Geographic Object-based Image Analysis Approach Research Articles

Semiautomatic Mapping of Center Pivot Irrigated Areas Using Sentinel-2 Images and GEOBIA Approach

Image analysis and feature extraction of remoted sensing data are significant for mapping irrigated agriculture areas as a source of information to improve water management and agricultural planning. This paper presents an image segmented base approach GEOBIA (Geographic Object-Based Image Analysis) to extract irrigated areas by Center Pivot Irrigation System (CPIS). This study suggests a semi-automated recognition of circular patterns for the mapping of irrigated regions by center pivots, using Sentinel -2 MSI images, 10 meters spatial resolution. A set of images from different seasons, humid and dry are used to maximize de CPIS’s occurrence. A multiresolution segmentation method was applied, and a large number of segment-based shape features was extracted and used as input to a feature selection procedure (shape descriptors: Area; Compactness; Circularity Factor; Length/Width; Radius of smallest enclosing ellipse; and Roundness). In addition, another shape descriptor “Circularity Factor” was developed in this research and played an important role during preliminaries classification processes. The accuracy assessment of preliminaries classifications has validated used the Circularity Factor together with the other chosen shape descriptors to reach better results to CPIS’s detection. Furthermore, 86.23% of the CPIS mapped in the classification process is in accordance with the ground truth map. This methodology can be used to map large areas in a relatively short time and provides a tool for monitoring irrigated areas.

Leveraging NAIP Imagery for Accurate Large-Area Land Use/land Cover Mapping: A Case Study in Central Texas

Large-area land use land cover (LULC) mapping using high-resolution imagery remains challenging due to radiometric differences between scenes, the low spectral depth of the imagery, landscape heterogeneity, and computational limitations. Using a random forest (RF)- supervised machine-learning algorithm, we present a geographic object-based image analysis approach to classifying a large mosaic of 220 National Agriculture Imagery Program orthoimagery into lulc categories. The approach was applied in central Texas, USA, covering over 6000 km2. We generated 36 variables for each object and accounted for spatial structures of sample data to determine the distance at which samples were spatially independent. The final rf model produced 94.8% accuracy on independent stratified random samples. In addition, vegetation and water indices, the mean and standard deviation of principal components, and texture features improved classification accuracy. This study demonstrates a cost-effective way of producing an accurate multi-class land use/land cover map using high-spatial/low-spectral resolution orthoimagery.

Geographic object-based image analysis and artificial neural networks for digital soil mapping

The use of techniques to calibrate soil prediction models based on legacy maps is restricted to areas where conventional survey and soil mapping were performed. It is necessary to seek alternatives to calibrate soil predictive models without a legacy soil map. The problem is the difficulty to delineate polygons of soil classes based on soil sampling points. This paper presents a novel digital soil mapping strategy using only georeferenced points of soil profiles to delineate detailed polygons of soil classes by the Geographic Object-Based Image Analysis (GEOBIA) approach and Artificial Neural Networks (ANN) models. The main objective was to evaluate the integration of the GEOBIA approach at different segmentation levels with ANN’s models and sampling points to produce a digital soil map of the Vale dos Vinhedos region, in Rio Grande do Sul, Brazil. From a Digital Terrain Model, 10 predictive variables were extracted. From the RapidEye remote sensing image the following spectral variables were extracted: Normalized Difference Vegetation Index and Normalized Difference Water Index. A new set of variables was produced using principal component analysis after standardization of the original variables. The multiresolution segmentation algorithm was used to create image objects at different segmentation levels. Four segmentation levels were tested with scale parameter (SP) 1, 2, 5 and 10; and the Cheesboard segmentation (CS) was used to transform the original pixels into polygons. Information on types of soil was obtained from 163 georeferenced soil profiles. Five ANN’s structures were implemented: four for the segmented data (SP01, SP02, SP05 and SP10) and one for the per-pixel approach (CS); 16 repetitions of each ANN were performed. With the best models for each segmentation level, extrapolations of soil classes were performed for the entire study area and validated with a detailed legacy soil map. The per-pixel approach model obtained the worst result. The best result in the extrapolation of soil types found was with the object-based approach (SP1). In our understanding, the scientific community can profit from the proposed methodology particularly when it only has georeferenced points of soil classes and not a legacy map, i.e., immediately after a field survey without the need for photointerpretation for delineation of the soil mapping units. The use of GEOBIA for digital mapping of soil classes seems to be a promising approach for future research.

