Showing papers in "Isprs Journal of Photogrammetry and Remote Sensing in 2015"
TL;DR: In this article, an approach based on the integration of pixel-and object-based methods with knowledge (POK-based) has been developed to handle the classification process of 10 land cover types, i.e., firstly each class identified in a prioritized sequence and then results are merged together.
Abstract: Global Land Cover (GLC) information is fundamental for environmental change studies, land resource management, sustainable development, and many other societal benefits. Although GLC data exists at spatial resolutions of 300 m and 1000 m, a 30 m resolution mapping approach is now a feasible option for the next generation of GLC products. Since most significant human impacts on the land system can be captured at this scale, a number of researchers are focusing on such products. This paper reports the operational approach used in such a project, which aims to deliver reliable data products. Over 10,000 Landsat-like satellite images are required to cover the entire Earth at 30 m resolution. To derive a GLC map from such a large volume of data necessitates the development of effective, efficient, economic and operational approaches. Automated approaches usually provide higher efficiency and thus more economic solutions, yet existing automated classification has been deemed ineffective because of the low classification accuracy achievable (typically below 65%) at global scale at 30 m resolution. As a result, an approach based on the integration of pixel- and object-based methods with knowledge (POK-based) has been developed. To handle the classification process of 10 land cover types, a split-and-merge strategy was employed, i.e. firstly each class identified in a prioritized sequence and then results are merged together. For the identification of each class, a robust integration of pixel-and object-based classification was developed. To improve the quality of the classification results, a knowledge-based interactive verification procedure was developed with the support of web service technology. The performance of the POK-based approach was tested using eight selected areas with differing landscapes from five different continents. An overall classification accuracy of over 80% was achieved. This indicates that the developed POK-based approach is effective and feasible for operational GLC mapping at 30 m resolution.
1,260 citations
TL;DR: It is demonstrated that the selection of optimal neighborhoods for individual 3D points significantly improves the results of 3D scene analysis and may even further increase the quality of the derived results while significantly reducing both processing time and memory consumption.
Abstract: 3D scene analysis in terms of automatically assigning 3D points a respective semantic label has become a topic of great importance in photogrammetry, remote sensing, computer vision and robotics. In this paper, we address the issue of how to increase the distinctiveness of geometric features and select the most relevant ones among these for 3D scene analysis. We present a new, fully automated and versatile framework composed of four components: (i) neighborhood selection, (ii) feature extraction, (iii) feature selection and (iv) classification. For each component, we consider a variety of approaches which allow applicability in terms of simplicity, efficiency and reproducibility, so that end-users can easily apply the different components and do not require expert knowledge in the respective domains. In a detailed evaluation involving 7 neighborhood definitions, 21 geometric features, 7 approaches for feature selection, 10 classifiers and 2 benchmark datasets, we demonstrate that the selection of optimal neighborhoods for individual 3D points significantly improves the results of 3D scene analysis. Additionally, we show that the selection of adequate feature subsets may even further increase the quality of the derived results while significantly reducing both processing time and memory consumption.
513 citations
TL;DR: A review of state-of-the-art retrieval methods for quantitative terrestrial bio-geophysical variable extraction using optical remote sensing imagery and the prospects of implementing these methods into future processing chains for operational retrieval of vegetation properties are presented and discussed.
Abstract: Forthcoming superspectral satellite missions dedicated to land monitoring, as well as planned imaging spectrometers, will unleash an unprecedented data stream. The processing requirements for such large data streams involve processing techniques enabling the spatio-temporally explicit quantification of vegetation properties. Typically retrieval must be accurate, robust and fast. Hence, there is a strict requirement to identify next-generation bio-geophysical variable retrieval algorithms which can be molded into an operational processing chain. This paper offers a review of state-of-the-art retrieval methods for quantitative terrestrial bio-geophysical variable extraction using optical remote sensing imagery. We can categorize these methods into (1) parametric regression, (2) non-parametric regression, (3) physically-based and (4) hybrid methods. Hybrid methods combine generic capabilities of physically-based methods with flexible and computationally efficient methods, typically non-parametric regression methods. A review of the theoretical basis of all these methods is given first and followed by published applications. This paper focusses on: (1) retrievability of bio-geophysical variables, (2) ability to generate multiple outputs, (3) possibilities for model transparency description, (4) mapping speed, and (5) possibilities for uncertainty retrieval. Finally, the prospects of implementing these methods into future processing chains for operational retrieval of vegetation properties are presented and discussed.
471 citations
TL;DR: In this article, a very light plane (SwingletCam) equipped with a very cheap, non-metric camera was used to acquire images with ground resolutions better than 5 cm.
Abstract: Coastal areas suffer degradation due to the action of the sea and other natural and human-induced causes. Topographical changes in beaches and sand dunes need to be assessed, both after severe events and on a regular basis, to build models that can predict the evolution of these natural environments. This is an important application for airborne LIDAR, and conventional photogrammetry is also being used for regular monitoring programs of sensitive coastal areas. This paper analyses the use of unmanned aerial vehicles (UAV) to map and monitor sand dunes and beaches. A very light plane (SwingletCam) equipped with a very cheap, non-metric camera was used to acquire images with ground resolutions better than 5 cm. The Agisoft Photoscan software was used to orientate the images, extract point clouds, build a digital surface model and produce orthoimage mosaics. The processing, which includes automatic aerial triangulation with camera calibration and subsequent model generation, was mostly automated. To achieve the best positional accuracy for the whole process, signalised ground control points were surveyed with a differential GPS receiver. Two very sensitive test areas on the Portuguese northwest coast were analysed. Detailed DSMs were obtained with 10 cm grid spacing and vertical accuracy (RMS) ranging from 3.5 to 5.0 cm, which is very similar to the image ground resolution (3.2–4.5 cm). Where possible to assess, the planimetric accuracy of the orthoimage mosaics was found to be subpixel. Within the regular coastal monitoring programme being carried out in the region, UAVs can replace many of the conventional flights, with considerable gains in the cost of the data acquisition and without any loss in the quality of topographic and aerial imagery data.
