Object-based image analysis

Offering rapidly and periodically achievable and very high resolution (VHR) data at low cost, unmanned aerial vehicles (UAVs) have recently become one of the most popular remote sensing technologies. In this study, a novel Normalized Digital Surface Model (NDSM) fusion approach is proposed to improve the object-based land cover classification (LCC) performance of the RGB camera equipped UAVs. The UAV flights were completed in a multi-class sloping urban area with oblique viewing geometry and a VHR orthomosaic was generated. The objectbased LCC was performed applying bottom-up (BU) multiresolution segmentation and Nearest Neighbor and Rule-Based classification algorithms. In order to improve the limited segmentation performance of the RGB orthomosaic, high-quality NDSM was produced and fused as an additional imaging band. The highest LCC accuracy utilizing four imaging bands were achieved with a weight of 0.3 for NDSM and 1 for RGB bands. To preserve the information provided by the NDSM data for each image object, a single-stage segmentation process has been performed in contrast to BU segmentation. The results demonstrated that the proposed approach increased the LCC precision up to 15%, recall up to 4.71%, F1 score up to 8.58%, Kappa as 7% and overall accuracy as 4.57%.

Impervious surface or man-made surface is an important factor in the hydrologic cycle and water management of a developing city. Quantifying and understanding the extent of impervious surface area should be carried out to assess accurately water discharge during heavy rainfall, land use change detection, development and many other impacts on the community. Iligan City is one of the most devastated areas when tropical storm Washi hit Mindanao in December 2011.This research aimed to extract impervious surfaces like concrete and asphalt roads, buildings, parking areas and concrete pavement within a 7.05 km² city proper of Iligan. Utilizing Light Detection and Ranging (LiDAR) datasets and derivatives, an object-based image analysis (OBIA), support vector machine (SVM) classifier, and ground truthing, five classes were extracted. These classes were building, vegetation, impervious ground surface, pervious ground surface, and water. Building class and impervious ground surface were subsequently aggregated to a class of impervious surface. Furthermore, vegetation and pervious ground surface were assigned to a class of pervious surface. Results showed an overall accuracy of 0.83 with a Kappa Index of Agreement (KIA) of 0.78. Greater accuracy was attained when further refinement using rule-based classification approach was conducted to misclassified objects. Percent coverage of impervious surface area is greater than 50% of the total study area.

Uzaktan algilama cisimlere temas etmeksizin hava fotograflari ve uydu goruntulerini kullanarak haritalama ve dijital goruntu isleme seklinde veri elde etmektir. Dogal kaynaklara ait verileri depolama, analiz etme ve sunma konularinda bir kac metod bulunmaktadir. Bu surecte uzaktan algilama ve Cografi Bilgi sistemleri oldukca onemli rol oynamaktadir. Uzaktan algilama verileri arazi kullaniminda olusan degisimleri orman amenajmani, havza ve kent planlamasi, su kaynaklari ve dogal kaynak planlamasi icin detayli bilgiler saglamaktadir. Konumsal orman envanteri icin farkli uzaktan algilama verilerinin dogruluk degerleri oldukca onemlidir. Bu nedenle mescere tiplerinin belirlenmesi sirasinda kullanilan uzaktan algilama verileri degerlendirilmistir. Bu uzaktan algilama verileri; hava fotograflari (1/15000 olcekli renki kizilotesi hava fotografi) ve uydu goruntuleri (LANDSAT TM, IKONOS)’dir. Bu uzaktan algilama verilerinin hepsinde hedeflen mescere tipleri haritasi uretmektir. Butun harita ...

Lucian Drăguţ 1,2 , Clemens Eisank 1 , Thomas Strasser 1 , Thomas Blaschke 1 ... 1Department of Geography and Geology, University of Salzburg, Hellbrunnerstraße 34, Salzburg 5020, Austria ... 2Department of Geography, West University of Timişoara, V. Pârvan Blv. 4, Timişoara ...

The aim of this article consists in a presentation and a comparison of filters for the detection of archaeological structures developed through open-source Geographical Information Systems tools applied to the surface of the archaeological site known as Mesas del Esparragal, located in the municipality of Alcalá de los Gazules (Cádiz, Spain). The site, identified by some researchers as Lascuta because of its relationship with the famous Lascuta Bronze, has exceptional characteristics in terms of its strategic position for controlling the territory. For this reason, the Romans at the hands of the consul Lucius Emilius Paulus, were able to establish their camp in this important place with the aim of putting down the uprising perpetuated by the city of Hasta Regia. For this purpose, the data obtained through the LiDAR sensor collected by the National Plan for Aerial Orthophotography (PNOA) will be used. These data will be classified by means of the LASTools complement and later a series of filters will be made through the Relief Visualization Toolbox (RVT) software and the SAGA software, with the purpose of comparing them and determining which filter or filters are the most appropriate when locating possible archaeological structures under the surface. The filters to be compared are: The Three-directional Hillshade, the Simple Local Relief Model (SLRM) and the Red Relief Image Map (RRIM).

The designations employed and the presentation of material in this information product do not imply the expression of any opinion whatsoever on the part of the Food and Agriculture Organization of the United Nations (FAO) concerning the legal or development status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. The mention of specific companies or products of manufacturers, whether or not these have been patented, does not imply that these have been endorsed or recommended by FAO in preference to others of a similar nature that are not mentioned. The views expressed in this information product are those of the author(s) and do not necessarily reflect the views or policies of FAO.

