The aim of this study is to develop an automated model to extract buildings from aerial images. The fuzzy k-NN classification method is used to extract the buildings by using color information. Also in the thesis, the advantages of the relevance feedback systems are discussed. The software, BuildingLS, is developed in C#. The model is evaluated in 5 different test areas with more than 700 building.

Submitted by GILMAR RENAN KISAKI OLIVEIRA null (renan.kisaki@gmail.com) on 2016-12-20T03:09:27Z No. of bitstreams: 1 2016_MSc_Gilmar_Renan_Unesp.pdf: 3787156 bytes, checksum: 03bf116759b5ae85b2e5302f167eb985 (MD5) / Approved for entry into archive by Felipe Augusto Arakaki (arakaki@reitoria.unesp.br) on 2016-12-21T19:54:13Z (GMT) No. of bitstreams: 1 oliveira_grk_me_prud.pdf: 3787156 bytes, checksum: 03bf116759b5ae85b2e5302f167eb985 (MD5) / Made available in DSpace on 2016-12-21T19:54:13Z (GMT). No. of bitstreams: 1 oliveira_grk_me_prud.pdf: 3787156 bytes, checksum: 03bf116759b5ae85b2e5302f167eb985 (MD5) Previous issue date: 2016-02-29 / Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) / Esta dissertação contempla o desenvolvimento de um método que combina os dados LiDAR (Light Detection And Ranging) obtidos por sistema de varredura LASER (Light Amplification by Stimulated Emission of Radiation) e imagens aéreas de uma mesma região a fim de extrair os contornos de telhados de edificações, onde os parâmetros de orientação das imagens são conhecidos. O método proposto neste trabalho pode ser dividido nas seguintes etapas: extração das edificações nos dados LiDAR; extração dos contornos das edificações nos dados LiDAR; e refinamento dos contornos das edificações integrando dados LiDAR e imagens. Primeiramente, as edificações são extraídas dos dados LiDAR, seguida da determinação dos seus pontos de contorno, que por sua vez, são projetados em duas imagens que formam um modelo estereoscópico. Às imagens do par é aplicado o algoritmo de detecção de bordas de Canny com o objetivo de identificar as bordas de edificações. Tendo os contornos dos telhados de edificações (provenientes dos dados LiDAR) projetados nas imagens de bordas, é realizado um procedimento de busca dos pontos de bordas de edificações nas imagens. Com base nos pixels identificados como bordas de edificações e com o propósito de obter uma figura geométrica que represente os contornos, é aplicado o ajuste de retas 2D pelo Método dos Mínimos Quadrados (MMQ) integrado à filtragem de pontos espúrios por meio do teste Tau. Para avaliar o método proposto e implementado foram utilizados dados LiDAR com densidade média de 6,7 pontos/m² e imagens aéreas digitais com GSD de 8 cm. Os resultados obtidos na avaliação dos experimentos mostraram que o método proposto conseguiu extrair os contornos dos telhados, com melhores resultados para edificações isoladas que não possuíam projeção de sombras ou objetos sobre elas, atingindo valores da ordem de 0,97 GSD e 1,80 GSD, para o REMQ em planimetria e altimetria, respectivamente. / This dissertation proposes a methodology to extract building boundary through the integration of LiDAR data and aerial imagery where the image orientation parameters are known. The proposed method can be divided into following steps: building extraction from LiDAR data; building boundary extraction from LiDAR data; and refinement of building boundary through the integration of LiDAR data and optical imaging. Building are first extracted from LiDAR data, then building boundaries are determined in LiDAR data and projected onto the stereo pair of aerial images. These aerial images are results from the application of Canny edge detector in order to identify building boundaries from images. Since the 3D building boundaries (determined from LiDAR data) are projected onto the Canny images, a search mechanism is performed to find the building edge points in these images. A 2D line adjustment by Least Squares Method (LSM) is performed, followed by outlier detection based on Tau Statistical Test, for generating a geometric shape to represent the buildings through the building edge pixels identified. In order to evaluate the proposed approach, LiDAR data with approximate density of 6.7 pts/m² and digital aerial images with GSD around 8 cm were used. The results showed that the proposed method enabled to extract building roof boundaries with best results for isolated buildings without objects or shadow’s projection on them with the root mean square error (RMSE) around 0.97 GSD and 1.80 GSD in planimetry and altimetry, respectively. / CNPq: 130473/2013-8

