Search

Due to the complex spectral and spatial structures of remotely sensed images, the delineation of land use/land cover classes using conventional approaches is a challenging task. This article tackles the problem of seeking optimal parameters of multi-resolution segmentation for a classification task using WorldView-2 imagery. Taguchi optimization was applied to search optimal parameters using the plateau objective function (POF) and quality rate (Qr) as fitness criteria. Analysis of variance was also used to estimate the contributions of the parameters for POF and Qr, separately. The scale parameter was the most effective one, with contribution levels of 87.45% and 56.87% for POF and Qr, respectively. Linear regression and support-vector regression methods were used to predict the results of the experiment. Test results revealed that Taguchi optimization was more effective than linear regression and support-vector regression for predicting POF and Qr values.

The poplar species in the forest ecosystems are one of the most valuable and beneficial species for the society and environment. Conventional methods require high cost, time and labor need, and the results obtained vary and are insu˚cient in terms of achieved accuracy level. Determination of poplar cultivated fields and mapping of their spatial sites play a vital role for decision-makers and planners to enhance the economic and ecological value of poplar trees. The study aims to map Poplar (P. deltoides) cultivated areas in Akyazi district of Sakarya, Turkey province using various combinations of the Sentinel-2A image bands. For this purpose, object-based classification based on multi-resolution segmentation algorithm was utilized to produce image objects and ensemble learning algorithms, namely, Adaboost (AdaB), Random Forest (RF), Rotation Forest (RotFor) and Canonical correlation forest (CCF) were applied to produce thematic maps. In order to analyze the effects of the spectral bands of the Sentinel-2A image on the object-based classification performance, three datasets consisting of different spectral band combinations (i.e. four 10 m bands, six 20 m bands and ten 10m pan-sharpened bands) were used. The results showed that the RotFor and CCF classifiers produced superior classification performances compared to the AdaB and RF classifiers for the band combinations regarded in this study. Moreover, it was found that determination of poplar tree class level accuracy reached to ~94% in terms of F-score. It was also observed that the inclusion of the six spectral bands at 20 m resolution resulted in a noteworthy increase in classification accuracy (up to 6%) compared to single 10m band combination.

Climate change and global warming along with human activities have caused abrupt changes in the atmosphere, marine, and terrestrial ecosystems. One of these changes is the rising number of mucilage events in marine ecosystems. During the recent two decades, mucilage blooms have begun to appear more frequently in the Sea of Marmara in Turkey, surrounded by seven densely populated cities with various anthropogenic activities including household pollution, heavy industrialization, agricultural pollution, commercial fishing, shipyards, and specialized marine terminals with high maritime traffic density. In Spring 2021, a massive mucilage event observed in the Sea of Marmara plagued the entire ecosystem and raised awareness among the government and the public to detect and monitor this phenomenon. In this research, daily monitoring and detection of mucilage formations from the coarse spatial resolution MODIS products were investigated during the 2021 bloom period. The results were validated with the reference mucilage datasets derived from Sentinel-2A imagery and in situ spectroradiometer measurements over mucilage formations. The results revealed that the MODIS surface reflectance profiles were highly correlated with the field spectral measurements and estimated mucilage formations were spatially overlapped with Sentinel-2A. Overall, the produced maps accurately depicted the mucilage-covered areas despite the limitations of unreliable estimates along the land-water transition lines, and no-data areas due to the low-quality observations and high cloud coverage.