L'osservazione del livello e della distribuzione del danno dopo un evento catastrofico sono di primaria importanza nei momenti successivi ad un evento, per la pianificazione immediata per la risposta all'evento e, successivamente, per una migliore comprensione degli effetti del terremoto sull'ambiente costruito. Dopo il terremoto di Bam, che ha colpito l'Iran il 26 dicembre 2003, satelliti ad alta risoluzione (Quickbird e Ikonos) sono diventati delle valide fonti di informazioni, sia per le caratteristiche divulgative (diffusione di informazioni tramite Internet), sia con lo scopo vero e proprio di mappatura della distribuzione del danno e per la ricognizione delle aree maggiormente colpite. Saranno l'argomento della presente comunicazione l'analisi delle immagini satellitari, tramite tecnologia ad oggetti: lo specifico obiettivo della ricerca è la descrizione e la definizione di un inventario degli edifici estratto esclusivamente da immagini satellitari, attraverso specifici algoritmi di parametrizzazione della segmentazione. La possibilità di estrarre informazioni di tipo quantitativo dalle immagini satellitari in modo semi-automatico è, infatti, particolarmente utile per lo sviluppo di applicazioni in regioni non coperte da un sistema informativo geografico o da cartografia adeguati. Le immagini QuickBird saranno poi analizzate nell'ottica della definizione di un indice di cambiamento, che potrà essere utilizzato per definire la distribuzione del danneggiamento.

This study investigates the use of Object-based Machine Learning (OBML) algorithms to detect features such as building rooftop materials and vegetation in Red-Green-Blue (RGB) images captured by Unmanned Aerial Vehicles (UAVs) over the Chenkalady Divisional Secretariat in Sri Lanka. It employs Object-based Image Analysis (OBIA) for feature identification through object-based segmentation, feature extraction, and classification using random forest classifiers. The analysis utilizes high-spatial resolution RGB imagery from the OpenAerialMap (OAM) platform, calibrated radiometrically via the Empirical Line Method (ELM). The method achieves over 95% accuracy in segregating rooftop materials based on distinct texture and colour patterns at an optimal segmentation scale. Vegetation indices, such as the Normalized Green-Red Difference Index (NGRDI) and the RGB Vegetation Index (RGBVI) enhance the resources for identifying tree species and building types. Results demonstrate the algorithm's effectiveness in identifying features, particularly in classifying coconut trees with a 45% precision rate using a 100-sample examination for location accuracy. The nearest neighbour method achieves an 86% accuracy rate within a 1-meter radius. In addition, vegetation indices highlight vegetation zones and help identify different tree species. Accuracy assessment using 100 randomly distributed points shows that both NGRDI and RGBVI vegetation indices achieve an overall accuracy of 96% and a kappa coefficient of 0.97. This study advances feature detection methods in UAV imagery, emphasizing the synergy between deep learning techniques. The proposed algorithm demonstrates promising results in accurately identifying diverse landscape features, paving the way for enhanced applications in environmental monitoring, disaster management, and urban planning using UAVs equipped with visible light sensors that provide high-resolution imagery.

The early detection and management of plant diseases in short-statured crops such as beans, black gram, cowpea, groundnut, and tomato are crucial for minimizing yield losses and improving agricultural resilience in regions like the Karnataka-Tamil Nadu border. This study presents the design, development, and evaluation of a convolutional neural network (CNN)-based plant disease detection model integrated into a low-height rover system tailored for smallstatured crop fields. A dataset of 18,256 annotated images spanning 23 classes of healthy and diseased leaf samples was compiled, pre-processed, and used to train a multi-layered CNN architecture. While batch-level prediction of 18,256 test samples revealed overfitting evidenced by a uniform misclassification of all samples into a single class-the system demonstrated reliable disease identification during real-time field deployment. The rover's performance was further enhanced through the integration of soil moisture sensors, enabling preliminary analysis of environmental correlations with disease occurrence. The model and its hardware platform were evaluated qualitatively and quantitatively, with findings presented using confusion matrices, class distribution charts, and visual documentation from field trials. Limitations such as reduced effectiveness under fluctuating weather conditions, limited battery life in large-scale fields, and incompatibility with taller crops were acknowledged. Recommendations include integrating solar energy systems and improving data augmentation techniques to enhance robustness. This study contributes significantly to the underexplored domain of precision agriculture for short-statured crops, offering an accessible, scalable framework for early disease diagnosis using AI and embedded systems. Despite its constraints, the system presents a viable tool for smallholder farmers, fostering sustainable agricultural practices. Future research should address model generalization, environmental adaptability, and integration with cloud-based farm management systems. This work underscores the potential of AIdriven solutions in transforming low-input farming ecosystems by offering context-specific technological interventions rooted in local agronomic realities.

