Estimating fractional land cover in semi-arid central Kalahari: the impact of mapping method (spectral unmixing vs. object-based image analysis) and vegetation morphology

Focusing on the central Kalahari, this study utilized fractional cover of photosynthetic vegetation (f_PV), non-photosynthetic vegetation (f_NPV) and bare soil (f_BS), derived in situ and estimated from GeoEye-1 imagery using Multiple Endmember Spectral Mixture Analysis (MESMA) and object-based image analysis (OBIA) to determine superior method for fractional cover estimation and the impact of vegetation morphology on the estimation accuracy. MESMA mapped fractional cover by testing endmember models of varying complexity. Based on OBIA, image was segmented at five segmentation scales followed by classification. MESMA provided more accurate fractional cover estimates than OBIA. The increasing segmentation scale in OBIA resulted in a consistent increase in error. Different vegetation morphology types showed varied responses to the changing segmentation scale, reflecting their unique ecology and physiognomy. While areas under woody cover produced lower error even at coarse segmentation scales, those with herbaceous cover provided low error only at the fine segmentation scale.     

Assessment of the image misregistration effects on object-based change detection

High-spatial resolution remote sensing imagery provides unique opportunities for detailed characterization and monitoring of landscape dynamics. To better handle such data sets, change detection using the object-based paradigm, i.e., object-based change detection (OBCD), have demonstrated improved performances over the classic pixel-based paradigm. However, image registration remains a critical pre-process, with new challenges arising, because objects in OBCD are of various sizes and shapes. In this study, we quantified the effects of misregistration on OBCD using high-spatial resolution SPOT 5 imagery (5 m) for three types of landscapes dominated by urban, suburban and rural features, representing diverse geographic objects. The experiments were conducted in four steps: (i) Images were purposely shifted to simulate the misregistration effect. (ii) Image differencing change detection was employed to generate difference images with all the image-objects projected to a feature space consisting of both spectral and texture variables. (iii) The changes were extracted using the Mahalanobis distance and a change ratio. (iv) The results were compared to the ‘real’ changes from the image pairs that contained no purposely introduced registration error. A pixel-based change detection method using similar steps was also developed for comparisons. Results indicate that misregistration had a relatively low impact on object size and shape for most areas. When the landscape is comprised of small mean object sizes (e.g., in urban and suburban areas), the mean size of ‘change’ objects was smaller than the mean of all objects and their size discrepancy became larger with the decrease in object size. Compared to the results using the pixel-based paradigm, OBCD was less sensitive to the misregistration effect, and the sensitivity further decreased with an increase in local mean object size. However, high-spatial resolution images typically have higher spectral variability within neighboring pixels than the relatively low resolution datasets. As a result, accurate image registration remains crucial to change detection even if an object-based approach is used.     

Pixel-based and object-based classifications using high- and medium-spatial-resolution imageries in the urban and suburban landscapes

With the increasing availability of high-spatial-resolution remote sensing imageries and with the observed limitations of pixel-based techniques, the development and testing of geographic object-based image analysis (GEOBIA) techniques for image classification have become one of the main research areas in geospatial science. This paper examines and compares the classification performance of a pixel-based method and an object-based method as applied to high- (QuickBird satellite image) and medium- (Landsat TM image) spatial-resolution imageries in the context of urban and suburban landscapes. For the pixel-based classification, the maximum-likelihood supervised classification approach was employed. And for the object-based classification, the pixel-based classified maps were integrated with a set of image segments produced using various calibrations. The results show evidence that the object-based method can produce classifications that are more accurate for both high- and medium-spatial- resolution imageries in the context of urban and suburban landscapes.     