465 citations
TL;DR: Empirical studies show the proposed approach to be at least an order of magnitude faster when compared to a conventional region growing method and able to incorporate semantic-based feature criteria, while achieving precision, recall, and fitness scores of at least 75% and as much as 95%.
Abstract: This paper introduces a novel, region-growing algorithm for the fast surface patch segmentation of three-dimensional point clouds of urban environments. The proposed algorithm is composed of two stages based on a coarse-to-fine concept. First, a region-growing step is performed on an octree-based voxelized representation of the input point cloud to extract major (coarse) segments. The output is then passed through a refinement process. As part of this, there are two competing factors related to voxel size selection. To balance the constraints, an adaptive octree is created in two stages. Empirical studies on real terrestrial and airborne laser scanning data for complex buildings and an urban setting show the proposed approach to be at least an order of magnitude faster when compared to a conventional region growing method and able to incorporate semantic-based feature criteria, while achieving precision, recall, and fitness scores of at least 75% and as much as 95%.
430 citations
TL;DR: A novel method to derive 3D hyperspectral information from lightweight snapshot cameras for unmanned aerial vehicles for vegetation monitoring and applies the approach to data from a flight campaign in a barley experiment to demonstrate the feasibility of vegetation monitoring in the context of precision agriculture.
Abstract: This paper describes a novel method to derive 3D hyperspectral information from lightweight snapshot cameras for unmanned aerial vehicles for vegetation monitoring. Snapshot cameras record an image cube with one spectral and two spatial dimensions with every exposure. First, we describe and apply methods to radiometrically characterize and calibrate these cameras. Then, we introduce our processing chain to derive 3D hyperspectral information from the calibrated image cubes based on structure from motion. The approach includes a novel way for quality assurance of the data which is used to assess the quality of the hyperspectral data for every single pixel in the final data product. The result is a hyperspectral digital surface model as a representation of the surface in 3D space linked with the hyperspectral information emitted and reflected by the objects covered by the surface. In this study we use the hyperspectral camera Cubert UHD 185-Firefly, which collects 125 bands from 450 to 950 nm. The obtained data product has a spatial resolution of approximately 1 cm for the spatial and 21 cm for the hyperspectral information. The radiometric calibration yields good results with less than 1% offset in reflectance compared to an ASD FieldSpec 3 for most of the spectral range. The quality assurance information shows that the radiometric precision is better than 0.13% for the derived data product. We apply the approach to data from a flight campaign in a barley experiment with different varieties during the growth stage heading (BBCH 52 – 59) to demonstrate the feasibility for vegetation monitoring in the context of precision agriculture. The plant parameters retrieved from the data product correspond to in-field measurements of a single date field campaign for plant height (R2 = 0.7), chlorophyll (BGI2, R2 = 0.52), LAI (RDVI, R2 = 0.32) and biomass (RDVI, R2 = 0.29). Our approach can also be applied for other image-frame cameras as long as the individual bands of the image cube are spatially co-registered beforehand.
376 citations
TL;DR: Experimental results with three Radarsat-2 images in quad polarization mode indicate that classification accuracies could be significantly increased by integrating spatial and polarimetric features using ensemble learning strategies.
Abstract: Fully Polarimetric Synthetic Aperture Radar (PolSAR) has the advantages of all-weather, day and night observation and high resolution capabilities. The collected data are usually sorted in Sinclair matrix, coherence or covariance matrices which are directly related to physical properties of natural media and backscattering mechanism. Additional information related to the nature of scattering medium can be exploited through polarimetric decomposition theorems. Accordingly, PolSAR image classification gains increasing attentions from remote sensing communities in recent years. However, the above polarimetric measurements or parameters cannot provide sufficient information for accurate PolSAR image classification in some scenarios, e.g. in complex urban areas where different scattering mediums may exhibit similar PolSAR response due to couples of unavoidable reasons. Inspired by the complementarity between spectral and spatial features bringing remarkable improvements in optical image classification, the complementary information between polarimetric and spatial features may also contribute to PolSAR image classification. Therefore, the roles of textural features such as contrast, dissimilarity, homogeneity and local range, morphological profiles (MPs) in PolSAR image classification are investigated using two advanced ensemble learning (EL) classifiers: Random Forest and Rotation Forest. Supervised Wishart classifier and support vector machines (SVMs) are used as benchmark classifiers for the evaluation and comparison purposes. Experimental results with three Radarsat-2 images in quad polarization mode indicate that classification accuracies could be significantly increased by integrating spatial and polarimetric features using ensemble learning strategies. Rotation Forest can get better accuracy than SVM and Random Forest, in the meantime, Random Forest is much faster than Rotation Forest.
340 citations
TL;DR: In this article, the authors present a compendium of satellites under civilian and/or commercial control with the potential to gather global land-cover observations and analyze the changes and shows how innovation, the need for secure data-supply, national pride, falling costs and technological advances may underpin the trends.