Center pivot field delineation and mapping: A satellite-driven object-based image analysis approach for national scale accounting

Center pivot irrigation systems are used to enhance crop production in many countries around the world. Establishing the location and extent of such fields provides information that assists in describing a range of agricultural metrics, including crop identification, yield forecasts, monitoring of irrigation requirements and crop water use, as well as supporting national and regional auditing, licensing and compliance efforts. However, detailed information on the number, extent and changing dynamics of agricultural fields is often lacking in many countries: nowhere more so than in developing regions. To address this lack, we performed a national scale accounting of center pivot fields in Saudi Arabia, using a three year multi-temporal analysis of Landsat-8 satellite data. A geographic object-based image analysis approach was developed based on five 50 × 50 km sub-areas extracted from Landsat data for the year 2015, and applied to delineate individual center pivot fields at a national scale for 2013, 2014 and 2015. The extent of fields was determined via a map of the annual maximum Normalized Difference Vegetation Index (NDVI), while a 15 m spatial resolution map of annual panchromatic band minimums was used to produce an edge detection layer to delineate individual adjoining fields. Amongst a range of classification parameters that were included in the object-based mapping approach, shape information, such as the center pivot field length, length:width ratio, and elliptic fit, were identified as critical parameters. Applying the rule-set that was developed from the five 50 × 50 km sub-regions to the national scale resulted in the identification of 36,052 (11,103 km2), 38,114 (11,902 km2), and 37,254 (11,555 km2) individual fields in 2013, 2014, and 2015, respectively. Approximately 94% of these fields were correctly detected, while their individually measured area was mapped with >95% combined accuracy for fields >0.225 km2 when evaluated against manually delineated fields. Smaller center pivot fields, and specifically those adjoining neighbouring fields, had lower area mapping accuracies (>91% in 75% of cases). The object-based approach allowed a national scale and multi-temporal assessment of center pivot field delineation and mapping, affording new insights into agricultural practice and providing a methodological basis for examining the impact of water management and related policy initiatives, amongst many other potential applications. Apart from filling a clear knowledge gap in Saudi Arabia, the approach has the potential to be expanded elsewhere: particularly to similar locations within the Middle East and North Africa.

Rapid Recent Deforestation Incursion in a Vulnerable Indigenous Land in the Brazilian Amazon and Fire-Driven Emissions of Fine Particulate Aerosol Pollutants

Deforestation in the Brazilian Amazon is related to the use of fire to remove natural vegetation and install crop cultures or pastures. In this study, we evaluated the relation between deforestation, land-use and land-cover (LULC) drivers and fire emissions in the Apyterewa Indigenous Land, Eastern Brazilian Amazon. In addition to the official Brazilian deforestation data, we used a geographic object-based image analysis (GEOBIA) approach to perform the LULC mapping in the Apyterewa Indigenous Land, and the Brazilian biomass burning emission model with fire radiative power (3BEM_FRP) to estimate emitted particulate matter with a diameter less than 2.5 µm (PM2.5), a primary human health risk. The GEOBIA approach showed a remarkable advancement of deforestation, agreeing with the official deforestation data, and, consequently, the conversion of primary forests to agriculture within the Apyterewa Indigenous Land in the past three years (200 km2), which is clearly associated with an increase in the PM2.5 emissions from fire. Between 2004 and 2016 the annual average emission of PM2.5 was estimated to be 3594 ton year−1, while the most recent interval of 2017–2019 had an average of 6258 ton year−1. This represented an increase of 58% in the annual average of PM2.5 associated with fires for the study period, contributing to respiratory health risks and the air quality crisis in Brazil in late 2019. These results expose an ongoing critical situation of intensifying forest degradation and potential forest collapse, including those due to a savannization forest-climate feedback, within “protected areas” in the Brazilian Amazon. To reverse this scenario, the implementation of sustainable agricultural practices and development of conservation policies to promote forest regrowth in degraded preserves are essential.