One traditional, although visionary goal of the remote sensing (RS) community is the development of operational satellite-based measurement systems suitable for automating the quantitative analysis of large-scale spaceborne multi-source multi-resolution image databases . In past years this goal was almost exclusively dealt with by research programs focused on land cover (LC) and land cover change (LCC) detection at global scale (Gutman et al., 2004) (pp. 451, 452). In recent years the objective of developing operational satellite-based measurement systems has become increasingly urgent due to multiple drivers. While cost-free access to large-scale low spatial resolution (SR) (above 40 m) and medium SR (from 40 to 20 m) spaceborne image databases has become a reality

To deliver sample estimates provided with the necessary probability foundation to permit generalization from the sample data subset to the whole target population being sampled, probability sampling strategies are required to satisfy three necessary not sufficient conditions: (i) All inclusion probabilities be greater than zero in the target population to be sampled. If some sampling units have an inclusion probability of zero, then a map accuracy assessment does not represent the entire target region depicted in the map to be assessed. (ii) The inclusion probabilities must be: (a) knowable for nonsampled units and (b) known for those units selected in the sample: since the inclusion probability determines the weight attached to each sampling unit in the accuracy estimation formulas, if the inclusion probabilities are unknown, so are the estimation weights. This original work presents a novel (to the best of these authors' knowledge, the first) probability sampling protocol for ...

According to existing literature and despite their commercial success, state-of-the-art two-stage non-iterative geographic object-based image analysis (GEOBIA) systems and three-stage iterative geographic object-oriented image analysis (GEOOIA) systems, where GEOOIA  GEOBIA, remain affected by a lack of productivity, general consensus and research. To outperform the degree of automation, accuracy, efficiency, robustness, scalability and timeliness of existing GEOBIA/GEOOIA systems in compliance with the Quality Assurance Framework for Earth Observation (QA4EO) guidelines, this methodological work is split into two parts. The present first paper provides a multi-disciplinary Strengths, Weaknesses, Opportunities and Threats (SWOT) analysis of the GEOBIA/GEOOIA approaches that augments similar analyses proposed in recent years. In line with constraints stemming from human vision, this SWOT analysis promotes a shift of learning paradigm in the pre-attentive vision first stage of a remote sensing (RS) image understanding system (RS-IUS), from sub-symbolic statistical model-based (inductive) image segmentation to symbolic physical model-based (deductive) image preliminary classification. Hence, a symbolic deductive pre-attentive vision first stage accomplishes image sub-symbolic segmentation and image symbolic pre-classification simultaneously. In the second part of this work a novel hybrid (combined deductive and inductive) RS-IUS architecture featuring a symbolic deductive pre-attentive vision first stage is proposed and discussed in terms of: (a) computational theory (system design); (b) information/knowledge representation; (c) algorithm design; and

The availability of high-resolution satellite images allowed new applications, especially in urban areas. In Brazil, where up-to-date information is frequently missing in data banks at Municipal, State and Federal level, the information obtained from these images could eventually fill this gap. In this paper, a case study, performed in the city of São José dos Campos, São Paulo State, Brazil is presented. A QUICKBIRD scene was classified using an object-oriented approach. After presenting the fundamentals of this new approach, a land cover map of a section from this city is presented and discussed. This is a contribution for urban planning. 1.

Semantic segmentation requires methods capable of learning high-level features while dealing with large volume of data. Towards such goal, Convolutional Networks can learn specific and adaptable features based on the data. However, these networks are not capable of processing a whole remote sensing image, given its huge size. To overcome such limitation, the image is processed using fixed size patches. The definition of the input patch size is usually performed empirically (evaluating several sizes) or imposed (by network constraint). Both strategies suffer from drawbacks and could not lead to the best patch size. To alleviate this problem, several works exploited multi-context information by combining networks or layers. This process increases the number of parameters resulting in a more difficult model to train. In this work, we propose a novel technique to perform semantic segmentation of remote sensing images that exploits a multi-context paradigm without increasing the number of parameters while defining, in training time, the best patch size. The main idea is to train a dilated network with distinct patch sizes, allowing it to capture multi-context characteristics from heterogeneous contexts. While processing these varying patches, the network provides a score for each patch size, helping in the definition of the best size for the current scenario. A systematic evaluation of the proposed algorithm is conducted using four high-resolution remote sensing datasets with very distinct properties. Our results show that the proposed algorithm provides improvements in pixelwise classification accuracy when compared to state-of-the-art methods.

An airport with its runways, terminals, connecting roads etc. is a complex of many units and even its size may vary, still it occupies a large area of the earth's surface. For this reason, the construction of an airport causes a significant change of land cover / land use in the area in which it is located. Istanbul, most important economic, political and social centre of Turkey, has a special geographical position linking the Asian and the European continents, which makes it one of the most important point of world airlines traffic. There are two airports which are still in use in Istanbul, but due to increasing needs, the 3 rd Airport is being constructed on the European side of the city and on the Black Sea coast, at 35km from the city centre. The work, which began in May 2015, is designed to be carried out on four stages over an area of 76.5 million square meters and is still ongoing. 78% of the total project will be completed when the construction of the three runways and a terminal that forms the first stage of 3 rd Airport, and it is planned to be the world's largest and busiest airport, when it is completed in October 2018. Being constructed on a very wide forest land, the land cover / land use change caused by the 3 rd Airport construction is of great importance in terms of environmental change. In this context, it is important to evaluate the land cover / land use change in the region by the end of 2018, when the first phase of the project is over. Today, multitemporal satellites are actively used to determine land cover / land use change. In this study, the change in the region was assessed using Sentinel 2A satellite data. The multi-temporal image dataset used in the study are acquired in 2015 and 2017, consists of 4 spectral bands in the visible-infrared (VNIR) region of the electromagnetic spectrum and has 10 m spatial resolution. An object-based classification method was used to determine the land change over the three-year period following the beginning of the airport construction. The results of the study were evaluated by visual interpretation and classification accuracy assessment and the advantages and disadvantages of the method used were examined.