Abstract: This paper presents a compendium of satellites under civilian and/or commercial control with the potential to gather global land-cover observations. From this we show that a growing number of sovereign states are acquiring capacity for space based land-cover observations and show how geopolitical patterns of ownership are changing. We discuss how the number of satellites flying at any time has progressed as a function of increased launch rates and mission longevity, and how the spatial resolutions of the data they collect has evolved. The first such satellite was launched by the USA in 1972. Since then government and/or private entities in 33 other sovereign states and geopolitical groups have chosen to finance such missions and 197 individual satellites with a global land-cover observing capacity have been successfully launched. Of these 98 were still operating at the end of 2013. Since the 1970s the number of such missions failing within 3 years of launch has dropped from around 60% to less than 20%, the average operational life of a mission has almost tripled, increasing from 3.3 years in the 1970s to 8.6 years (and still lengthening), the average number of satellites launched per-year/per-decade has increased from 2 to 12 and spatial resolution increased from around 80 m to less than 1 m multispectral and less than half a meter for panchromatic; synthetic aperture radar resolution has also fallen, from 25 m in the 1970s to 1 m post 2007. More people in more countries have access to data from global land-cover observing spaceborne missions at a greater range of spatial resolutions than ever before. We provide a compendium of such missions, analyze the changes and shows how innovation, the need for secure data-supply, national pride, falling costs and technological advances may underpin the trends we document.
318 citations
TL;DR: In this paper, the utility of the newly-launched medium-resolution multispectral Landsat 8 Operational Land Imager (OLI) dataset with a large swath width, in quantifying aboveground biomass (AGB) in a forest plantation was assessed.
Abstract: Aboveground biomass estimation is critical in understanding forest contribution to regional carbon cycles. Despite the successful application of high spatial and spectral resolution sensors in aboveground biomass (AGB) estimation, there are challenges related to high acquisition costs, small area coverage, multicollinearity and limited availability. These challenges hamper the successful regional scale AGB quantification. The aim of this study was to assess the utility of the newly-launched medium-resolution multispectral Landsat 8 Operational Land Imager (OLI) dataset with a large swath width, in quantifying AGB in a forest plantation. We applied different sets of spectral analysis (test I: spectral bands; test II: spectral vegetation indices and test III: spectral bands + spectral vegetation indices) in testing the utility of Landsat 8 OLI using two non-parametric algorithms: stochastic gradient boosting and the random forest ensembles. The results of the study show that the medium-resolution multispectral Landsat 8 OLI dataset provides better AGB estimates for Eucalyptus dunii , Eucalyptus grandis and Pinus taeda especially when using the extracted spectral information together with the derived spectral vegetation indices. We also noted that incorporating the optimal subset of the most important selected medium-resolution multispectral Landsat 8 OLI bands improved AGB accuracies. We compared medium-resolution multispectral Landsat 8 OLI AGB estimates with Landsat 7 ETM + estimates and the latter yielded lower estimation accuracies. Overall, this study demonstrates the invaluable potential and strength of applying the relatively affordable and readily available newly-launched medium-resolution Landsat 8 OLI dataset, with a large swath width (185-km) in precisely estimating AGB. This strength of the Landsat OLI dataset is crucial especially in sub-Saharan Africa where high-resolution remote sensing data availability remains a challenge.
240 citations
TL;DR: A systematic comparison of retrieval accuracy and processing speed of a multitude of parametric, non-parametric and physically-based retrieval methods using simulated S2 data concludes that the family of kernel-based MLRAs (e.g. GPR) is the most promising processing approach.
Abstract: Given the forthcoming availability of Sentinel-2 (S2) images, this paper provides a systematic comparison of retrieval accuracy and processing speed of a multitude of parametric, non-parametric and physically-based retrieval methods using simulated S2 data. An experimental field dataset (SPARC), collected at the agricultural site of Barrax (Spain), was used to evaluate different retrieval methods on their ability to estimate leaf area index (LAI). With regard to parametric methods, all possible band combinations for several two-band and three-band index formulations and a linear regression fitting function have been evaluated. From a set of over ten thousand indices evaluated, the best performing one was an optimized three-band combination according to ( ρ 560 - ρ 1610 - ρ 2190 ) / ( ρ 560 + ρ 1610 + ρ 2190 ) with a 10-fold cross-validation R CV 2 of 0.82 ( RMSE CV : 0.62). This family of methods excel for their fast processing speed, e.g., 0.05 s to calibrate and validate the regression function, and 3.8 s to map a simulated S2 image. With regard to non-parametric methods, 11 machine learning regression algorithms (MLRAs) have been evaluated. This methodological family has the advantage of making use of the full optical spectrum as well as flexible, nonlinear fitting. Particularly kernel-based MLRAs lead to excellent results, with variational heteroscedastic (VH) Gaussian Processes regression (GPR) as the best performing method, with a R CV 2 of 0.90 ( RMSE CV : 0.44). Additionally, the model is trained and validated relatively fast (1.70 s) and the processed image (taking 73.88 s) includes associated uncertainty estimates. More challenging is the inversion of a PROSAIL based radiative transfer model (RTM). After the generation of a look-up table (LUT), a multitude of cost functions and regularization options were evaluated. The best performing cost function is Pearson’s χ -square. It led to a R 2 of 0.74 ( RMSE : 0.80) against the validation dataset. While its validation went fast (0.33 s), due to a per-pixel LUT solving using a cost function, image processing took considerably more time (01:01:47). Summarizing, when it comes to accurate and sufficiently fast processing of imagery to generate vegetation attributes, this paper concludes that the family of kernel-based MLRAs (e.g. GPR) is the most promising processing approach.
240 citations
TL;DR: In this paper, the authors explored the use of NDVI seasonal time-series derived from Landsat images across different seasons to estimate aboveground biomass (AGB) in southeast Ohio by six empirical modeling approaches.