A method for monitoring building construction in urban sprawl areas using object-based analysis of Spot 5 images and existing GIS data

Urban sprawl has been identified as one of the most negative effects of global population growth on the environment and biodiversity. Frequent monitoring of urban sprawl is needed to limit the impact of this ongoing phenomenon. This paper proposes precise monitoring of building construction using an object-based classification methodology applied to Spot 5 images with a 2.5 m resolution. An application at a regional scale on Reunion Island in the Indian Ocean shows that this building extraction methodology has limitations in the production of reference urban maps because of difficulties in defining the shape and the number of buildings compared to classical photo-interpretation of aerial photography. However, these results are of great value for planning in urban sprawl areas where up-to-date information is lacking because of the rapid pace of house construction and residential development.     

Assessment of Nigeriasat-1 satellite data for urban land use/land cover analysis using object-based image analysis in Abuja,Nigeria

This study assesses the usefulness of Nigeriasat-1 satellite data for urban land cover analysis by comparing it with Landsat and SPOT data. The data-sets for Abuja were classified with pixel- and object-based methods. While the pixel-based method was classified with the spectral properties of the images, the object-based approach included an extra layer of land use cadastre data. The classification accuracy results for OBIA show that Landsat 7 ETM, Nigeriasat-1 SLIM and SPOT 5 HRG had overall accuracies of 92, 89 and 96%, respectively, while the classification accuracy for pixel-based classification were 88% for Landsat 7 ETM, 63% for Nigeriasat-1 SLIM and 89% for SPOT 5 HRG. The results indicate that given the right classification tools, the analysis of Nigeriasat-1 data can be compared with Landsat and SPOT data which are widely used for urban land use and land cover analysis.     

A multi-level context-guided classification method with object-based convolutional neural network for land cover classification using very high resolution remote sensing images

Classification of very high resolution imagery (VHRI) is challenging due to the difficulty in mining complex spatial and spectral patterns from rich image details. Various object-based Convolutional Neural Networks (OCNN) for VHRI classification have been proposed to overcome the drawbacks of the redundant pixel-wise CNNs, owing to their low computational cost and fine contour-preserving. However, classification performance of OCNN is still limited by geometric distortions, insufficient feature representation, and lack of contextual guidance. In this paper, an innovative multi-level context-guided classification method with the OCNN (MLCG-OCNN) is proposed. A feature-fusing OCNN, including the object contour-preserving mask strategy with the supplement of object deformation coefficient, is developed for accurate object discrimination by learning simultaneously high-level features from independent spectral patterns, geometric characteristics, and object-level contextual information. Then pixel-level contextual guidance is used to further improve the per-object classification results. The MLCG-OCNN method is intentionally tested on two validated small image datasets with limited training samples, to assess the performance in applications of land cover classification where a trade-off between time-consumption of sample training and overall accuracy needs to be found, as it is very common in the practice. Compared with traditional benchmark methods including the patch-based per-pixel CNN (PBPP), the patch-based per-object CNN (PBPO), the pixel-wise CNN with object segmentation refinement (PO), semantic segmentation U-Net (U-NET), and DeepLabV3+(DLV3+), MLCG-OCNN method achieves remarkable classification performance (> 80 %). Compared with the state-of-the-art architecture DeepLabV3+, the MLCG-OCNN method demonstrates high computational efficiency for VHRI classification (4–5 times faster).     

A combined spectral and object-based approach to transparent cloud removal in an operational setting for Landsat ETM+

The automated cloud cover assessment (ACCA) algorithm has provided automated estimates of cloud cover for the Landsat ETM+ mission since 2001. However, due to the lack of a band around 1.375 μm, cloud edges and transparent clouds such as cirrus cannot be detected. Use of Landsat ETM+ imagery for terrestrial land analysis is further hampered by the relatively long revisit period due to a nadir only viewing sensor. In this study, the ACCA threshold parameters were altered to minimise omission errors in the cloud masks. Object-based analysis was used to reduce the commission errors from the extended cloud filters. The method resulted in the removal of optically thin cirrus cloud and cloud edges which are often missed by other methods in sub-tropical areas. Although not fully automated, the principles of the method developed here provide an opportunity for using otherwise sub-optimal or completely unusable Landsat ETM+ imagery for operational applications. Where specific images are required for particular research goals the method can be used to remove cloud and transparent cloud helping to reduce bias in subsequent land cover classifications.     