Coastal zones play a particularly significant role in Sri Lanka, because it is an island with a broad coastline. However, natural processes and human activities are rapidly altering coastlines, with profound consequences for the environment, communities and industries that depend on. This study investigates the causes of coastal erosion and its ongoing challenges, with a special focus on Hikkaduwa Divisional Secretariat Division in Galle district. The required data for DSAS, NDVI, and NDBI analyses were collected through questionnaire surveys, interviews, and

L'articolo mette a confronto due differenti approcci di classificazione automatica orientata agli oggetti per ottenere mappe tematiche Uso del Suolo GIS ready: eCognition della Definiens e T-MAP della SI2G. Entrambe le soluzioni sono basate su procedure che non sfruttano solo il contenuto spettrale del singolo pixel ma definiscono, nella fase di segmentazione, le regioni che costituiscono le entità del territorio stesso, in modo da predisporre una mappa degli oggetti e dei loro attributi. Selezionando un'area test relativa ad una parte della Provincia di Ancona è stato effettuato un confronto fra le mappe Uso del Suolo (scala 1:10000) ottenute dalle due differenti metodologie. L'obiettivo è stato quello di analizzare sia i risultati della segmentazione, nella maniera più oggettiva e rigorosa possibile, per valutare la qualità geometrica delle entità estratte, sia la coerenza della classificazione tematica derivata. Sono state analizzate in particolar modo i risultati della classificazione relativa alle classi 1, 2 e 3 della Corine Land Cover (CLC) al secondo livello, generalizzando i risultati di T-MAP che ottiene precisioni e dettaglio al terzo livello della CLC. La sovrastima della classe CLC 2 del 10% rilevata sul prodotto eCognition è risultata subordinata alla fase di segmentazione che governa la definizione delle classi e quindi degli oggetti di appartenenza.

Son yıllarda yüksek konumsal çözünürlüklü uydu görüntülerinin sınıflandırılmasında geleneksel piksel tabanlı sınıflandırma yaklaşımı yerine obje tabanlı yaklaşımın kullanımı önem kazanmıştır. Obje tabanlı görüntü analizi iki temel işlem adımından oluşur. Bunlardan ilki benzer spektral özelliklere sahip piksellerin gruplandırılarak homojen yapılı görüntü objelerinin oluşturulmasını içeren segmentasyon işlemidir. İkincisi ise oluşturulan görüntü objelerin spektral, istatistiksel, geometrik ve ilişkisel özelliklerinden yararlanarak gerçekleştirilen sınıflandırma işlemidir. Sınıflandırma işleminde üretilen görüntü objeleri temel sınıflandırma birimi olarak değerlendirmeye alındığından, obje tabanlı görüntü analizinin doğruluğu segmentasyon işleminin kalitesi ile doğrudan ilişkilidir. Görüntü segmentasyonu ölçek, şekil ve yoğunluk olarak bilinen üç farklı parametre yardımıyla gerçekleştirilmekte olup, bunlar arasında görüntü objelerinin büyüklüğünü kontrol eden ölçek parametresinin tespiti segmentasyonun başarısı açısından büyük bir öneme sahiptir. Bu çalışmada yüksek çözünürlüklü WorldView-2 uydu görüntüsünün obje tabanlı yaklaşımla sınıflandırılması problemi ele alınmıştır. Bu çalışmada, görüntü segmentasyonu işleminin gerçekleştirilmesinde tüm görüntü için tek bir ölçek parametresi uygulamak yerine, görüntüdeki doğal ve yapay nesnelerin dağılımları göz önüne alan bölgesel ölçek parametresi kullanımı önerilmiştir. Tek ölçek parametresi ve farklı ölçek parametreleri ile oluşturulan görüntü objelerinin sınıflandırılmasında en yakın komşuluk (EYK) ve destek vektör makineleri (DVM) algoritmaları kullanılmıştır. Sonuçlar, farklı ölçek parametreleri kullanımıyla DVM ve EYK algoritmalarının hesaplanan sınıflandırma doğruluklarında %6 ve %3’e varan seviyede önemli iyileşmeler olduğunu göstermektedir. McNemar’s testi sonuçlarına göre algoritmaların sınıflandırma performansındaki bu artışların istatistiksel olarak anlamlı olduğu ortaya koyulmuştur