Abstract: Spatially explicit knowledge of aboveground biomass (AGB) in large areas is important for accurate carbon accounting. Landsat data have been widely used to provide efficient and timely estimates of forest AGB because of their long archive and relatively high spatial resolution. Previous studies have explored different empirical modeling approaches to estimate AGB, but most of them only used a single Landsat image in the peak season, which may cause a saturation problem and low accuracy. To improve the accuracy of AGB estimation using Landsat images, this study explored the use of NDVI seasonal time-series derived from Landsat images across different seasons to estimate AGB in southeast Ohio by six empirical modeling approaches. Results clearly show that NDVI in the fall season has a stronger correlation to AGB than in the peak season, and using seasonal NDVI time-series can result in a more accurate AGB estimation and less saturation than using a single NDVI. In comparing these different empirical approaches, it is difficult to decide which one is superior to the other because they have different strengths and their accuracy is generally similar, indicating that modeling methods may not be the key issue for improving the accuracy of AGB estimation from Landsat data. This study suggests that future research should pay more attention to seasonal time-series data, and especially the data from the fall season.
TL;DR: In this paper, a fully automated processing chain for near real-time flood detection using high resolution TerraSAR-X Synthetic Aperture Radar (SAR) data is presented.
Abstract: In this paper, a fully automated processing chain for near real-time flood detection using high resolution TerraSAR-X Synthetic Aperture Radar (SAR) data is presented. The processing chain including SAR data pre-processing, computation and adaption of global auxiliary data, unsupervised initialization of the classification as well as post-classification refinement by using a fuzzy logic-based approach is automatically triggered after satellite data delivery. The dissemination of flood maps resulting from this service is performed through an online service which can be activated on-demand for emergency response purposes (i.e., when a flood situation evolves). The classification methodology is based on previous work of the authors but was substantially refined and extended for robustness and transferability to guarantee high classification accuracy under different environmental conditions and sensor configurations. With respect to accuracy and computational effort, experiments performed on a data set of 175 different TerraSAR-X scenes acquired during flooding all over the world with different sensor configurations confirm the robustness and effectiveness of the proposed flood mapping service. These promising results have been further confirmed by means of an in-depth validation performed for three study sites in Germany, Thailand, and Albania/Montenegro.
TL;DR: The proposed method is efficient and robust for extracting buildings, streetlamps, trees, telegraph poles, traffic signs, cars, and enclosures from mobile laser scanning (MLS) point clouds, with an overall accuracy of 92.3%.
Abstract: Point clouds collected in urban scenes contain a huge number of points (e.g., billions), numerous objects with significant size variability, complex and incomplete structures, and variable point densities, raising great challenges for the automated extraction of urban objects in the field of photogrammetry, computer vision, and robotics. This paper addresses these challenges by proposing an automated method to extract urban objects robustly and efficiently. The proposed method generates multi-scale supervoxels from 3D point clouds using the point attributes (e.g., colors, intensities) and spatial distances between points, and then segments the supervoxels rather than individual points by combining graph based segmentation with multiple cues (e.g., principal direction, colors) of the supervoxels. The proposed method defines a set of rules for merging segments into meaningful units according to types of urban objects and forms the semantic knowledge of urban objects for the classification of objects. Finally, the proposed method extracts and classifies urban objects in a hierarchical order ranked by the saliency of the segments. Experiments show that the proposed method is efficient and robust for extracting buildings, streetlamps, trees, telegraph poles, traffic signs, cars, and enclosures from mobile laser scanning (MLS) point clouds, with an overall accuracy of 92.3%.
TL;DR: This study demonstrated that the improved algorithm by using both thermal and optical MODIS data, provides a robust, simple and automated approach to identify and map paddy rice fields in temperate and cold temperate zones, the northern frontier of rice planting.
Abstract: Knowledge of the area and spatial distribution of paddy rice is important for assessment of food security, management of water resources, and estimation of greenhouse gas (methane) emissions. Paddy rice agriculture has expanded rapidly in northeastern China in the last decade, but there are no updated maps of paddy rice fields in the region. Existing algorithms for identifying paddy rice fields are based on the unique physical features of paddy rice during the flooding and transplanting phases and use vegetation indices that are sensitive to the dynamics of the canopy and surface water content. However, the flooding phenomena in high latitude area could also be from spring snowmelt flooding. We used land surface temperature (LST) data from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor to determine the temporal window of flooding and rice transplantation over a year to improve the existing phenology-based approach. Other land cover types (e.g., evergreen vegetation, permanent water bodies, and sparse vegetation) with potential influences on paddy rice identification were removed (masked out) due to their different temporal profiles. The accuracy assessment using high-resolution images showed that the resultant MODIS-derived paddy rice map of northeastern China in 2010 had a high accuracy (producer and user accuracies of 92% and 96%, respectively). The MODIS-based map also had a comparable accuracy to the 2010 Landsat-based National Land Cover Dataset (NLCD) of China in terms of both area and spatial pattern. This study demonstrated that our improved algorithm by using both thermal and optical MODIS data, provides a robust, simple and automated approach to identify and map paddy rice fields in temperate and cold temperate zones, the northern frontier of rice planting.
TL;DR: New ensemble margin criteria are introduced to evaluate the performance of Random Forests in the context of large area land cover classification and the effect of different training data characteristics (imbalance and mislabelling) on classification accuracy and uncertainty is examined.
Abstract: Studies have demonstrated the robust performance of the ensemble machine learning classifier, random forests, for remote sensing land cover classification, particularly across complex landscapes. This study introduces new ensemble margin criteria to evaluate the performance of Random Forests (RF) in the context of large area land cover classification and examines the effect of different training data characteristics (imbalance and mislabelling) on classification accuracy and uncertainty. The study presents a new margin weighted confusion matrix, which used in combination with the traditional confusion matrix, provides confidence estimates associated with correctly and misclassified instances in the RF classification model. Landsat TM satellite imagery, topographic and climate ancillary data are used to build binary (forest/non-forest) and multiclass (forest canopy cover classes) classification models, trained using sample aerial photograph maps, across Victoria, Australia. Experiments were undertaken to reveal insights into the behaviour of RF over large and complex data, in which training data are not evenly distributed among classes (imbalance) and contain systematically mislabelled instances. Results of experiments reveal that while the error rate of the RF classifier is relatively insensitive to mislabelled training data (in the multiclass experiment, overall 78.3% Kappa with no mislabelled instances to 70.1% with 25% mislabelling in each class), the level of associated confidence falls at a faster rate than overall accuracy with increasing amounts of mislabelled training data. In general, balanced training data resulted in the lowest overall error rates for classification experiments (82.3% and 78.3% for the binary and multiclass experiments respectively). However, results of the study demonstrate that imbalance can be introduced to improve error rates of more difficult classes, without adversely affecting overall classification accuracy.