GEOBIA for mangrove mapping using UAV-modified NIR 1 camera sensor

Mangrove ecosystem is a natural resource that can be beneficial for coastal community. One of its benefits is it can be utilized as an ecotourism spot such as Baros Mangrove Conservation Area in Kretek, Bantul. Hence, mangrove ecosystem must be conserved or expanded if possible. The first step to conserving mangrove ecosystem is to map its existing condition. UAV can be used to acquire a high-resolution imagery quickly. With modified near-infrared (NIR) sensor from a pocket camera can produce three band i.e. NIR, green, and blue that can be used to calculate several vegetation indices such as Simple Ration (SR), Normalized Difference Vegetation Index (NDVI), and Enhanced Difference Vegetation Index (ENDVI). Those bands and vegetation indices then used to map mangrove using Geographic Object-Based Image Analysis approach. The result indicated that high-resolution imagery from modified NIR camera sensor can be used to detect mangrove and distinguish it from other object using supervised classification. The digitally classified imagery then compared to manually digitized mangrove map to calculate the total error of the map.

Land Cover Changes in Open-Cast Mining Complexes Based on High-Resolution Remote Sensing Data

Remote sensing technologies can play a fundamental role in the environmental assessment of open-cast mining and the accurate quantification of mine land rehabilitation efforts. Here, we developed a systematic geographic object-based image analysis (GEOBIA) approach to map the amount of revegetated area and quantify the land use changes in open-cast mines in the Carajás region in the eastern Amazon, Brazil. Based on high-resolution satellite images from 2011 to 2015 from different sensors (GeoEye, WorldView-3 and IKONOS), we quantified forests, cangas (natural metalliferous savanna ecosystems), mine land, revegetated areas and water bodies. Based on the GEOBIA approach, threshold values were established to discriminate land cover classes using spectral bands, the normalized difference vegetation index (NDVI), normalized difference water index (NDWI) and a light detection and range sensor (LiDAR) digital terrain model and slope map. The overall accuracy was higher than 90%, and the kappa indices varied between 0.82 and 0.88. During the observation period, the mining complex expanded, which led to the conversion of canga and forest vegetation to mine land. At the same time, the amount of revegetated area increased. Thus, we conclude that our approach is capable of providing consistent information regarding land cover changes in mines, with a special focus on the amount of revegetation necessary to fulfill environmental liabilities.

Comparing Deep Neural Networks, Ensemble Classifiers, and Support Vector Machine Algorithms for Object-Based Urban Land Use/Land Cover Classification

With the advent of high-spatial resolution (HSR) satellite imagery, urban land use/land cover (LULC) mapping has become one of the most popular applications in remote sensing. Due to the importance of context information (e.g., size/shape/texture) for classifying urban LULC features, Geographic Object-Based Image Analysis (GEOBIA) techniques are commonly employed for mapping urban areas. Regardless of adopting a pixel- or object-based framework, the selection of a suitable classifier is of critical importance for urban mapping. The popularity of deep learning (DL) (or deep neural networks (DNNs)) for image classification has recently skyrocketed, but it is still arguable if, or to what extent, DL methods can outperform other state-of-the art ensemble and/or Support Vector Machines (SVM) algorithms in the context of urban LULC classification using GEOBIA. In this study, we carried out an experimental comparison among different architectures of DNNs (i.e., regular deep multilayer perceptron (MLP), regular autoencoder (RAE), sparse, autoencoder (SAE), variational autoencoder (AE), convolutional neural networks (CNN)), common ensemble algorithms (Random Forests (RF), Bagging Trees (BT), Gradient Boosting Trees (GB), and Extreme Gradient Boosting (XGB)), and SVM to investigate their potential for urban mapping using a GEOBIA approach. We tested the classifiers on two RS images (with spatial resolutions of 30 cm and 50 cm). Based on our experiments, we drew three main conclusions: First, we found that the MLP model was the most accurate classifier. Second, unsupervised pretraining with the use of autoencoders led to no improvement in the classification result. In addition, the small difference in the classification accuracies of MLP from those of other models like SVM, GB, and XGB classifiers demonstrated that other state-of-the-art machine learning classifiers are still versatile enough to handle mapping of complex landscapes. Finally, the experiments showed that the integration of CNN and GEOBIA could not lead to more accurate results than the other classifiers applied.