Combining spatiotemporal fusion and object-based image analysis for improving wetland mapping in complex and heterogeneous urban landscapes

Remote sensing has been proven promising in wetland mapping. However, conventional methods in a complex and heterogeneous urban landscape usually use mono temporal Landsat TM/ETM + images, which have great uncertainty due to the spectral similarity of different land covers, and pixel-based classifications may not meet the accuracy requirement. This paper proposes an approach that combines spatiotemporal fusion and object-based image analysis, using the spatial and temporal adaptive reflectance fusion model to generate a time series of Landsat 8 OLI images on critical dates of sedge swamp and paddy rice, and the time series of MODIS NDVI to calculate phenological parameters for identifying wetlands with an object-based method. The results of a case study indicate that different types of wetlands can be successfully identified, with 92.38%. The overall accuracy and 0.85 Kappa coefficient, and 85% and 90% for the user’s accuracies of sedge swamp and paddy respectively.     

Assessment of SPOT 5 and QuickBird remotely sensed imagery for mapping tree cover in savannas

The relative abundance and distribution of trees in savannas has important implications for ecosystem function. High spatial resolution satellite sensors, including QuickBird and IKONOS, have been successfully used to map tree cover patterns in savannas. SPOT 5, with a 2.5 m panchromatic band and 10 m multispectral bands, represents a relatively coarse resolution sensor within this context, but has the advantage of being relatively inexpensive and more widely available. This study evaluates the performance of NDVI threshold and object based image analysis techniques for mapping tree canopies from QuickBird and SPOT 5 imagery in two savanna systems in southern Africa. High thematic mapping accuracies were obtained with the QuickBird imagery, independent of mapping technique. Geometric properties of the mapping indicated that the NDVI threshold produced smaller patch sizes, but that overall patch size distributions were similar. Tree canopy mapping using SPOT 5 imagery and an NDVI threshold approach performed poorly, however acceptable thematic accuracies were obtained from the object based image analysis. Although patch sizes were generally larger than those mapped from the QuickBird image data, patch size distributions mapped with object based image analysis of SPOT 5 have a similar form to the QuickBird mapping. This indicates that SPOT 5 imagery is suitable for regional studies of tree canopy cover patterns.     

Coupling high-resolution satellite imagery with ALS-based canopy height model and digital elevation model in object-based boreal forest habitat type classification

We developed a classification workflow for boreal forest habitat type mapping. In object-based image analysis framework, Fractal Net Evolution Approach segmentation was combined with random forest classification. High-resolution WorldView-2 imagery was coupled with ALS based canopy height model and digital terrain model. We calculated several features (e.g. spectral, textural and topographic) per image object from the used datasets. We tested different feature set alternatives; a classification accuracy of 78.0% was obtained when all features were used. The highest classification accuracy (79.1%) was obtained when the amount of features was reduced from the initial 328 to the 100 most important using Boruta feature selection algorithm and when ancillary soil and land-use GIS-datasets were used. Although Boruta could rank the importance of features, it could not separate unimportant features from the important ones. Classification accuracy was bit lower (78.7%) when the classification was performed separately on two areas: the areas above and below 1 m vertical distance from the nearest stream. The data split, however, improved the classification accuracy of mire habitat types and streamside habitats, probably because their proportion in the below 1 m data was higher than in the other datasets. It was found that several types of data are needed to get the highest classification accuracy whereas omitting some feature groups reduced the classification accuracy. A major habitat type in the study area was mesic forests in different successional stages. It was found that the inner heterogeneity of different mesic forest age groups was large and other habitat types were often inside this heterogeneity.     