The essential procedure of constructing 3-D building models in urban areas is to extract the building boundary footprint. In the past researches, the common procedures used in extracting the building footprint are applying edge detection, vectorization, and generalization. However, the derived boundary lines occasionally occur zigzag patterns, thus, it still needs further building footprint regularization. This study proposed a new approach in the point of view that the points, lines and polygons are the essential elements in reconstructing 3-D building models. The proposed new method is based on ¡§corner detection approach (CDA)¡¨ and ¡§Adjustment of building footprints and corner points (ABFCO)¡¨ algorithm on Light Detection And Ranging (LiDAR) or binary classification resultant imagery. This study implements Harris and Local Binary Pattern (LBP) corner detection, afterward, connects all detected points by using convex hull algorithm. However, ortho-non-rectangle buildings would compose poor outlines after convex hull. This study combines open and dilation morphology with the find ignored point algorithm to improve any incorrect connections. Finally, performs the ABFCO algorithm to those points which belong to the same boundary to generalize a line segment, and to figure out the intersections and boundary lines of the buildings. The experiment results have proved that the overall accuracy of LBP corner detection is about 3.5% higher than Harris corner detection, its overall accuracy is about 92% in rectangular buildings and about 91% in non-rectangular buildings, its standard deviation of boundary length is 0.29m and better than Harris¡¦s 0.55m. We also compared LBP corner detection with edge detection. The overall accuracy of corner detection is about 3% higher than edge detection, standard deviation of boundary length 0.37m is also better than edge detection 0.75m. This study not only proved the corner detection is better than edge detection from data, but also developed ABFCO algorithm is helpful for extracting more accurate building footprint lines.

Extraction of buildings from remote sensing sources is an important GIS application and has been the subject of extensive research over the last three decades. An accurate building inventory is required for applications such as GIS database maintenance and revision; impervious surfaces mapping; storm water management; hazard mitigation and risk assessment. Despite all the progress within the fields of photogrammetry and image processing, the problem of automated feature extraction is still unresolved. A methodology for automatic building extraction that integrates remote sensing sources and GIS data was proposed. The methodology consists of a series of image processing and spatial analysis techniques. It incorporates initial simplification procedure and multiple feature analysis components. The extraction process was implemented and tested on three distinct types of buildings including commercial, residential and high-rise. Aerial imagery and GIS data from Shelby County, Tennessee were identified for the testing and validation of the results. The contribution of each component to the overall methodology was quantitatively evaluated as relates to each type of building. The automatic process was compared to manual building extraction and provided means to alleviate the manual procedure effort. A separate module was implemented to identify the 2D shape of a building. Indices for two specific shapes were developed based on the moment theory. The indices were tested and evaluated on multiple feature segments and proved to be successful. The research identifies the successful building extraction scenarios as well as the challenges, difficulties and drawbacks of the process. Recommendations are provided based on the testing and evaluation for future extraction projects.

The recent growth in inexpensive laser scanning sensors has created entire fields of research aimed at processing this data. One application is determining the polygonal boundaries of roofs, as seen from an overhead view. The resulting building outlines have many commercial as well as military applications. My work in this area has created a segmentation algorithm where the descriptive features are computationally and theoretically simpler than previous methods. A support vector machine is used to segment data points using these features, and their use is not common for roof detection to date. Despite the simplicity of the feature calculations, the accuracy of our algorithm is similar to previous work. I also describe a basic polygonal extraction method, which is acceptable for basic roofs.

<p>Buildings are one of the fundamental sources of geospatial information for urban planning, population estimation, and infrastructure management. Although building extraction research has gained considerable progress through neural network methods, the labeling of training data still requires manual operations which are time-consuming and labor-intensive. Aiming to improve this process, this thesis developed an automated building extraction method based on the boundary following technique and the Mask Regional Convolutional Neural Network (Mask R-CNN) model. First, assisted by known building footprints, a boundary following method was used to automatically best label the training image datasets. In the next step, the Mask R-CNN model was trained with the labeling results and then applied to building extraction. Experiments with datasets of urban areas of Bloomington and Indianapolis with 2016 high resolution aerial images verified the effectiveness of the proposed approach. With the help of existing building footprints, the automatic labeling process took only five seconds for a 500*500 pixel image without human interaction. A 0.951 intersection over union (IoU) between the labeled mask and the ground truth was achieved due to the high quality of the automatic labeling step. In the training process, the Resnet50 network and the feature pyramid network (FPN) were adopted for feature extraction. The region proposal network (RPN) then was trained end-to-end to create region proposals. The performance of the proposed approach was evaluated in terms of building detection and mask segmentation in the two datasets. The building detection results of 40 test tiles respectively in Bloomington and Indianapolis showed that the Mask R-CNN model achieved 0.951 and 0.968 F1-scores. In addition, 84.2% of the newly built buildings in the Indianapolis dataset were successfully detected. According to the segmentation results on these two datasets, the Mask R-CNN model achieved the mean pixel accuracy (MPA) of 92% and 88%, respectively for Bloomington and Indianapolis. It was found that the performance of the mask segmentation and contour extraction became less satisfactory as the building shapes and roofs became more complex. It is expected that the method developed in this thesis can be adapted for large-scale use under varying urban setups.</p>