Advancing barrier island habitat mapping using landscape position information

Barrier islands are dynamic ecosystems that change gradually from coastal processes, including currents and tides, and rapidly from episodic events, such as storms. These islands provide many important ecosystem services, including storm protection and erosion control to the mainland, habitat for fish and wildlife, and tourism. Habitat maps, developed by scientists, provide a critical tool for monitoring changes to these dynamic ecosystems. Barrier island monitoring often requires custom habitat maps due to several factors, including island size and the classification of unique geomorphology-based habitats, such as beach, dune, and barrier flats. In this study, we reviewed barrier-island-specific habitat mapping efforts and highlighted common habitat class types, source data, and mapping approaches. We also developed a framework for mapping geomorphology-based barrier island habitats using a rule-based, geographic object-based image analysis approach, which included the use of field data, tide data, high-resolution orthophotography, and lidar data. This framework integrates several barrier island mapping advancements with regard to the use of landscape position information for automated dune extraction and the use of Monte Carlo analyses for the treatment of elevation uncertainty for elevation-dependent habitats. Specifically, we used the uncertainty analyses to refine automated dune delineation based on elevation relative to extreme storm water levels and to increase the accuracy of intertidal and supratidal/upland habitat delineation. We found that dune extraction results were enhanced when elevation relative to storm water levels and visual interpretation were also applied. This framework could also be applied to beach–dune systems found along a mainland.

Object-based urban landcover mapping methodology using high spatial resolution imagery and airborne laser scanning

Mapping landcover in cities is essential for urban ecology and landuse management, yet urban landcover is often highly heterogeneous at fine spatial scales. Pixel-based approaches are shown to be less successful for effectively mapping urban landcover due to high heterogeneity, with relatively low accuracies reported despite the use of high spatial resolution optical imagery. Alternatively, geographic object-based image analysis (GEOBIA) has yielded higher accuracies across a range of urban applications. We combine three-dimensional (3-D) information from airborne laser scanning (ALS) data with RapidEye high-spatial-resolution imagery in a GEOBIA approach to classify urban landcover in a large metropolitan region in Vancouver, Canada. Results indicate that 12 urban classes could be accurately mapped at 2-m spatial resolution across 150,000 ha with an overall accuracy of 88% (kappa 0.87). Though 5-m RapidEye multispectral pixels were often mixed in heterogeneous urban areas, the additional insight provided by the 3-D ALS information enabled accurate classification of fine spatial objects such as street trees and single-family dwellings.