There will be nearly 2 billion new residents by 2030 due to human migration and population growth, yet the consequences of both current and future urbanization on biodiversity conservation are poorly known. However, an increase in the awareness of the consequences on the role of biodiversity in sustaining livelihoods and human wellbeing has been translated into the various movements, in line with the Sustainable Development Goals (SDG), especially SDG15 and SDG3. Conservation areas (CAs) have become one of the solutions to uphold ecosystem sustainability for biodiversity conservation. Thus, the study aimed to analyze the changes in land use and land cover (LULC) surrounding the CA - Forest Research Institute Malaysia (FRIM) Natural Heritage Site, Malaysia. The objectives were: 1) to quantify the changes of LULC, 2) to construct the Digital Elevation Model (DEM), of FRIM and the surroundings, thus representing the Malaysian terrestrial CAs and, 3) to suggest LULC changes- based criteria for CA buffer zone (CABAs). The data used were spatial data and satellite images and, ERDAS Imagine and ArcGIS software to execute the analysis. The study revealed a reduction in forest areas and addition in built-up (housing) and water bodies. There was an encroachment of built-up area (housing) into the FRIM. Therefore, the study suggested buffer zones for FRIM, a shield to prevent the stresses caused by urbanization and encroachment activities that may jeopardize the intactness of the conservation areas. The study also provided an effort to formulate a more effective strategy for enhancing sustainable development to the policymakers.

Küresel ısınmanın karasal alanlarda olduğu gibi denizlerde de doğal dengeyi bozan etkileri bulunmaktadır. Adriyatik ve Tiren denizlerinde 200 yılı aşkın süredir sıklıkla yaşanan müsilaj olayları, Marmara Denizinde ilk kez 2007 yılında gözlemlenmiş, 2021 yılı mayıs ayından itibaren ise tekrar yoğun şekilde ortaya çıkarak doğal bir felakete dönüşmüştür. Deniz yüzeyinde geniş alanlar kaplayan müsilaj, deniz altındaki canlı hayatını da ciddi şekilde tehdit etmektedir. Müsilajın alansal büyüklüğünün, yoğunluğunun, zamansal hareketinin ve deniz suyu ile arasındaki ısı farkının takibi ve sahada yürütülen temizlik çalışmalarının etkilerinin tespiti için uzaktan algılama teknolojileri etkin bir çözümdür. Periyodik olarak üretilen müsilaj dağılım haritaları, deniz yüzeyindeki toplam müsilaj hacminin hesaplanması, rehabilitasyon çalışmalarının yürütülmesi ve alınacak önlemlerin belirlenmesi başta olmak üzere birçok faaliyet için temel altlık konumundadır. Bu çalışmada, Sentinel-2 uydu görüntüsünün dört spektral bandı arasındaki ilişkiyi kullanan ‘müsilaj indeksi’ ile müsilaj kaplı alanların tespiti hedeflenmiştir. Bu indeksten faydalanılarak 14, 19 ve 24 Mayıs 2021 tarihli uydu görüntüleri için müsilaj indeks haritaları oluşturulmuştur. Sentinel-2 için tanımlanan müsilaj indeksi ile müsilaj oluşumlarının hızlı bir şekilde tespit ve analizinde önemli bir veri seti oluşturulabileceği görülmüştür. Ayrıca 17 Mayıs 2021 tarihli Landsat-8 termal uydu görüntüsü yardımıyla temiz deniz yüzeyi ve müsilaj oluşumlarının yüzey sıcaklıkları analiz edilmiştir. Sonuç olarak, müsilaj alanlarının yüzey sıcaklığının ortalama deniz yüzey sıcaklığına göre 3°C’ye kadar daha yüksek olduğu tespit edilmiştir. Elde edilen bulgular, uzaktan algılamanın erken uyarı ve hazırlık çalışmalarındaki etkinliğini ve afet riski yönetiminde vazgeçilmez bir araç olduğunu ortaya koymuştur.

In recent years, ensemble learning methods have become popular in landslide susceptibility mapping (LSM) with varying degrees of success. Within classifier ensemble concept, decision tree b...

With smaller pixel coverages from recent sensors, considering single pixels in the classification process has become ineffective and incapable for delineation of the characteristics of targeted lan...