Evaluation of semivariogram features for object-based image classification

Inclusion of textures in image classification has been shown beneficial. This paper studies an efficient use of semivariogram features for object-based high-resolution image classification. First, an input image is divided into segments, for each of which a semivariogram is then calculated. Second, candidate features are extracted as a number of key locations of the semivariogram functions. Then we use an improved Relief algorithm and the principal component analysis to select independent and significant features. Then the selected prominent semivariogram features and the conventional spectral features are combined to constitute a feature vector for a support vector machine classifier. The effect of such selected semivariogram features is compared with those of the gray-level co-occurrence matrix (GLCM) features and window-based semivariogram texture features (STFs). Tests with aerial and satellite images show that such selected semivariogram features are of a more beneficial supplement to spectral features. The described method in this paper yields a higher classification accuracy than the combination of spectral and GLCM features or STFs.     

Area-based and location-based validation of classified image objects

Geographic object-based image analysis (GEOBIA) produces results that have both thematic and geometric properties. Classified objects not only belong to particular classes but also have spatial properties such as location and shape. Therefore, any accuracy assessment where quantification of area is required must (but often does not) take into account both thematic and geometric properties of the classified objects. By using location-based and area-based measures to compare classified objects to corresponding reference objects, accuracy information for both thematic and geometric assessment is available. Our methods provide location-based and area-based measures with application to both a single-class feature detection and a multi-class object-based land cover analysis. In each case the classification was compared to a GIS layer of associated reference data using randomly selected sample areas. Error is able to be pin-pointed spatially on per-object, per class and per-sample area bases although there is no indication whether the errors exist in the classification product or the reference data. This work showcases the utility of the methods for assessing the accuracy of GEOBIA derived classifications provided the reference data is accurate and of comparable scale.     

Fuel type mapping using object-based image analysis of DMC and Landsat-8 OLI imagery

Efficient forest fire management requires precise and up-to-date knowledge regarding the composition and spatial distribution of forest fuels at various spatial and temporal scales. Fuel-type maps are essential for effective fire prevention strategies planning, as well as the alleviation of the environmental impacts of potential wildfire events. The aim of this study was to assess and compare the potential of Disaster Monitoring Constellation and Landsat-8 OLI satellite images (Operational Land Imager), combined with Object-Based Image Analysis (GEOBIA), in operational mapping of the Mediterranean fuel types at a regional scale. The results showcase that although the images of both sensors can be used with GEOBIA analysis for the generation of accurate fuel-type maps, only the OLI images can be considered as applicable for regional mapping of the Mediterranean fuel types on an operational basis.     

Mapping the distribution of the main host for plague in a complex landscape in Kazakhstan: An object-based approach using SPOT-5 XS,Landsat 7 ETM+, SRTM and multiple Random Forests

Plague is a zoonotic infectious disease present in great gerbil populations in Kazakhstan. Infectious disease dynamics are influenced by the spatial distribution of the carriers (hosts) of the disease. The great gerbil, the main host in our study area, lives in burrows, which can be recognized on high resolution satellite imagery. In this study, using earth observation data at various spatial scales, we map the spatial distribution of burrows in a semi-desert landscape.The study area consists of various landscape types. To evaluate whether identification of burrows by classification is possible in these landscape types, the study area was subdivided into eight landscape units, on the basis of Landsat 7 ETM+ derived Tasselled Cap Greenness and Brightness, and SRTM derived standard deviation in elevation.In the field, 904 burrows were mapped. Using two segmented 2.5 m resolution SPOT-5 XS satellite scenes, reference object sets were created. Random Forests were built for both SPOT scenes and used to classify the images. Additionally, a stratified classification was carried out, by building separate Random Forests per landscape unit.Burrows were successfully classified in all landscape units. In the ‘steppe on floodplain’ areas, classification worked best: producer's and user's accuracy in those areas reached 88% and 100%, respectively. In the ‘floodplain’ areas with a more heterogeneous vegetation cover, classification worked least well; there, accuracies were 86 and 58% respectively. Stratified classification improved the results in all landscape units where comparison was possible (four), increasing kappa coefficients by 13, 10, 9 and 1%, respectively.In this study, an innovative stratification method using high- and medium resolution imagery was applied in order to map host distribution on a large spatial scale. The burrow maps we developed will help to detect changes in the distribution of great gerbil populations and, moreover, serve as a unique empirical data set which can be used as input for epidemiological plague models. This is an important step in understanding the dynamics of plague.     