A GEOBIA Approach for Multitemporal Land-Cover and Land-Use Change Analysis in a Tropical Watershed in the Southeastern Amazon

The southeastern Amazon region has been intensively occupied by human settlements over the past three decades. To evaluate the effects of human settlements on land-cover and land-use (LCLU) changes over time in the study site, we evaluated multitemporal Landsat images from the years 1984, 1994, 2004, 2013 and Sentinel to the year 2017. Then, we defined the LCLU classes, and a detailed “from-to” change detection approach based on a geographic object-based image analysis (GEOBIA) was employed to determine the trajectories of the LCLU changes. Three land-cover (forest, montane savanna and water bodies) and three land-use types (pasturelands, mining and urban areas) were mapped. The overall accuracies and kappa values of the classification were higher than 0.91 for each of the classified images. Throughout the change detection period, ~47% (19,320 km2) of the forest was preserved mainly within protected areas, while almost 42% (17,398 km2) of the area was converted from forests to pasturelands. An intrinsic connection between the increase in mining activity and the expansion of urban areas also exists. The direct impacts of mining activities were more significant throughout the montane savanna areas. We concluded that the GEOBIA approach adopted in this study combines the advantages of quality human interpretation and the capacities of quantitative computing.

Forest cover changes in Gorce NP (Poland) using photointerpretation of analogue photographs and GEOBIA of orthophotos and nDSM based on image-matching based approach

ABSTRACTForest cover change can be detected with high precision using 3D geospatial data and semi-automatic analyses of Remote Sensing data. The aim of our study, performed in Gorce National Park in Poland, was to generate a land use land cover (LULC) map and use it to analyse forest cover change. The study area is a subalpine forest region that has been affected by bark beetle and wind disturbances. The Geographic Object-Based Image Analysis approach was used for classification, with Colour Infrared orthophotos and normalized Digital Surface Models generated using image-matching approach. Gathered results showed that dominating LULC class is coniferous forests (3380 ha; 47% of study area), when second largest class is deciduous forests (2204 ha; 30%). The dead Norway spruce stands (465.5 ha; 6.5%) showed significant increase comparing to 114.1 ha mapped in 1997.

COMPARATIVE ASSESSMENT BETWEEN PER-PIXEL AND OBJECT-ORIENTED FOR MAPPING LAND COVER AND USE

ABSTRACT: The traditional per-pixel classification methods consider only spectral information, and may be limited. Object-based classifiers, however, also consider shape and texture, firstly segmenting the image, and then classifying individual objects. Thus, a Geographic Object-Based Image Analysis (GEOBIA) was compared in conjunction with data mining techniques and a traditional per-pixel method. A cut of Landsat-8, bands 2 to 7, orbit/point 223/77, located between the municipalities of Cascavel, Corbelia, Cafelândia and Tupassi, in the west part of the state of Parana, from 12/18/2013 was used. In the GEOBIA approach was realized image segmentation, spatial and spectral attribute extraction, and classification using the decision tree supervised algorithm, J48. For the per-pixel method, we used the supervised Maximum Likelihood Classifier. Both approaches presented equivalent results, with Kappa Index of 0.75 and Global Accuracy (GA) of 78.97% for the approach by GEOBIA and Kappa Index of 0.72 and GA of 77.44% for the perpixel classification. The classification by GEOBIA showed better accuracy for the soil, forest and soybean classes, and did not show the splash aspect, which visually improves the classification result.

Developing Lebanon's fire danger forecast

Lebanon lacks a system for forecasting wildfire danger with the combined use of dynamic weather forecasts and comprehensive fire risk information (i.e., hazard and vulnerability). Accordingly, the aim of this study was to develop a new national fire danger forecast system for use as an operational fire management tool in Lebanon. The specific objectives were to 1) produce an advanced fire risk map of Lebanon, 2) customize an index for use in fire danger forecasting, and 3) develop an automated system for daily fire danger forecasts. Producing an advanced fire risk map involved the combined use of collected fire risk data and predicted wildfire occurrence. Forward stepwise binary logistic regression analysis of 24 socio-economic and biophysical variables was initially adopted to predict wildfire occurrence. Geographic object-based image analysis (GEOBIA) approach was employed to map the different classes of fire risk. Consequently, the customization of an index for fire danger forecasting involved the use of both fire risk and Fire Weather Index (FWI) data using also the GEOBIA approach. Overall, modeling socio-economic and biophysical factors correctly predicted an average of 85.7% of fire occurrences between 1998 and 2012 across the country. In addition, a comparison of FWI forecasts and fire occurrence (i.e., observed fire events) during the fire season of 2015 resulted in sensitivity (76.2%) and specificity (18.6%) estimates. Finally, automated daily fire danger forecasts are expected to serve as essential tools for both scientific and operational purposes.