TL;DR: A strategy for the semi-automatic optimization of object-based classification is developed, which involves an area-based accuracy assessment that analyzes the relationship between scale and the training set size and suggests that the optimal SSP for each class has a high positive correlation with the mean area obtained by manual interpretation.
Abstract: Unmanned Aerial Vehicle (UAV) has been used increasingly for natural resource applications in recent years due to their greater availability and the miniaturization of sensors. In addition, Geographic Object-Based Image Analysis (GEOBIA) has received more attention as a novel paradigm for remote sensing earth observation data. However, GEOBIA generates some new problems compared with pixel-based methods. In this study, we developed a strategy for the semi-automatic optimization of object-based classification, which involves an area-based accuracy assessment that analyzes the relationship between scale and the training set size. We found that the Overall Accuracy (OA) increased as the training set ratio (proportion of the segmented objects used for training) increased when the Segmentation Scale Parameter (SSP) was fixed. The OA increased more slowly as the training set ratio became larger and a similar rule was obtained according to the pixel-based image analysis. The OA decreased as the SSP increased when the training set ratio was fixed. Consequently, the SSP should not be too large during classification using a small training set ratio. By contrast, a large training set ratio is required if classification is performed using a high SSP. In addition, we suggest that the optimal SSP for each class has a high positive correlation with the mean area obtained by manual interpretation, which can be summarized by a linear correlation equation. We expect that these results will be applicable to UAV imagery classification to determine the optimal SSP for each class.
TL;DR: The contextual constraints for objects extracted by graph-cut segmentation are used to optimize the initial classification results obtained by the JointBoost classifier and indicate that the proposed features and method are effective for classification of airborne LiDAR data from complex scenarios.
Abstract: The demands for automatic point cloud classification have dramatically increased with the wide-spread use of airborne LiDAR. Existing research has mainly concentrated on a few dominant objects such as terrain, buildings and vegetation. In addition to those key objects, this paper proposes a supervised classification method to identify other types of objects including power-lines and pylons from point clouds using a JointBoost classifier. The parameters for the learning model are estimated with various features computed based on the geometry and echo information of a LiDAR point cloud. In order to overcome the shortcomings stemming from the inclusion of bare ground data before classification, the proposed classifier directly distinguishes terrain using a feature step-off count. Feature selection is conducted using JointBoost to evaluate feature correlations thus improving both classification accuracy and operational efficiency. In this paper, the contextual constraints for objects extracted by graph-cut segmentation are used to optimize the initial classification results obtained by the JointBoost classifier. Our experimental results show that the step-off count significantly contributes to classification. Seventeen effective features are selected for the initial classification results using the JointBoost classifier. Our experiments indicate that the proposed features and method are effective for classification of airborne LiDAR data from complex scenarios.
TL;DR: In this paper, satellite images from Thematic Mapper (TM) and enhanced thematic mapper (ETM) were used to quantify the spatio-temporal changes that took place in the coastal zone of Hatiya Island during the specified period.
Abstract: A large percentage of the world’s population is concentrated along the coastal zones. These environmentally sensitive areas are under intense pressure from natural processes such as erosion, accretion and natural disasters as well as anthropogenic processes such as urban growth, resource development and pollution. These threats have made the coastal zone a priority for coastline monitoring programs and sustainable coastal management. This research utilizes integrated techniques of remote sensing and geographic information system (GIS) to monitor coastline changes from 1989 to 2010 at Hatiya Island, Bangladesh. In this study, satellite images from Thematic Mapper (TM) and Enhanced Thematic Mapper (ETM) were used to quantify the spatio-temporal changes that took place in the coastal zone of Hatiya Island during the specified period. The modified normalized difference water index (MNDWI) algorithm was applied to TM (1989 and 2010) and ETM (2000) images to discriminate the land–water interface and the on-screen digitizing approach was used over the MNDWI images of 1989, 2000 and 2010 for coastline extraction. Afterwards, the extent of changes in the coastline was estimated through overlaying the digitized maps of Hatiya Island of all three years. Coastline positions were highlighted to infer the erosion/accretion sectors along the coast, and the coastline changes were calculated. The results showed that erosion was severe in the northern and western parts of the island, whereas the southern and eastern parts of the island gained land through sedimentation. Over the study period (1989–2010), this offshore island witnessed the erosion of 6476 hectares. In contrast it experienced an accretion of 9916 hectares. These erosion and accretion processes played an active role in the changes of coastline during the study period.
TL;DR: A methodology was developed to delineate buildings from a point cloud and classify the present gaps, and two learning algorithms – SVM and Random Forests were tested for mapping the damaged regions based on radiometric descriptors.