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第18654号 / 工博第3963号 / 新制||工||1610(附属図書館) / 31568 / 京都大学大学院工学研究科都市環境工学専攻 / (主査)教授 田村 正行, 准教授 須﨑 純一, 准教授 横松 宗太 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

Approaches were developed for building extraction and updating from high resolution satellite imagery. The developed approaches include two main stages: (i) detecting the building patches and (ii) delineating the building boundaries. The building patches are detected from high resolution satellite imagery using the Support Vector Machines (SVM) classification, which is performed for both the building extraction and updating approaches. In the building extraction part of the study, the previously detected building patches are delineated using the Hough transform and boundary tracing based techniques. In the Hough transform based technique, the boundary delineation is carried out using the processing operations of edge detection, Hough transformation, and perceptual grouping. In the boundary tracing based technique, the detected edges are vectorized using the boundary tracing algorithm. The results are then refined through line simplification and vector filters. In the building updating part of the study, the destroyed buildings are determined through analyzing the existing building boundaries and the previously detected building patches. The new buildings are delineated using the developed model based approach, in which the building models are selected from an existing building database by utilizing the shape parameters. The developed approaches were tested in the Batikent district of Ankara, Turkey, using the IKONOS panchromatic and pan-sharpened stereo images (2002) and existing vector database (1999). The results indicate that the proposed approaches are quite satisfactory with the accuracies computed in the range from 68.60% to 98.26% for building extraction, and from 82.44% to 88.95% for building updating.

With the advent of very high spatial resolution (VHR) satellite, spatial details within the image scene have increased considerably. This led to the development of object-based image analysis (OBIA) for the analysis of VHR satellite images. Image segmentation is the fundamental step for OBIA. However, a large number of techniques exist for RS image segmentation. To identify the best ones for VHR imagery, a comprehensive literature review on image segmentation is performed. Based on that review, it is found that the multiresolution segmentation, as implemented in the commercial software eCognition, is the most widely-used technique and has been successfully applied for wide variety of VHR images. However, the multiresolution segmentation suffers from the parameter estimation problem. Therefore, this study proposes a solution to the problem of the parameter estimation for improving its efficiency in VHR image segmentation. The solution aims to identify the optimal parameters, which correspond to optimal segmentation. The solution to the parameter estimation is drawn from the Equations related to the merging of any two adjacent objects in multiresolution segmentation. The solution utilizes spectral, shape, size, and neighbourhood relationships for a supervised solution. In order to justify the results of the solution, a global segmentation accuracy evaluation technique is also proposed. The solution performs excellently with the VHR images of different sensors, scenes, and land cover classes. In order to justify the applicability of solution to a real life problem, a building detection application based on multiresolution segmentation from the estimated parameters, is carried out. The accuracy of the building detection is found nearly to be eighty percent. Finally, it can be concluded that the proposed solution is fast, easy to implement and effective for the intended applications.

Cette thèse se situe dans le cadre de l'interprétation d'images satellite à haute résolution, et concerne plus spécifiquement la mise à jour de bases de données cartographiques grâce à des images optique et radar à haute résolution. Cette étude présente une chaîne de traitement générique pour la création ou la mise à jour de bases de données représentant les routes ou les bâtiments en milieu urbain. En fonction des données disponibles, différents scénarios sont envisagés. Le traitement est effectué en deux étapes. D'abord nous cherchons les objets qui doivent être retirés de la base de données. La seconde étape consiste à rechercher dans les images de nouveaux objets à ajouter dans la base de données. Pour réaliser ces deux étapes, des descripteurs sont construits dans le but de caractériser les objets d'intérêt dans les images d'entrée. L'inclusion ou élimination des objets dans la base de données est basée sur un score obtenu après fusion des descripteurs dans le cadre de la théorie de Dempster-Shafer. Les résultats présentés dans cette thèse illustrent l'intérêt d'une fusion multi-capteurs. De plus l'intégration aisée de nouveaux descripteurs permet à la chaîne d'être améliorable et adaptable à d'autres objets. / This work takes place in the framework of high resolution remote sensing image analysis. It focuses on the issue of cartographic database creation or updating with optical and SAR images. The goal of this work is to build a generic processing chain to update or create a cartographic database representing roads and buildings in built-up areas. According to available data, various scenarios are foreseen. The proposed processing chain is composed of two steps. First, if a database is available, the presence of each database object is checked in the images. The second step consist of looking for new objects that should be included in the database. To determine if an object should be present in the updated database, relevant features are extracted from images in the neighborhood of the considered object. Those features are based on caracteristics of roads and buildings in SAR and optical images. The object removal/inclusion in the DB is based on a score obtained by the fusion of features in the framework of the Dempster-Shafer evidence theory. Results highlight the interest of multi sensor fusion. Moreover the chosen framework allows the easy integration of new features in the processing chain.