GEOBIA-based identification of alluvial fans and bajadas through geomorphometry, image analysis and fuzzy ontology

ABSTRACTAlluvial fan and bajada mapping is important for practical and economic importance to society. In this study, advanced geomorphometric, image processing and knowledge representation methods were investigated, enhanced and developed, towards the simultaneous identification of alluvial fans and bajadas. The developed method was based on the landform-pattern element approach. Data used included the 10 m spatial resolution National Elevation Dataset, provided by USGS and a pan-sharpened 15 m Landsat OLi imagery. The study area was located in the Death Valley, Nevada, USA. Digital Terrain Model (DTM) processing included noise-removal filtering and depression treatment. A two-phase GEographic Object-Based Image Analysis approach involving multi-scale segmentation and fuzzy ontology-based reasoning was designed. On the first phase, topographic forms were identified, by examining their morphometry and spatial relationships. On the second phase, the alluvial fans and bajadas were identified based on drainage pattern texture and topographic features such as landform shape. Due to a systematic omission on the fan toe outline, each fan and bajada spectral signature was taken into account to determine the omitted areas through region growing on the multispectral imagery. Accuracy assessment indicated satisfactory results, since a 90.3% completeness and 88.8% quality were achieved for both alluvial fans and bajadas.

A Geographic Object-Based Approach for Land Classification Using LiDAR Elevation and Intensity

By providing detailed height and intensity land surface information, high-resolution LiDAR data have proved to be effective in supporting land classification when combined with other major geospatial data sources, such as hyperspectral images. However, rectifying and fusing multisource geospatial data involves what normally is a manual and time-consuming process. In this letter, we propose a geographic object-based image analysis approach to enable semiautomatic land classification and mapping using LiDAR elevation and intensity data. The methodological framework consists of a series of operations, including preprocessing, object-based segmentation, creation of statistical variables from elevation and intensity, and semisupervised classification. We have successfully applied this approach to the classification of multiple land features, including asphalt, grass, barren land, swimming pool, shrubland, pavement, and buildings. Results show that our proposed approach performs better than LiDAR analysis methods in classifying different land parcels.

Evaluating the influence of geo-environmental factors on gully erosion in a semi-arid region of Iran: An integrated framework

Despite the importance of soil erosion in sustainable development goals in arid and semi-arid areas, the study of the geo-environmental conditions and factors influencing gully erosion occurrence is rarely undertaken. As effort to this challenge, the main objective of this study is to apply an integrated approach of Geographic Object-Based Image Analysis (GEOBIA) together with high-spatial resolution imagery (SPOT-5) for detecting gully erosion features at the Kashkan-Poldokhtar watershed, Iran. We also aimed to apply a Conditional Probability (CP) model for establishing the spatial relationship between gullies and the Geo-Environmental Factors (GEFs). The gully erosion inventory map prepared using GEOBIA and field surveying was randomly partitioned into two subsets: (1) part 1 that contains 70% was used in the training phase of the CP model; (2) part 2 is a validation dataset (30%) for validation of the model and to confirm its accuracy. Prediction performances of the GEOBIA and CP model were checked by overall accuracy and Receiver Operating Characteristics (ROC) curve methods, respectively. In addition, the influence of all GEFs on gully erosion was evaluated by performing a sensitivity analysis model. The validation findings illustrated that overall accuracy for GEOBIA approach and the area under the ROC curve for the CP model were 92.4% and 89.9%, respectively. Also, based on sensitivity analysis, soil texture, drainage density, and lithology represent significantly effects on the gully erosion occurrence. This study has shown that the integrated framework can be successfully used for modeling gully erosion occurrence in a data-poor environment.