In an attempt to replicate Global Forest Watch Canada (2009) and , this research project evaluated five core methods to determine which one should be used for clearcut detection. The five methods assessed were: Band 5 Differencing, Band 7 Differencing, Normalized Difference Vegetation Index Differencing, Enhanced Wetness Difference Index Differencing, and Unsupervised Classification. The project used three phases to determine which method was the most accurate, whether or not a supervised classification increased method accuracy, and whether an increase in the interval between images dates had an impact on method accuracy. It was hypothesized that Normalized Difference Vegetation Index Differencing would be the most accurate method, a supervised classification would increase accuracy, and an increase in time interval between image dates would decrease method accuracy. A set of study areas in Nova Scotia were selected for the project. Trout Lake Training Area was used to learn how clearcut pixels appeared on Landsat images by using a high resolution IKONOS satellite image as a reference. The five aforementioned methods, were then applied to the Long Lake Evaluation Area. A supervised classification was conducted prior to method implementation for Governors Lake Evaluation Area, in an attempt to determine if the methods' accuracy would increase. Once the first two phases were completed, the methods were assessed for accuracy. The most accurate method was applied to Kelly Lake Application Area, Guysborough County, to determine if the accuracy would decrease with an increased time interval between image dates. Band 7 Differencing post-supervised classification was determined to be the most accurate method, contrary to the hypothesis. The supervised classification did increase accuracy of clearcut detection. When Band 7 Differencing with prior supervised classification was applied to images with an increased time interval, the accuracy decreased. May 4 th 2015 iii RÉSUMÉ An Assessment of Methods for Clearcut Detection Using Landsat Satellite Images of Nova Scotia By: Marc A. Therrien Dans une tentative de reproduire Global Forest Watch Canada (2009) et Weeks, et al. (2013), ce projet de recherche a évalué cinq méthodes de base pour déterminer quelles méthodes devraient être utilisées pour la détection de la coupe à blanc. Les cinq méthodes évaluées étaient : Band 5 Differencing, Band 7 Differencing, Normalized Difference Vegetation Index Differencing, Enhanced Wetness Index Difference Differencing et Unsupervised Classification. Le projet a utilisé trois phases pour déterminer quelle était la méthode la plus précise, si une classification supervisée augmente la précision de la méthode, et si un allongement de l'intervalle d'image a eu un impact sur la précision de la méthode. Les trois hypothèses sont : la Normalized Difference Vegetation Index Difference serait la méthode la plus précise, une classification supervisée augmenterait la précision de la méthode, et une augmentation de l'intervalle de temps entre les dates d'image diminuerait la précision de la méthode. Pour compléter le projet quatre régions situées en Nouvelle Écosse ont été utilisé. La première region a été utilisée pour apprendre l'apparence d'un pixel du coup à blanc sur une image Landsat en utilisant une image satellite de haute résolution IKONOS come vérification. Les cinq méthodes, mentionnées ci-dessus, ont ensuite été appliquées à la zone d'évaluation Long Lake situé dans la Vallée d'Annapolis. Une classification supervisée a été réalisée avant l'application des méthodes d'évaluation pour la zone du Lac Gouverneurs dans une tentative de déterminer si la précision des méthodes augmenterait. Une fois les deux premières étapes ont été réalisées, les méthodes ont été évaluées pour la précision. La méthode la plus précise a été appliquée à Kelly Lake situé dans le Comté de Guysborough, afin de déterminer si la précision diminuerait avec l'augmentation de l'intervalle de temps entre les dates d'image. Band 7 Differencing après la classification supervisée était la méthode la plus précise, contrairement à l'hypothèse. La classification supervisée a augmenté la précision de la méthode. Lorsque l'intervalle de temps avait augmenté entre les dates des images, la précision du Band 7 Differencing avait diminué. Le 4 May 2015 finally said, "this is a huge improvement from your last draft". I am in forever debt to you. Will Flanagan also deserves an acknowledgement, as his advice and knowledge throughout this past year saved me from some sticky situations. He also provided great conversations and distractions when needed during the course of the project. Other acknowledgements include, my parents, Helen Whitman and Michel Therrien who have supported me through this process, and have told me "You better get a good job with this". Finally, I would like to thank, and acknowledge Katelynn Carter-Rogers, as she has been my support, motivation, and has given me much input and guidance throughout my undergraduate, and now thesis project. v

Özet: Ülkemizde bölgeler arası gelişmişlik farklılıkları ülkemizin en temel sorunlarından biri olup, birçok sosyo-ekonomik sorunun kaynağını da oluşturmaktadır. Bu farklılıkları yaratan temel unsurlar, eğitim yetersizliği, sermaye birikiminin azlığı, yerel bir pazarın yeterince gelişmemesi, girdi temini, kalifiye eleman temininde karşılaşılan güçlükler, coğrafi konum, alt yapı eksikliği ve tarım arazilerinin ekonomik büyüklüğün altında olması, çok parçalı tarım arazisi sayılabilir. Tarım makinaları alanında yapılan akademik çalışmalar da bölgeler arası gelişmişlikten etkilenmekte olup, sosyo-ekonomik yapının bir parçası görünümünü sergilemektedir. Ancak, teknolojik gelişmeler ve teknolojik düzeyin içselleştirilerek tarım makinaları araştırmalarında kullanılması hem sektörde, hem de bu alanda araştırma yapan araştırıcılarda paradigma değişikliğini gerektirecektir. Bu çalışma ile yeni yeni tarım makinaları alanında kullanım alanı bulan elektromanyetik spektrum içerisinde yer alan dalga boylarından nasıl yararlanıldığı, nerelerde kullanıldığı ve gelecekte yapılacak araştırmalarda nelerin kullanılabileceği özetlenmiştir.

Bu çalışmanın amacı, videolar içerisinde bulunan çıplak gözle görülmesi, açığa çıkarılması veya gösterilmesi zor olan zamansal değişimleri ortaya çıkarmak ve daha önce yapılan farklı benzer çalışmaların deneysel olarak analiz etmektir. Bu çalışmada ortaya çıkarılması zor olan değişimleri vurgulamak için iki yaklaşım üzerinde çalışılmıştır. Bunlardan ilki olan Euler (piksel bazlı işleme) yaklaşımında, ilk olarak ana video büyültme metodu çağırılır. Bu metot vasıtasıyla standart video dizileri girdi olarak alınır. İkinci olarak, uzaysal ayrışma işlemi yapılır ve sonrasını takip eden işlemde de her bir kareye zamansal filtreleme uygulanır. Buna mukabil, ikinci yaklaşım olan Lagrangian (nokta izleme) yaklaşımında sırasıyla video kayıt, öznitelik noktaları takibi, yörünge kümelenmesi, yoğun optik akış interpolasyonu, katman segmentasyonu, büyütme, doku doldurma ve uyum çevirimi işlemi sırasıyla yapılarak görüntüler üzerinden veri analizinin yapılması sağlanır. Bu tarama makalesi kapsamında her iki yöntemden bahsedilerek karşılaştırılmaları gerçekleştirilmiştir.