Abstract: Point clouds generated from airborne oblique images have become a suitable source for detailed building damage assessment after a disaster event, since they provide the essential geometric and radiometric features of both roof and facades of the building. However, they often contain gaps that result either from physical damage or from a range of image artefacts or data acquisition conditions. A clear understanding of those reasons, and accurate classification of gap-type, are critical for 3D geometry-based damage assessment. In this study, a methodology was developed to delineate buildings from a point cloud and classify the present gaps. The building delineation process was carried out by identifying and merging the roof segments of single buildings from the pre-segmented 3D point cloud. This approach detected 96% of the buildings from a point cloud generated using airborne oblique images. The gap detection and classification methods were tested using two other data sets obtained with Unmanned Aerial Vehicle (UAV) images with a ground resolution of around 1–2 cm. The methods detected all significant gaps and correctly identified the gaps due to damage. The gaps due to damage were identified based on the surrounding damage pattern, applying Gabor wavelets and a histogram of gradient orientation features. Two learning algorithms – SVM and Random Forests were tested for mapping the damaged regions based on radiometric descriptors. The learning model based on Gabor features with Random Forests performed best, identifying 95% of the damaged regions. The generalization performance of the supervised model, however, was less successful: quality measures decreased by around 15–30%.
TL;DR: In this paper, the authors present a global land cover mapping using earth observation satellite data: recent progresses and challenges, recent progress, and challenges of using satellite data for land cover analysis.
Abstract: Global land cover mapping using earth observation satellite data : recent progresses and challenges
TL;DR: In this article, the authors used hyperspectral data from the recently launched Sentinel 2 Multispectral Imager (MSI) and Landsat 8 OLI for comparison purposes.
Abstract: The major constraint in understanding grass above ground biomass variations using remotely sensed data are the expenses associated with the data, as well as the limited number of techniques that can be applied to different management practices with minimal errors. New generation multispectral sensors such as Sentinel 2 Multispectral Imager (MSI) are promising for effective rangeland management due to their unique spectral bands and higher signal to noise ratio. This study resampled hyperspectral data to spectral resolutions of the newly launched Sentinel 2 MSI and the recently launched Landsat 8 OLI for comparison purposes. Using Sparse partial least squares regression, the resampled data was applied in estimating above ground biomass of grasses treated with different fertilizer combinations of ammonium sulfate, ammonium nitrate, phosphorus and lime as well as unfertilized experimental plots. Sentinel 2 MSI derived models satisfactorily performed (R2 = 0.81, RMSEP = 1.07 kg/m2, RMSEP_rel = 14.97) in estimating grass above ground biomass across different fertilizer treatments relative to Landsat 8 OLI (Landsat 8 OLI: R2 = 0.76, RMSEP = 1.15 kg/m2, RMSEP_rel = 16.04). In comparison, hyperspectral data derived models exhibited better grass above ground biomass estimation across complex fertilizer combinations (R2 = 0.92, RMSEP = 0.69 kg/m2, RMSEP_rel = 9.61). Although Sentinel 2 MSI bands and indices better predicted above ground biomass compared with Landsat 8 OLI bands and indices, there were no significant differences (α = 0.05) in the errors of prediction between the two new generational sensors across all fertilizer treatments. The findings of this study portrays Sentinel 2 MSI and Landsat 8 OLI as promising remotely sensed datasets for regional scale biomass estimation, particularly in resource scarce areas.
TL;DR: In this paper, the authors focus on measuring the ground displacements due to seismotectonic events using three sub-pixel correlators: COSI-Corr, a free, closed-source correlator, dependent on commercial software (ENVI).
Abstract: Image correlation is one of the most efficient techniques to determine horizontal ground displacements due to earthquakes, landslides, ice flows or sand dune migrations. Analyzing these deformations allows a better understanding of the causes and mechanisms of the events. By using sub-pixel correlation on before- and after-event ortho-images obtained from high resolution satellite images it is possible to compute the displacement field with high planimetric resolution. In this paper, we focus on measuring the ground displacements due to seismotectonic events. The three sub-pixel correlators used are: COSI-Corr – developed by Caltech, a free, closed-source correlator, dependent on commercial software (ENVI) and widely used by the geoscience community for measuring ground displacement; Medicis – developed by CNES, also a closed-source correlator capable of measuring this type of deformation; and MicMac – developed by IGN, the free open-source correlator we study and tune for measuring fine ground displacements. We measured horizontal ground deformation using these three correlators on SPOT images in three study cases: the 2001 Kokoxili earthquake, the 2005 dyke intrusion in the Afar depression and the 2008 Yutian earthquake.
TL;DR: It was shown that interpolation of the knowledge-based features increases the stability of the classifier allowing for its re-use from year to year without recalibration, and shows potential for application at larger scale as well as for delivering cropland map in near real time.
Abstract: Global, timely, accurate and cost-effective cropland mapping is a prerequisite for reliable crop condition monitoring. This article presented a simple and comprehensive methodology capable to meet the requirements of operational cropland mapping by proposing (1) five knowledge-based temporal features that remain stable over time, (2) a cleaning method that discards misleading pixels from a baseline land cover map and (3) a classifier that delivers high accuracy cropland maps (>80%). This was demonstrated over four contrasted agrosystems in Argentina, Belgium, China and Ukraine. It was found that the quality and accuracy of the baseline impact more the certainty of the classification rather than the classification output itself. In addition, it was shown that interpolation of the knowledge-based features increases the stability of the classifier allowing for its re-use from year to year without recalibration. Hence, the method shows potential for application at larger scale as well as for delivering cropland map in near real time.
TL;DR: In this article, the authors investigated the potential of Random Forest (RF), a machine learning technique, to estimate LiDAR measured canopy structure using a time series of Landsat imagery and demonstrated the value of using disturbance and successional history to inform estimates of canopy structure and obtain improved estimates of forest canopy cover and height using the RF algorithm.
