terrain information to reduce the effects of topography on spectral information while the second technique involves the classification of land-cover types using training data derived from spectral feature space. Test areas in Snowdonia and the Somerset Levels were used to evaluate the methodology and promising results were achieved. However, the low classification accuracies obtained suggest that spectral classification alone is not a suitable tool to use in the mapping of semi-natural cover types.     

嵌入式深度神经网络高光谱图像聚类

目的 高光谱图像的高维特性和非线性结构给聚类任务带来了"维数灾难"和线性不可分问题，以往的工作将特征提取过程与聚类过程互相剥离，难以同时优化。为了解决上述问题，提出了一种新的嵌入式深度神经网络模糊C均值聚类方法（EDFCC）。方法 EDFCC算法为了提取更加有效的深层特征，联合优化高光谱图像的特征提取和聚类过程，将模糊C均值聚类算法嵌入至深度自编码器网络中，可以保持两任务联合优化的优势，同时利用深度自编码器网络降维以及逼近任意非线性函数的能力，逐步将原始数据映射到潜在特征空间，提取数据的深层特征。所提方法采用模糊C均值聚类算法约束特征提取过程，学习适用于聚类的高光谱数据深层特征，动态调整聚类指示矩阵。结果 实验结果表明，EDFCC算法在Indian Pines和Pavia University两个高光谱数据集上的聚类精度分别达到了42.95%和60.59%，与当前流行的低秩子空间聚类算法（LRSC）相比分别提高了3%和4%，相比于基于自编码器的数据聚类算法（AEKM）分别提高了2%和3%。结论 EDFCC算法能够从高光谱图像的高维光谱信息中提取更加有效的深层特征，提升聚类精度，并且由于EDFCC算法不需要额外的训练过程，大大提升了聚类效率。     

Assessing the efficiency of multispectral satellite and airborne hyperspectral images for land cover mapping in an aquatic environment with emphasis on the water caltrop (Trapa natans)

A number of clear issues are pertinent when considering whether, or not, to use a remotely sensed dataset. We evaluate these issues here by comparing an aerial hyperspectral image at 1.5 m geometric resolution that comprises 128 narrow bands within a spectral range between 400 nm and 1,000 nm as well as a nine-band Landsat 8 image at 30.0 m geometric resolution. We therefore applied Random Forest (RF) and Support Vector Machine (SVM) classifiers utilizing different input data sets to determine the best thematic accuracy for both types of images by involving all possible bands and then minimized them using variable selection and dimension reduction via Minimum Noise Fraction (MNF). We then compared Landsat images to an aerial hyperspectral one. The results of this analysis revealed that band selections based on variable importance and MNF-transformation improved thematic accuracy assessed as Overall Accuracy (OA). Results reveal a 1.00% improvement in OA via variable selection as 59 bands instead of 128 bands and a 1.50% via MNF-transformation of the hyperspectral image. This improvement was 4.52% in the Landsat image when using a MNF-transformation compared to the best performances without transformation or variable selection. Data also showed that application of Landsat spectral range on hyperspectral bands resulted in different outcomes; specifically, SVM resulted in a 91.50% OA while RF resulted in 95.50% OA. Landscape ecology results show that use of the Landsat image provided fewer land cover patches and that differences encompassed 6.30% of the whole area. We therefore conclude that Landsat data can be used with a number of limitations for accurate ecological mapping.     

Estimating and mapping crop residues cover on agricultural lands using hyperspectral and IKONOS data

Crop residues left on agricultural lands after harvest play an important role in controlling and protecting soil against water and wind erosion. One challenge of remote sensing is to differentiate crop residues from bare soil and crop cover, especially when the residues have been weathered and/or when the crop cover phenology is more advanced. Several techniques for mapping and estimating crop residues exist in the literature. However, these methods are time consuming and not suited for quantitative evaluation. They have the disadvantage of being less rigorous and accurate because they do not consider the spectral mixture of different materials in the same pixel. In this study, the potential of hyperspectral (Probe-1) and multispectral high spatial resolution (IKONOS) data were compared for estimating and mapping crop residues on agricultural lands using the constrained linear spectral mixture analysis approach. Image data were spectrally and radiometrically calibrated, atmospherically corrected, as well as geometrically rectified. Pure spectral signatures of residues, bare soil and crop cover were manually extracted from image data based on prior knowledge of the fields. Percent (fraction) cover for each sampling point was extracted using unmixing and validated against ground reference measurements. The best results were achieved for the crop cover (index of agreement (D) = 0.92 and root mean square error (RMSE) = 0.09) adjusted for the impurity of the endmembers canola, pea and wheat, followed by the wheat residues (D = 0.76 and RMSE = 0.12). Considering only the wheat residues in fields with a canola crop, D increases to 0.86. The soil fractions were generally underestimated with D = 0.72, and no significant improvements could be made after adjusting for the shadow effect. The estimations from the IKONOS data were poorer for the same cover types (residues: D = 0.40 and RMSE = 0.24; crop: D = 0.51 and RMSE = 0.38; soil: D = 0.58 and RMSE = 0.29). Relative to the IKONOS data, the better performance of the hyperspectral data is mainly due to the improved spectral band characteristics, especially in the SWIR, which is sensitive to the residues (lignin and cellulose absorption features), soil and crop cover.     