An automated object-based approach for the multiscale image segmentation of forest scenes

Over the last decade the analysis of Earth observation data has evolved from what were predominantly per-pixel multispectral-based approaches, to the development and application of multiscale object-based methods. To empower users with these emerging object-based approaches, methods need to be intuitive, easy to use, require little user intervention, and provide results closely matching those generated by human interpreters. In an attempt to facilitate this, we present multiscale object-specific segmentation (MOSS) as an integrative object-based approach for automatically delineating image-objects (i.e., segments) at multiple scales from a high-spatial resolution remotely sensed forest scene. We further illustrate that these segments cognitively correspond to individual tree crowns, ranging up to forest stands, and describe how such a tool may be used in computer-assisted forest inventory mapping. MOSS is composed of three primary components: object-specific analysis (OSA), object-specific upscaling (OSU), and a new segmentation algorithm referred to as size constrained region merging (SCRM). The rationale for integrating these methods is that the first two provide the third with object-size parameters that otherwise would need to be specified by a user. Analysis is performed on an IKONOS-2 panchromatic image that represents a highly fragmented forested landscape in the Sooke Watershed on southern Vancouver Island, BC, Canada.     

Proof of concept of a novel cloud computing approach for object-based remote sensing data analysis and classification

Advances in the development of Earth observation data acquisition systems have led to the continuously growing production of remote sensing datasets, for which timely analysis has become a major challenge. In this context, distributed computing technology can provide support for efficiently handling large amounts of data. Moreover, the use of distributed computing techniques, once restricted by the availability of physical computer clusters, is currently widespread due to the increasing offer of cloud computing infrastructure services. In this work, we introduce a cloud computing approach for object-based image analysis and classification of arbitrarily large remote sensing datasets. The approach is an original combination of different distributed methods which enables exploiting machine learning methods in the creation of classification models, through the use of a web-based notebook system. A prototype of the proposed approach was implemented with the methods available in the InterCloud system integrated with the Apache Zeppelin notebook system, for collaborative data analysis and visualization. In this implementation, the Apache Zeppelin system provided the means for using the scikit-learn Python machine learning library in the design of a classification model. In this work we also evaluated the approach with an object-based image land-cover classification of a GeoEye-1 scene, using resources from a commercial cloud computing infrastructure service provided. The obtained results showed the effectiveness of the approach in efficiently handling a large data volume in a scalable way, in terms of the number of allocated computing resources.     

Spatial application of Random Forest models for fine-scale coastal vegetation classification using object based analysis of aerial orthophoto and DEM data

High spatial resolution mapping of natural resources is much needed for monitoring and management of species, habitats and landscapes. Generally, detailed surveillance has been conducted as fieldwork, numerical analysis of satellite images or manual interpretation of aerial images, but methods of object-based image analysis (OBIA) and machine learning have recently produced promising examples of automated classifications of aerial imagery. The spatial application potential of such models is however still questionable since the transferability has rarely been evaluated.We investigated the potential of mosaic aerial orthophoto red, green and blue (RGB)/near infrared (NIR) imagery and digital elevation model (DEM) data for mapping very fine-scale vegetation structure in semi-natural terrestrial coastal areas in Denmark. The Random Forest (RF) algorithm, with a wide range of object-derived image and DEM variables, was applied for classification of vegetation structure types using two hierarchical levels of complexity. Models were constructed and validated by cross-validation using three scenarios: (1) training and validation data without spatial separation, (2) training and validation data spatially separated within sites, and (3) training and validation data spatially separated between different sites.Without spatial separation of training and validation data, high classification accuracies of coastal structures of 92.1% and 91.8% were achieved on coarse and fine thematic levels, respectively. When models were applied to spatially separated observations within sites classification accuracies dropped to 85.8% accuracy at the coarse thematic level, and 81.9% at the fine thematic level. When the models were applied to observations from other sites than those trained upon the ability to discriminate vegetation structures was low, with 69.0% and 54.2% accuracy at the coarse and fine thematic levels, respectively.Evaluating classification models with different degrees of spatial correlation between training and validation data was shown to give highly different prediction accuracies, thereby highlighting model transferability and application potential. Aerial image and DEM-based RF models had low transferability to new areas due to lack of representation of aerial image, landscape and vegetation variation in training data. They do, however, show promise at local scale for supporting conservation and management with vegetation mappings of high spatial and thematic detail based on low-cost image data.     