With the increase in spatial resolution of recent sensors, object-based image analysis (OBIA) has gained importance for producing detailed land use maps. One of the main advantages of OBIA is that a variety of spectral, spatial and textural features can be extracted for the segmented image objects that are later utilized in classification. However, using a large number of features not only increases the required computational time, but also requires a large number of ground samples, which is unavailable in most cases. For these reasons, feature selection (FS) has become an important research topic for OBIA based classification studies. In this study, three filter-based FS algorithms namely, Chi square, information gain and ReliefF were applied to determine the most effective object features that ensure high separability among landscape features. For this purpose, importance degree (i.e. ranks) of 110 input object features were firstly estimated by the algorithms, and correlation-based merit function was then applied to determine optimum feature subset size. Multi-resolution segmentation algorithm was applied for segmenting a WorldView-2 image. Support vector machine, random forest and nearest neighbour classifiers were all utilized to classify segmented image objects using the selected object features. Results revealed that the FS algorithms were effective for selecting the most relevant features. Also, the classifiers produced the highest performances with 24 out of 110 features selected by the information gain (IG) algorithm. Particularly, the support vector machine classifier produced the highest overall accuracy (92.00%) with 24 selected features determined by the IG algorithm. A significant improvement of about 4% was achieved by applying FS procedures that was found statistically significant in terms of Wilcoxon signed-ranks test.

Object-based image analysis (OBIA) has attained great importance for the delineation of landscape features, particularly with the accessibility to satellite images with high spatial resolution acquired by recent sensors. Statistical parametric classifiers have become ineffective mainly due to their assumption of normal distribution, vast increase in the dimensions of the data and availability of limited ground sample data. Despite pixel-based approaches, OBIA takes semantic information of extracted image objects into consideration, and thus provides more comprehensive image analysis. In this study, Indian Pines hyperspectral data set, which was recorded by the AVIRIS hyperspectral sensor, was used to analyse the effects of high dimensional data with limited ground reference data. To avoid the dimensionality curse, principal component analysis (PCA) and feature selection based on Jeffries–Matusita (JM) distance were utilized. First 19 principal components representing 98.5% of the image were selected using the PCA technique whilst 30 spectral bands of the image were determined using JM distance. Nearest neighbour (NN) and random forest (RF) classifiers were employed to test the performances of pixel- and object-based classification using conventional accuracy metrics. It was found that object-based approach outperformed the traditional pixel-based approach for all cases (up to 18% improvement). Also, the RF classifier produced significantly more accurate results (up to 10%) than the NN classifier.

Classifying spectrally similar crop types in fragmented landscapes is a difficult task due to the low spectral and spatial resolution of satellite imagery. The objective of this study is twofold: (I) to evaluate the performance of a recent ensemble methodology, namely canonical correlation forest (CCF), and (ii) to investigate the potential of recently launched Sentinel-2 (S-2) image for crop classification. The algorithm is based on building a forest-type ensemble model using multiple canonical correlation trees. Its performance was compared to widely-used random forest (RF) and rotation forest (RotFor) for the classification of 4-band and 10-band S-2 and Landsat-8 (L-8) OLI datasets. In addition to overall accuracy estimations, results of the methods were assessed using recently introduced map-level and category-level disagreement measures. Results showed that the CCF and RotFor algorithms produced statistically similar results for the S-2 datasets but statistically different results for the OLI...

Very high-resolution images obtained with recently launched satellite sensors have been used intensively in the remote sensing area. The widespread use of high-resolution images has greatly facilitated the creation and updating of land use/land cover (LULC) maps. Traditional pixel-based image analysis methods that extract information based solely on the spectral values of pixels are generally not suitable for high-resolution images. Unlike pixel-based approaches, object-based image analysis (OBIA) uses pixel clustering (image objects) instead of pixels by considering the shape, texture, context and spectral features and provide richer information extraction. Image segmentation is an important process and prerequisite for the OBIA process. It is essential to evaluate the performance of segmentation algorithms for the determination of effective segmentation methods and optimization of segmentation parameters. In this study, the multi-resolution segmentation algorithm is used for the segmentation process. The effect of spectral bands on segmentation quality was analysed using a Worldview-2 high-resolution satellite image. In order to analyze segmentation quality, two unsupervised quality metrics, namely, F-measure and Plateau Objective Function (POF) values were calculated for each band separately. In this manner, optimum parameter values were determined using different variations of Moran's I Index and variance values. Image segmentation was performed by using different scale, shape and compactness parameter values. In this context, 30 segmentation analysis was performed considering three different spectral bands (red, green and near-infrared bands). The results showed that the highest segmentation quality was acquired for the NIR band among the spectral bands for the F-measure method, while the highest segmentation quality value was achieved for the green band for the POF metric. In addition, the optimum segmentation parameter values of the scale, shape and compactness were determined as 30-0.3-0.5 and 50-0.1-0.3 for F-measure and POF approaches, respectively.