Satellite survey is widely used for archaeological site discovery, but the efficacy of the method has received little systematic investigation. In this analysis, twelve study participants of different experience levels performed an unstructured remote survey of 197 km 2 in the Sama and Moquegua valleys of south central Peru where previous pedestrian surveys recorded 546 archaeological sites. Results indicate an average site discovery rate of 9.3% (0-18%, 95% range). The most experienced participants detect up to 20% (17-22%) of known archaeological sites. These detection rates can be used to derive reliable site frequency estimates on the Andean coast, which can be used in planning and budgeting for field efforts and estimating demographic patterns at large spatial scales that are difficult to achieve through pedestrian survey. More generally, this analysis offers a method for deriving correction terms specific to other parts of the world. Additionally, the results can serve as a baseline for evaluating the effectiveness of emerging artificial intelligence routines for archaeological site detection.

This research investigates the various RADARSAT-2 polarimetric SAR features for urban land cover classification using object-based method combining with support vector machine (SVM) and ruled-based approach. Six-dates of RADARSAT-2 fine-beam polarimetric SAR data were acquired in the rural-urban fringe of Greater Toronto Area during June to September, 2008. The major landuse/land-cover classes were high-density built-up areas, low-density built-up areas, roads, forests, parks, golf courses, water and several types of agricultural crops. The polarimetric SAR features examined are the parameters from Pauli, Freeman and Cloude-Pottier decompositions as well as the elements from coherence matrix and the intensities and their logarithm form of each channel. For urban land cover classification, SVM is combined with rule-based method for the object-based classification. The image objects containing the multitemporal polarimetric features were classified using the SVM classifier first. The SVM classification results were further refined using a rule-based approach. Rules were built to recognize specific classes defined by the shape features and the spatial relationships within the context. In terms of the effectiveness of different SAR ploarimtric parameters, the results indicated that the processed Pauli feature set could produce best classification result while the use of all the polarimetric features did not produce the best classification result. The raw Pauli parameters could generate similar result as all T elements. The logarithm parameters such as log intensity and processed Pauli parameters perform better than the intensity and raw Pauli respectively. The proposed object-based classification using SVM and rule-based approach yielded higher classification accuracies than the object-based classification using nearest neighbor classifier.

Semantic segmentation of remotely sensed images for land-use and land-cover classes plays a
significant role in various ecosystem management applications. State-of-the-art results in assigning land-use
and land-cover classes are primarily achieved using fully convolutional encoder-decoder architectures.
However, the uneven distribution of the land-use and land-cover classes becomes a major hurdle leading
to performance skewness towards majority classes over minority classes. This paper proposes a novel dualphase training, with the first phase proposing a new undersampling technique using minority class focused
class normalization and the second phase that uses this learnt knowledge for ensembling to prevent overfitting
and compensate for the loss of information due to undersampling. The proposed method achieved an overall
performance gain of up to 2% in MIoU, Kappa, and F1 Score metrics and up to 3% in class-wise F1-score
when compared to the baseline models on Wuhan Dense Labeling, Vaihingen and Potsdam datasets.

Computational (or "artificial") intelligence is not just about robots. It is about understanding the nature of intelligent thought and action using computers as experimental devices. It also deals with the nature of inferential mechanisms and how computer programs allow us to discover how we produce inferences. In this paper I introduce some of the key points in Computational Intelligence in Archaeology, exploring the implications in our discipline, both theoretically and methodologically, of Machine Learning tools and techniques. Theoretical and practical aspects of computer programs able to reproduce the same tasks archaeologists do are reviewed in this paper. The question of whether it is possible to automate the archaeological knowledge production is of both great theoretical interest and increasing practical importance, because knowledge and information are being generated much faster than they can be effectively analyzed. Computable archaeology-if you do not like the expression "automatic archaeology"-is the proper way of exploring new ways of answering the questions we have not yet answered.

Optical Coherence Tomography (OCT) emerged as a technology for the detection of retinal disease. Owing to its excellent performance and ability to provide in-vivo high-resolution images, the popularity increased dramatically across various application domains. Consequently, OCT has been widely used in other branches of medical applications, i.e., oncology and otolaryngology, industry, and agriculture. Despite its widespread use, OCT image analysis is an inherently subjective, laborious, and time-intensive task that requires expertise. Deep learning (DL) stands as the current state-of-the-art method for image analysis. Hence, several research groups have directed their efforts toward incorporating DL algorithms with OCT imaging to reduce the time as well as the subjectivity. This article comprehensively reviews the principal technological advancements in DL methods applied across various OCT applications. Additionally, it explores the latest trends in developing DL methods for OCT, highlights their limitations, and discusses future opportunities in a comprehensive manner.