Ground object information extraction from hyperspectral remote sensing images using deep learning algorithm

Mining and utilizing coal resources play an influential role in economic development. In this regard, the feature information extraction in the area is researched to accurately and efficiently assist the production arrangement and deployment in the mining area. First, the detection ability of Hyperspectral Remote Sensing Image (HRSI) technology is analyzed. It has high spectral resolution and many bands. Specific bands can be extracted as needed to highlight target features. According to the characteristics of HRSIs, the data spectrum information and spatial information are comprehensively utilized, and the Convolutional Neural Network (CNN) based on deep learning is employed for feature extraction. CNN allows the machine to automatically obtain data features by learning and guide the classification of features. Taking the Liuyuan research area in Gansu as an example, three CNN models are used to extract and classify the ground features in the area. The VGG-19 model can provide the highest classification accuracy rate, reaching 87.3%; the VGG-16 model has the highest classification accuracy rate of the ground in the mining area, reaching 95.2%. ResNet model has the best effect on road classification. Then, the lithology classification is applied based on Thermal Airborne Hyperspectral Imager (TASI) data. The noise level of the first 20 bands is comparatively stable; afterward, it increases exponentially, showing a higher noise level, and the spectrum curve of the data after denoising becomes smoother. The end-member extraction method is employed to extract 25 end-member spectra of almost all lithology in the research area from the image. The similarity coefficient clustering analysis is employed to group the curves, which are divided into six categories in total. The separability of similar categories can be constrained by the objective function using the dictionary learning method, and the accuracy of the sparse representation of the category spectrum can be improved. The spectral matching method is used to subdivide each group of mapping results, suggesting that in the research area, granite is the most widely distributed, followed by diorite, andesite, and quartzite. Deep learning algorithms are applied to extract ground feature information, which is of great significance to the safety production in the mining area. The hyperspectral remote sensing rock and mineral thematic information extraction module is developed, which preliminarily realizes the quantitative acquisition and high-precision identification of typical mineral information, and provides technical support for the research of remote sensing geological evaluation technology of resource exploration in the new era.     

Automating land cover mapping of Scotland using expert system and knowledge integration methods

Effective land cover mapping often requires the use of multiple data sources and data interpretation methods, particularly when no one data source or interpretation method provides sufficiently good results. Method-oriented approaches are often only effective for specific land cover class/data source combinations, and cannot be applied when different classification systems or data sources are required or available. Here we present a method, based on Endorsement Theory, of pooling evidence from multiple expert systems and spatial datasets to produce land cover maps. Individual ‘experts’ are trained to produce evidence for or against a class, with this evidence being categorised according to strength. An evidence integration rule set is applied to evidence lists to produce conclusions of different strength regarding individual classes, and the most likely class identified. The only expert system design implemented currently within the methodology is a neural network model, although the system has been designed to accept information from decision trees, fuzzy k-means and Bayesian statistics as well. We have used the technique to produce land cover maps of Scotland using three classification systems of varying complexity. Mapping accuracy varied between 52.6% for a map with 96 classes to 88.8% for a map with eight classes. The accuracy of the maps generated is higher than when individual datasets are used, showing that the evidence integration method applied is suitable for improving land cover mapping accuracy. We showed that imagery was not necessarily the most important data source for mapping where a large number of classes are used, and also showed that even data sources that produce low accuracy scores when used for mapping by themselves do improve the accuracy of maps produced using this integrative approach. Future work in developing the method is identified, including the inclusion of additional expert systems and improvement of the evidence integration, and evaluation carried out of the overall effectiveness of the approach.     

A hybrid of deep learning and hand-crafted features based approach for snow cover mapping

Monitoring the extent of snow cover plays a vital role for a better understanding of current and future climatic, ecological, and water cycle conditions. Previously, several traditional machine learning models have been applied for accomplishing this while exploring a variety of feature extraction techniques on various information sources. However, the laborious process of any amount of hand-crafted feature extraction has not helped to obtain high accuracies. Recently, deep learning models have shown that feature extraction can be made automatic and that they can achieve the required high accuracies but at the cost of requiring a large amount of labelled data. Fortunately, despite the absence of such large amounts of labelled data for this task, we can rely on pre-trained models, which accept red-green-blue (RGB) information (or dimensions-reduced spectral data). However, it is always better to include a variety of information sources to solve any problem, especially with the availability of other important information sources like synthetic aperture radar (SAR) imagery and elevation. We propose a hybrid model where the deep learning is assisted by these information sources which have until now been left out. Particularly, our model learns from both the deep learning features (derived from spectral data) and the hand-crafted features (derived from SAR and elevation). Such an approach shows interesting performance-improvement from 96.02% (through deep learning alone) to 98.10% when experiments were conducted for Khiroi village of the Himalayan region in India.