Satellite based remote sensing technologies and Geographical Information Systems (GIS) present operable and cost-effective solutions for mapping fires and observing post-fire regeneration. Mersin-Gülnar wildfire, which occurred in August 2008 in Turkey, selected as study site. The fire was devastating and continued 55 days. According to Turkish General Directorate of Forestry reports, it caused two deaths and left hundreds of people homeless. The aim of this study is to determine the fire severity and monitor vegetation recovery with using multitemporal spectral indices together with topographical factors. Pre-fire and post-fire Landsat ETM+ images were obtained to assess the related fire severity with using the widely-used differenced Normalized Burn Ratio (dNBR) algorithm. Also, the Normalized Vegetation Index (NDVI) and Soil Adjusted Vegetation Index (SAVI) were used to determine vegetation regeneration dynamics for a period of six consecutive years. In addition, aspect image derived from Aster Global Digital Elevation Model (GDEM) were used to determine vegetation regeneration regime of the study area. Results showed that 5388 ha of area burned with moderate to high severity damage. As expected, NDVI and SAVI values distinctly declined post-fire and then began to increase in the coming years. Mean NDVI value of burned area changed from 0.48 to 0.17 due to wildfire, whilst mean SAVI value changed from 0.61 to 0.26. Re-growth rates calculated for NDVI and SAVI 57% and 63% respectively, six years after the fire. Moreover, NDVI and SAVI were estimated six consecutive year period by taking into consideration east, south, north and west facing slopes. Analysis showed that north-facing and east-facing slopes have higher regeneration rates in compared to other aspects. This study serves as a window to an understanding of the process of fire severity and vegetation regeneration that is vital in wildfire management systems.

Within the last two decades, object-based image analysis (OBIA) considering objects (i.e. groups of pixels) instead of pixels has gained popularity and attracted increasing interest. The most important stage of the OBIA is image segmentation that groups spectrally similar adjacent pixels considering not only the spectral features but also spatial and textural features. Although there are several parameters (scale, shape, compactness and band weights) to be set by the analyst, scale parameter stands out the most important parameter in segmentation process. Estimating optimal scale parameter is crucially important to increase the classification accuracy that depends on image resolution, image object size and characteristics of the study area. In this study, two scale-selection strategies were implemented in the image segmentation process using pan-sharped Qickbird-2 image. The first strategy estimates optimal scale parameters for the eight sub-regions. For this purpose, the local variance/rate of change (LV-RoC) graphs produced by the ESP-2 tool were analysed to determine fine, moderate and coarse scales for each region. In the second strategy, the image was segmented using the three candidate scale values (fine, moderate, coarse) determined from the LV-RoC graph calculated for whole image. The nearest neighbour classifier was applied in all segmentation experiments and equal number of pixels was randomly selected to calculate accuracy metrics (overall accuracy and kappa coefficient). Comparison of region-based and image-based segmentation was carried out on the classified images and found that region-based multi-scale OBIA produced significantly more accurate results than image-based single-scale OBIA. The difference in classification accuracy reached to 10% in terms of overall accuracy.

Identification of landslide prone areas and production of accurate landslide susceptibility zonation maps have been crucial topics for hazard management studies. Since the prediction of susceptibility is one of the main processing steps in landslide susceptibility analysis, selection of a suitable prediction method plays an important role in the success of the susceptibility zonation process. Although simple statistical algorithms (e.g. logistic regression) have been widely used in the literature, the use of advanced non-parametric algorithms in landslide susceptibility zonation has recently become an active research topic. The main purpose of this study is to investigate the possible application of kernel-based Gaussian process regression (GPR) and support vector regression (SVR) for producing landslide susceptibility map of Tonya district of Trabzon, Turkey. Results of these two regression methods were compared with logistic regression (LR) method that is regarded as a benchmark method. Results showed that while kernel-based GPR and SVR methods generally produced similar results (90.46% and 90.37%, respectively), they outperformed the conventional LR method by about 18%. While confirming the superiority of the GPR method, statistical tests based on ROC statistics, success rate and prediction rate curves revealed the significant improvement in susceptibility map accuracy by applying kernel-based GPR and SVR methods.