Günümüzde hızla gelişmekte olan uydu teknolojileri ve uzaktan algılama tekniklerinden tarım uygulamalarında sıklıkla ve işlerlikle faydalanılmaktadır. Uydulardan elde edilen görüntüler, sahip olduğu yüksek mekânsal, radyometrik ve spektral çözünürlükleri sayesinde tarım alanlarının gözlenmesi ve arazi kullanım durumlarının tespit edilmesinde kullanılmaktadır. Uydu görüntülerinin sınıflandırılması sonucunda elde edilen veriler sayesinde arazi kullanım şekli ve örtüsü hakkında bilgi edinilebilmektedir. Ülkemiz için stratejik öneme sahip pamuk bitkisinin uydu görüntüleri yardımı ile tespiti için gerçekleştirilen bu çalışmada 5 bantlı RapidEye uydu görüntüleri kullanılmıştır. Çalışma alanında görüntü elde edilme tarihi ile eş zamanlı olarak arazi çalışmaları yapılmıştır. Arazi çalışmalarında yer örnekleri tespit edilmiş ve koordinatları GPS yard ımı ile belirlenmiştir. Böylece eğitim ve test verileri için arazi çalışması tamamlanmıştır. Elde edilen uydu görüntüsü eğitim verileri ile nesne tabanlı yöntem kullanılarak sınıflandırılmıştır. Test verileri ile de sınıflandırılmış görüntünün doğruluk analizi gerçekleştirilip güvenilirliği test edilmiştir. Pamuk bitkisinin tespitinde, nesne tabanlı sınıflandırma yöntemi ile elde edilen sınıflandırılmış sonuç 1 bitkisinin tespitinde, nesne tabanlı sınıflandırma yöntemi ile elde edilen sınıflandırılmış sonuç görüntünün genel doğruluğu % 98,19 olarak bulunmuştur.

Görüntü Bölütleme, resim içinde birbirinden kolayca ayrılabilen ve kendi içinde homojen olabilen farklı bölgelerin elde edilmesinde kullanılan zor bir tekniktir. Bu makale yerden 5 ila 30 metre yükseklikte bir drone tarafından elde edilen 6000*4000 pixel boyutundaki kuşbakışı görüntülerinin kentsel görüntülerim anlamsal olarak bölütlemesini sunar. Bölütleme için Fuzzy C-Means (FCM) algoritması kullanılmıştır. Görüntüler üzerinde herhangi bir ön işleme yapmadan elde edilen bölütleme sonuçları Dice, Jaccard ve Mutual Information, benzerlik ölçütleri ile değerlendirilmiştir. Elde edilen sonuçlara göre FCM algoritması kentsel görüntüleri başarılı bir şekilde bölütleme yapmaktadır.

Padang lamun mampu menyerap CO 2 sebagai gas penyebab pemanasan global. Penginderaan jauh dapat mengestimasi karbon yang diserap lamun dengan menggunakan persamaan regresi linier antara citra satelit spektral dengan nilai karbon atas lamun di lapang. Tujuan penelitian ini melihat kemampuan citra Sentinel-2 dalam memetakan habitat padang lamun dan mengestimasi stok karbon atas padang lamun di Gili Petarando, Lombok Timur. Metode penelitian menggunakan klasifikasi terbimbing algoritma maximum likelihood, algoritma NDVI, regresi linier, dan confusion matrix. Hasil klasifikasi terbimbing menghasilkan akurasi pemetaan habitat yang baik yaitu 69.44%. Area padang lamun mendominasi habitat bentik Gili Petarando dengan total 167.3 Ha yang menurut tingkat persen penutupan terbagi atas lamun padat yaitu 101.89 Ha, lamun sedang yaitu 50.29 Ha, dan lamun jarang yaitu 15.13 Ha. Terdapat korelasi tinggi pada regresi linier antara spektral band DII hijau merah (DII34) (x) dan nilai karbon lamun (y) yang dirumuskan dalam persamaan linier y = 151.41x -1770.8. Estimasi total stok karbon atas padang lamun di Gili Petarando adalah 9.33 tonC dengan kesalahan standar estimasi tergolong rendah yaitu 1.33 gC/m 2 yang menunjukkan estimasi yang baik. Hal ini membuktikan citra Sentinel-2 mempunyai kemampuan yang baik untuk digunakan sebagai alat untuk memetakan padang lamun dan mampu mengestimasi nilai stok karbon atas padang lamun.