ESTIMATING WOOD VOLUME FOR <i>PINUS BRUTIA</i> TREES IN FOREST STANDS FROM QUICKBIRD-2 IMAGERY

Knowledge of forest parameters, such as wood volume, is required for a sustainable forest management. Collecting such information in the field is laborious and even not feasible in inaccessible areas. In this study, tree wood volume is estimated utilizing remote sensing techniques, which can facilitate the extraction of relevant information. The study area is the University Forest of Taxiarchis, which is located in central Chalkidiki, Northern Greece and covers an area of 58km<sup>2</sup>. The tree species under study is the conifer evergreen species <i>P. brutia</i> (Calabrian pine). Three plot surfaces of 10m radius were used. VHR Quickbird-2 images are used in combination with an allometric relationship connecting the Tree Crown with the Diameter at breast height (Dbh), and a volume table developed for Greece. The overall methodology is based on individual tree crown delineation, based on (a) the marker-controlled watershed segmentation approach and (b) the GEographic Object-Based Image Analysis approach. The aim of the first approach is to extract separate segments each of them including a single tree and eventual lower vegetation, shadows, etc. The aim of the second approach is to detect and remove the “noisy” background. In the application of the first approach, the Blue, Green, Red, Infrared and PCA-1 bands are tested separately. In the application of the second approach, NDVI and image brightness thresholds are utilized. The achieved results are evaluated against field plot data. Their observed difference are between -5% to +10%.

ESTIMATING WOOD VOLUME FOR PINUS BRUTIA TREES IN FOREST STANDS FROM QUICKBIRD-2 IMAGERY

Abstract. Knowledge of forest parameters, such as wood volume, is required for a sustainable forest management. Collecting such information in the field is laborious and even not feasible in inaccessible areas. In this study, tree wood volume is estimated utilizing remote sensing techniques, which can facilitate the extraction of relevant information. The study area is the University Forest of Taxiarchis, which is located in central Chalkidiki, Northern Greece and covers an area of 58km2. The tree species under study is the conifer evergreen species P. brutia (Calabrian pine). Three plot surfaces of 10m radius were used. VHR Quickbird-2 images are used in combination with an allometric relationship connecting the Tree Crown with the Diameter at breast height (Dbh), and a volume table developed for Greece. The overall methodology is based on individual tree crown delineation, based on (a) the marker-controlled watershed segmentation approach and (b) the GEographic Object-Based Image Analysis approach. The aim of the first approach is to extract separate segments each of them including a single tree and eventual lower vegetation, shadows, etc. The aim of the second approach is to detect and remove the “noisy” background. In the application of the first approach, the Blue, Green, Red, Infrared and PCA-1 bands are tested separately. In the application of the second approach, NDVI and image brightness thresholds are utilized. The achieved results are evaluated against field plot data. Their observed difference are between -5% to +10%.

Building change detection through multi-scale GEOBIA approach by integrating deep belief networks with fuzzy ontologies

ABSTRACTMonitoring and mapping urban changes is of great importance for the development, planning and management of the urban zone, especially in countries with a rapidly growing urban area. The aim of this paper was to develop a GEographic Object-Based Image Analysis (GEOBIA) approach, by integrating Deep Learning classification and Fuzzy Ontologies through multi-scale analysis, to monitor building changes in suburban areas of Greece. Three suburban areas of east Attica, Greece were selected as representative to test the methodology. For each area, one QuickBird and one WorldView 2 image, taken in 2006 and 2011, respectively, were employed. Three segmentation levels and a three-level class hierarchy were developed for the extraction process. Deep Belief Networks were employed on the lowest level of the segmentation hierarchy (Level 1) for an initial detection of areas of possible change. To detect the changes in building infrastructure, the classification result of Level 1 was refined based on interpretation rules, developed on the upper levels of the hierarchy (Level 2 and Level 3). Accuracy assessment indicated that 93.5% of the total number of changes were successfully detected, while the commission error was less than 20%.