In landslide susceptibility mapping, factor weights have been usually determined by expert judgements. A novel methodology for weighting landslide causative factors by integrating statistical feature weighting algorithms was proposed. The primary focus of this study is to investigate the effectiveness of automatic feature weighting algorithms, namely Fisher, Chi-square and Relief-F algorithms. Analytic hierarchy process (AHP) method was used as a benchmark method to compare the performances of the weighting algorithms. All weighted factors were tested using factor-weighted overlay method, and quality of these maps was assessed using overall accuracy, area under the ROC curve (AUC) and success rate curve. In addition, Wilcoxon’s signed-rank test was applied to evaluate statistical differences between both estimated overall accuracies and AUCs, respectively. Results showed that the weights determined by feature weighting methods outperformed the conventional AHP method by about 6% and this level of ...

Image classification is a complex process, the accuracy of which is mainly related to the characteristics of the dataset, complexity of the problem under analysis, and the robustness of the classification algorithm. This study aims to assess and validate a recent classifier ensemble method called canonical correlation forest (CCF) for object-based image analysis. The proposed methodology is based on building a decision forest ensemble model using a set of decision trees constructed with the dataset obtained as a result of canonical correlation analysis. Its performance was compared to conventional nearest neighbor (NN) and popular random forest (RF) algorithms in the classification of WorldView-2 imagery. The number of image object features (total 128 features) was reduced to 42 features using a correlation-based feature selection algorithm. The results of the study showed that the CCF algorithm was found to be superior to RF and NN algorithms with respect to the estimated accuracy measures. The improvement in accuracy reached 4% for both 42- and 128-feature datasets and this level of improvement was found to be statistically significant when considering against the McNemar’s test results. Moreover, the performances of the CCF and NN algorithms were found to be sensitive for the dimension of the input dataset, whilst the performance of RF was more stable with or without feature selection. To sum up, the CCF algorithm, introduced as a new member of the tree-based ensemble learning algorithms, was found to be a powerful alternative to the RF algorithm considered dataset used in this study.

Machine learning techniques have been increasingly employed for solving many scientific and engineering problems. These data driven methods have been lately utilized with great success to produce landslide susceptibility maps. They give promising results particularly for mapping large landslide prone areas with limited geotechnical data. This chapter surveys their use in landslide susceptibility analysis and presents a case study investigating their effectiveness with regard to a conventional statistical method, namely logistic regression. It starts with the importance of spatial prediction of future landslides from past and present ones and discusses the requirement of advanced techniques for landslide susceptibility mapping. A critical literature survey is given under five main categories including core algorithms and their ensembles together with their hybrid forms. An application is presented for machine learning application using bagging, random forest, rotation forest and support vector machines with their optimal settings.

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 pixe...

Machine learning algorithms reported to be robust and superior to the conventional parametric classifiers have been recently employed in object-based classification. Within these algorithms, ensemble learning methods that construct set of individual classifiers and combining their predictions to make final decision about unlabelled data have been successfully applied. In this study, performance and effectiveness of a novel ensemble learning algorithm, rotation forest (RotFor) aiming to build diverse and accurate classifiers, was investigated for the first time in object-based classification using a WorldView-2 (WV-2) satellite image. Also, the combination of satellite imagery and ancillary data (i.e. normalized difference vegetation index and principal components) were assessed. Random forest (RF), support vector machine (SVM) and nearest neighbour (NN) algorithms were also used as benchmark classifiers to evaluate the power of RotFor. The classification results confirmed that integration of ancillary dat...