Abstract: Many forest management activities, including the development of forest inventories, require spatially detailed forest canopy cover and height data. Among the various remote sensing technologies, LiDAR (Light Detection and Ranging) offers the most accurate and consistent means for obtaining reliable canopy structure measurements. A potential solution to reduce the cost of LiDAR data, is to integrate transects (samples) of LiDAR data with frequently acquired and spatially comprehensive optical remotely sensed data. Although multiple regression is commonly used for such modeling, often it does not fully capture the complex relationships between forest structure variables. This study investigates the potential of Random Forest (RF), a machine learning technique, to estimate LiDAR measured canopy structure using a time series of Landsat imagery. The study is implemented over a 2600 ha area of industrially managed coastal temperate forests on Vancouver Island, British Columbia, Canada. We implemented a trajectory-based approach to time series analysis that generates time since disturbance (TSD) and disturbance intensity information for each pixel and we used this information to stratify the forest land base into two strata: mature forests and young forests. Canopy cover and height for three forest classes (i.e. mature, young and mature and young (combined)) were modeled separately using multiple regression and Random Forest (RF) techniques. For all forest classes, the RF models provided improved estimates relative to the multiple regression models. The lowest validation error was obtained for the mature forest strata in a RF model ( R 2 = 0.88, RMSE = 2.39 m and bias = −0.16 for canopy height; R 2 = 0.72, RMSE = 0.068% and bias = −0.0049 for canopy cover). This study demonstrates the value of using disturbance and successional history to inform estimates of canopy structure and obtain improved estimates of forest canopy cover and height using the RF algorithm.
TL;DR: A comparative shortest-path algorithm (CSP) for segmenting tree crowns scanned using terrestrial (T)-LiDAR and mobile LiDAR, inspired by the well-proved metabolic ecology theory and the ecological fact that vascular plants tend to minimize the transferring distance to the root.
Abstract: The rapid development of light detection and ranging (LiDAR) techniques is advancing ecological and forest research. During the last decade, numerous single tree segmentation techniques have been developed using airborne LiDAR data. However, accurate crown segmentation using terrestrial or mobile LiDAR data, which is an essential prerequisite for extracting branch level forest characteristics, is still challenging mainly because of the difficulties posed by tree crown intersection and irregular crown shape. In the current work, we developed a comparative shortest-path algorithm (CSP) for segmenting tree crowns scanned using terrestrial (T)-LiDAR and mobile LiDAR. The algorithm consists of two steps, namely trunk detection and subsequent crown segmentation, with the latter inspired by the well-proved metabolic ecology theory and the ecological fact that vascular plants tend to minimize the transferring distance to the root. We tested the algorithm on mobile-LiDAR-scanned roadside trees and T-LiDAR-scanned broadleaved and coniferous forests in China. Point-level quantitative assessments of the segmentation results showed that for mobile-LiDAR-scanned roadside trees, all the points were classified to their corresponding trees correctly, and for T-LiDAR-scanned broadleaved and coniferous forests, kappa coefficients ranging from 0.83 to 0.93 were obtained. We believe that our algorithm will make a contribution to solving the problem of crown segmentation in T-LiDAR scanned-forests, and might be of interest to researchers in LiDAR data processing and to forest ecologists. In addition, our research highlights the advantages of using ecological theories as guidelines for processing LiDAR data.
TL;DR: In this article, a spatial and spectral statistics-based scale parameter selection method for object-based information extraction from high spatial resolution remote sensing images is proposed, which uses the ALV graph to replace the semivariogram to pre-estimate the optimal spatial bandwidth.
Abstract: Geo-Object-Based Image Analysis (GEOBIA) is becoming an increasingly important technology for information extraction from remote sensing images. Multi-scale image segmentation is a key procedure that partitions an image into homogeneous parcels (image objects) in GEOBIA. Hierarchical image objects also provide a better representation result than a single-scale representation. However, scale selection in multi-scale image segmentation is always difficult for high-performance GEOBIA. This paper first generalizes the commonly used segmentation scale parameters into three aspects: spatial bandwidth (spatial distance between classes), attribute bandwidth (difference between classes) and merging threshold. Next, taking mean-shift multi-scale segmentation as an example, this paper proposes a spatial and spectral statistics-based scale parameter selection method for object-based information extraction from high spatial resolution remote sensing images. The main idea of this proposed method is to use the ALV graph to replace the semivariogram to pre-estimate the optimal spatial bandwidth. Next, the selection of the optimal attribute bandwidth and the merging threshold are based on the ALV histogram and simple geometric computation, respectively. This study uses Ikonos, Quickbird and aerial panchromatic images as the experimental data to verify the validity of the proposed scale parameter selection method. Experiments based on quantitative multi-scale segmentation evaluation testify to the validity of this method. This pre-estimation-based scale parameter selection method is practically helpful and efficient in GEOBIA. The idea of this method can be further extended to other segmentation algorithms and other sensor data.
TL;DR: The proposed approach to semantically classify buildings into much finer categories by learning random forest (RF) classifier from a large number of imbalanced samples with high-dimensional features is effective and accurate.
Abstract: While most existing studies have focused on extracting geometric information on buildings, only a few have concentrated on semantic information. The lack of semantic information cannot satisfy many demands on resolving environmental and social issues. This study presents an approach to semantically classify buildings into much finer categories than those of existing studies by learning random forest (RF) classifier from a large number of imbalanced samples with high-dimensional features. First, a two-level segmentation mechanism combining GIS and VHR image produces single image objects at a large scale and intra-object components at a small scale. Second, a semi-supervised method chooses a large number of unbiased samples by considering the spatial proximity and intra-cluster similarity of buildings. Third, two important improvements in RF classifier are made: a voting-distribution ranked rule for reducing the influences of imbalanced samples on classification accuracy and a feature importance measurement for evaluating each feature’s contribution to the recognition of each category. Fourth, the semantic classification of urban buildings is practically conducted in Beijing city, and the results demonstrate that the proposed approach is effective and accurate. The seven categories used in the study are finer than those in existing work and more helpful to studying many environmental and social problems.