Unsupervised image segmentation evaluation and refinement using a multi-scale approach

In this study, a multi-scale approach is used to improve the segmentation of a high spatial resolution (30 cm) color infrared image of a residential area. First, a series of 25 image segmentations are performed in Definiens Professional 5 using different scale parameters. The optimal image segmentation is identified using an unsupervised evaluation method of segmentation quality that takes into account global intra-segment and inter-segment heterogeneity measures (weighted variance and Moran’s I, respectively). Once the optimal segmentation is determined, under-segmented and over-segmented regions in this segmentation are identified using local heterogeneity measures (variance and Local Moran’s I). The under- and over-segmented regions are refined by (1) further segmenting under-segmented regions at finer scales, and (2) merging over-segmented regions with spectrally similar neighbors. This process leads to the creation of several segmentations consisting of segments generated at three different segmentation scales. Comparison of single- and multi-scale segmentations shows that identifying and refining under- and over-segmented regions using local statistics can improve global segmentation results.     

Hybrid region merging method for segmentation of high-resolution remote sensing images

Image segmentation remains a challenging problem for object-based image analysis. In this paper, a hybrid region merging (HRM) method is proposed to segment high-resolution remote sensing images. HRM integrates the advantages of global-oriented and local-oriented region merging strategies into a unified framework. The globally most-similar pair of regions is used to determine the starting point of a growing region, which provides an elegant way to avoid the problem of starting point assignment and to enhance the optimization ability for local-oriented region merging. During the region growing procedure, the merging iterations are constrained within the local vicinity, so that the segmentation is accelerated and can reflect the local context, as compared with the global-oriented method. A set of high-resolution remote sensing images is used to test the effectiveness of the HRM method, and three region-based remote sensing image segmentation methods are adopted for comparison, including the hierarchical stepwise optimization (HSWO) method, the local-mutual best region merging (LMM) method, and the multiresolution segmentation (MRS) method embedded in eCognition Developer software. Both the supervised evaluation and visual assessment show that HRM performs better than HSWO and LMM by combining both their advantages. The segmentation results of HRM and MRS are visually comparable, but HRM can describe objects as single regions better than MRS, and the supervised and unsupervised evaluation results further prove the superiority of HRM.     

The use of logistic model tree (LMT) for pixel- and object-based classifications using high-resolution WorldView-2 imagery

Logistic model tree (LMT), a new method integrating standard decision tree (DT) induction and linear logistic regression algorithm in a single tree, have been recently proposed as an alternative to DT-based learning algorithms. In this study, the LMT was applied in the context of pixel- and object-based classifications using high-resolution WorldView-2 imagery, and its performance was compared with C4.5, random forest and Adaboost. Results of the study showed that the LMT generally produced more accurate classification results than the other methods for both pixel- and object-based classifications. The improvement in classification accuracy reached to 3% in pixel-based and 5% in object-based classifications. It was also estimated that the LMT algorithm produced the most accurate results considering the allocation and overall disagreement errors. Based on the Wilcoxon’s Signed-Ranks tests, the performance differences between the LMT and the other methods were statistically significant for both pixel- and object-based image classifications.