This study presents experimental investigations on supervised ensemble classification for land cover classification. Despite the arrays of classifiers available in machine learning to create an ensemble, knowing and understanding the correct classifier to use for a particular dataset remains a major challenge. The ensemble method increases classification accuracy by consulting experts taking final decision in the classification process. This study generated various land cover maps, using image classification. This is to authenticate the number of classifiers that should be used for creating an ensemble. The study exploits feature selection techniques to create diversity in ensemble classification. Landsat imagery of Kampala (the capital of Uganda, East Africa), AVIRIS hyperspectral dataset of Indian pine of Indiana and Support Vector Machines were used to carry out the investigation. The research reveals that the superiority of different classification approaches employed depends on the datasets used. In addition, the pre-processing stage and the strategy used during the designing phase of each classifier is very essential. The results obtained from the experiments conducted showed that, there is no significant benefit in using many base classifiers for decision making in ensemble classification. The research outcome also reveals how to design better ensemble using feature selection approach for land cover mapping. The study also reports the experimental comparison of generalized support vector machines, random forests, C4.5, neural network and bagging classifiers for land cover classification of hyperspectral images. These classifiers are among the state-ofthe-art supervised machine learning methods for solving complex pattern recognition problems. The pixel purity index was used to obtain the endmembers from the Indiana pine and Washington DC mall hyperspectral image datasets. Generalized reduced gradient optimization algorithm was used to estimate fractional abundance in the image dataset thereafter obtained numeric values for land cover classification. v The fractional abundance of each pixel was obtained using the spectral signature values of the endmembers and pixel values of class labels. As the results of the experiments, the classifiers show promising results. Using Indiana pine and Washington DC mall hyperspectral datasets, experimental comparison of all the classifiers' performances reveals that random forests outperforms the other classifiers and it is computational effective. The study makes a positive contribution to the problem of classifying land cover hyperspectral images by exploring the use of generalized reduced gradient method and five supervised classifiers. The accuracy comparison of these classifiers is valuable for decision makers to consider tradeoffs in method accuracy versus complexity. The results from the research has attracted nine publications which include, six international and one local conference papers, one published in Computing Research Repository (CoRR), one Journal paper submitted and one Springer book chapter, Abe et al., 2012 obtained a merit award based on the reviewer reports and the score reports of the conference committee members during the conference period. vi Dedication This thesis is dedicated to the Almighty God, the Omnipotent, the Merciful and Omniscience Father. Who reigns and rules in the affairs of men forever and ever at all times. It is also dedicated to my beloved father Late Francis Olufunmilayo Akinmade. for the consistent encouragement and understanding shown to me during the course of this program. All the time you sacrificed to attend to my work painstakingly going through my work and offering useful suggestions. I express my gratitude to my co-

Mikroskopta incelenen karbonat kayalarda dokunun türünün belirlenmesi jeolojik araştırmalarda önemlidir. Petrol sondajlarında olduğu gibi kayaç örneklerinin hızlı bir şekilde değerlendirilmesi gereken durumlarda bu örneklerin tanımlanması görüntü işleme teknikleri ve yapay sinir ağları kullanılarak hızlı bir şekilde yapılabilir. Daha önceki bir çalışmada, bu yöntem kullanıldığı zaman kireçtaşı örneğinin, Dunham (1962) sınıflamasında çamurtaşı, vaketaşı, istiftaşı, tanetaşı olarak sınıflanan örneklerin yapay sinir ağları kullanılarak % 93 oranında dokunun doğru olarak belirlenebileceği yayınlanmıştır. Bu çalışma ise daha önce denenmiş yukarıdaki yöntemin İzmit ve çevresindeki Akveren formasyonu kayaç örneklerine uygulanmasıdır. Akveren formasyonundan alınan kayaç örneklerinden 45 adet ince kesit yapılmıştır. Bu ince kesitlerin fotoğrafları çekilerek bu görüntüler bilgisayarda MATLAB programında "Sobel" filtresi kullanılarak görüntü işleme çalışması ve sonra yapay sinir ağları kullanılarak yukarıdaki 4 sınıftan hangisine yerleşeceği konusu irdelenmiş ve mikroskopta belirlenen doku sınıflamasıyla karşılaştırılmıştır.

Penelitian ini bertujuan untuk : 1) Memetakan obyek penutup lahan skala detil dengan metode interpretasi visual dan klasifikasi berbasis obyek 2) Analisis kemampuan klasifikasi berbasis obyek dalam memetakan obyek penutup lahan. Penelitian ini menggunakan citra GeoEye-1 tahun perekaman 2009. Klasifikasi berbasis obyek menggunakan algoritma multiresolution segmentation untuk tahap segmentasi, dan nearest neighbor untuk tahap klasifikasi. Hasil klasifikasi berbasis obyek akan dibandingkan dengan hasil interpretasi visual kemudian dianalisis untuk mendapatkan informasi kemampuan klasifikasi berbasis obyek dalam memetakan obyek penutup lahan. Hasil perbandingan klasifikasi memperlihatkan klasifikasi berbasis obyek hanya mampu memetakan obyek yang sama dengan klasifikasi dengan interpretasi visual sebesar 46,37% pada skema klasifikasi tingkat IV dan 71,05% pada tingkat I. Hasil ini memperlihatkan bahwa kemampuan klasifikasi berbasis obyek dipengaruhi oleh jumlah obyek penutup lahan yang ...

Coastal marsh ecosystems are changing and being lost at a rapid rate around the world. One of the fastest rates of land loss, specifically coastal marsh habitat, occurs in Louisiana on the Northern Gulf Coast of the USA. To address this issue, state and federal agencies have undertaken massive wetland restoration efforts to preserve and restore coastal marsh habitats in Louisiana. For these efforts to be successful in the long-term, it is critical to understand what methodologies and techniques result in resilient restoration projects. However, traditional methods to monitor restoration success rely on labor intensive field measurements that are often limited in scope, difficult to maintain, and underfunded. Recent technological developments with uncrewed aircraft systems (UASs) and image processing have substantially improved the ability of restoration practitioners to use off-the-shelf UASs and cameras to map projects. We present a streamlined method using a commercially available drone with a high-resolution red, green, blue (RGB) camera to assess the effects of wetland restoration and integrate more modern tools into evaluation approaches. We conducted drone flights at restored brackish marshes of various ages using a space for time substitution with the goal of understanding the long-term success of marsh restoration. We observed that created marshes had higher land to water ratios than natural marshes. This finding suggests that these restored areas were gaining and maintaining elevation after approximately 10 years. Our method shows that drone surveys offer low-cost, minimally invasive methods for evaluating restored wetlands and ultimately tell us more about ecosystem function through realistic site-level habitat configurations.