International Institute of Minnesota1, Polish Academy of Sciences2, Lviv Polytechnic3, University of Leicester4, University of Freiburg5, Tsinghua University6, National Institute for Space Research7, University of the Sciences8, University of Vienna9, University of Modena and Reggio Emilia10, Lamont–Doherty Earth Observatory11, National Academy of Sciences of Ukraine12, International Trademark Association13, Damanhour University14, Gauhati University15
TL;DR: In this article, the authors used geographically weighted regression (GWR) and crowdsourced validation data from Geo-Wiki to create two hybrid global land cover maps that use medium resolution land cover products as an input.
Abstract: Land cover is of fundamental importance to many environmental applications and serves as critical baseline information for many large scale models e.g. in developing future scenarios of land use and climate change. Although there is an ongoing movement towards the development of higher resolution global land cover maps, medium resolution land cover products (e.g. GLC2000 and MODIS) are still very useful for modelling and assessment purposes. However, the current land cover products are not accurate enough for many applications so we need to develop approaches that can take existing land covers maps and produce a better overall product in a hybrid approach. This paper uses geographically weighted regression (GWR) and crowdsourced validation data from Geo-Wiki to create two hybrid global land cover maps that use medium resolution land cover products as an input. Two different methods were used: (a) the GWR was used to determine the best land cover product at each location; (b) the GWR was only used to determine the best land cover at those locations where all three land cover maps disagree, using the agreement of the land cover maps to determine land cover at the other cells. The results show that the hybrid land cover map developed using the first method resulted in a lower overall disagreement than the individual global land cover maps. The hybrid map produced by the second method was also better when compared to the GLC2000 and GlobCover but worse or similar in performance to the MODIS land cover product depending upon the metrics considered. The reason for this may be due to the use of the GLC2000 in the development of GlobCover, which may have resulted in areas where both maps agree with one another but not with MODIS, and where MODIS may in fact better represent land cover in those situations. These results serve to demonstrate that spatial analysis methods can be used to improve medium resolution global land cover information with existing products.
TL;DR: In this paper, the strength and performance of Landsat-8 image derived texture metrics (i.e. texture measures and texture ratios) in estimating plantation forest species AGB was investigated.
Abstract: The successful launch of the 30-m Landsat-8 Operational Land Imager (OLI) pushbroom sensor offers a new primary data source necessary for aboveground biomass (AGB) estimation, especially in resource-limited environments. In this work, the strength and performance of Landsat-8 OLI image derived texture metrics (i.e. texture measures and texture ratios) in estimating plantation forest species AGB was investigated. It was hypothesized that the sensor’s pushbroom design, coupled with the presence of refined spectral properties, enhanced radiometric resolution (i.e. from 8 bits to 12 bits) and improved signal-to-noise ratio have the potential to provide detailed spectral information necessary for significantly strengthening AGB estimation in medium-density forest canopies. The relationship between image texture metrics and measurements of forest attributes can be used to help characterize complex forests, and enhance fine vegetation biophysical properties, a difficult challenge when using spectral vegetation indices especially in closed canopies. This study examines the prospects of using Landsat-8 OLI sensor derived texture metrics for estimating AGB for three medium-density plantation forest species in KwaZulu Natal, South Africa. In order to achieve this objective, three unique data pre-processing techniques were tested (analysis I: Landsat-8 OLI raw spectral-bands vs. raw texture bands; analysis II: Landsat-8 OLI raw spectral-band ratios vs. texture band ratios and analysis III: Landsat-8 OLI derived vegetation indices vs. texture band ratios). The landsat-8 OLI derived texture parameters were examined for robustness in estimating AGB using linear regression, stepwise-multiple linear regression and stochastic gradient boosting regression models. The results of this study demonstrated that all texture parameters particularly band texture ratios calculated using a 3 × 3 window size, could enhance AGB estimation when compared to simple spectral reflectance, simple band ratios and the most popular spectral vegetation indices. For instance, the use of combined texture ratios yielded the highest R 2 values of 0.76 (RMSE = 9.55 t ha −1 (18.07%) and CV-RMSE of 0.18); 0.74 (RMSE = 12.81 t ha −1 (17.72%) and CV-RMSE of 0.08); 0.74 (RMSE = 12.67 t ha −1 (06.15%) and CV-RMSE of 0.06) and 0.53 (RMSE = 20.15 t ha −1 (14.40%) and CV-RMSE of 0.15) overall for Eucalyptus dunii , Eucalyptus grandis , Pinus taeda individually and all species, respectively. Overall, the findings of this study provide the necessary insight and motivation to the remote sensing community, particularly in resource constrained regions, to shift towards embracing various texture metrics obtained from the readily-available and cheap multispectral Landsat-8 OLI sensor.
TL;DR: This study proposes region-based precision and recall measures and uses them to compare two image partitions for the purpose of evaluating segmentation quality, and suggests using a combination of the region- based precision–recall curve and the F-measure for supervised segmentation evaluation.
Abstract: Segmentation of remote sensing images is a critical step in geographic object-based image analysis. Evaluating the performance of segmentation algorithms is essential to identify effective segmentation methods and optimize their parameters. In this study, we propose region-based precision and recall measures and use them to compare two image partitions for the purpose of evaluating segmentation quality. The two measures are calculated based on region overlapping and presented as a point or a curve in a precision–recall space, which can indicate segmentation quality in both geometric and arithmetic respects. Furthermore, the precision and recall measures are combined by using four different methods. We examine and compare the effectiveness of the combined indicators through geometric illustration, in an effort to reveal segmentation quality clearly and capture the trade-off between the two measures. In the experiments, we adopted the multiresolution segmentation (MRS) method for evaluation. The proposed measures are compared with four existing discrepancy measures to further confirm their capabilities. Finally, we suggest using a combination of the region-based precision–recall curve and the F-measure for supervised segmentation evaluation.