Extração de contornos de telhados de edifícios a partir da integração de imagem aérea de alta-resolução e dados LASER, utilizando campos aleatórios de Markov
Carregando...
Data
2016-12-19
Autores
Fernandes, Vanessa Jordão Marcato [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Esse trabalho propõe o desenvolvimento de um método para a extração automática de contornos de telhados de edifícios com a combinação de dados de Varredura a LASER Aerotransportado (VLA) e dados fotogramétricos e campos aleatórios de Markov (MRF). Inicialmente, um Modelo Digital de Superfície normalizado (MDSn) é gerado através da diferença entre o Modelo Digital de Superfície (MDS) e o Modelo Digital de Terreno (MDT), obtidos a partir da nuvem de pontos LASER. Em seguida, o MDSn é segmentado para a obtenção dos polígonos que representam objetos altos da cena. Esses polígonos são projetados na imagem para restringir o espaço de busca para a segmentação da imagem em regiões. Esse processo possibilita a extração de polígonos na imagem que representem objetos altos. O processo de identificação de contornos de telhados, em meio aos objetos altos detectados na imagem, na etapa anterior, é realizado através da otimização de uma função de energia estabelecida com base em MRF que modela propriedades específicas de contornos de telhados de edifícios. No modelo MRF são utilizados tanto os polígonos extraídos da imagem quanto os extraídos dos dados VLA. A função de energia é otimizada pelo método Algoritmo Genético (AG). O método proposto nesse trabalho foi avaliado com base em dados reais - imagens aéreas de alta resolução e dados VLA. Os resultados obtidos na avaliação experimental mostraram que a metodologia funciona adequadamente na tarefa de extrair os contornos de telhados de edifícios. A função de energia proposta associada ao método de otimização AG diferenciou corretamente os contornos de telhados de edifícios dos demais objetos altos presentes nas cenas. Os contornos de telhados extraídos apresentam boa qualidade, o que é evidenciado por meio dos índices de completeza e correção obtidos pela avaliação numérica. Com base nos índices médios obtidos para cada experimento, têm-se as médias de completeza e correção para os experimentos iguais a 90,96% e 98,99%, respectivamente. Os valores máximos de completeza e correção são de 99,19% e 99,94%, respectivamente, e os valores mínimos de 78,08% e 97,46%, respectivamente. Os menores valores de completeza estão associados às áreas de oclusão por vegetação e presença de sombras.
This paper proposes a method for the automatic extraction of building roof contours through a combination of Airborne Laser Scanner (ALS) and photogrammetric data, and Markov Random Field (MRF). Initially, a normalized digital surface model (nDSM) is generated on the basis of the difference between the digital surface model and the digital terrain model, obtained from the LiDAR point cloud. Then the nDSM is segmented to obtain the polygons representing aboveground objects. These polygons are projected onto image to restrict the search space for image segmentation into regions. This process enables the extraction of polygons in the image representing aboveground objects. Building roof contours are identified from among the aboveground objects in the image by optimizing a Markov-random-field-based energy function that embodies roof contour specific properties. In the MRF model are used both polygons extracted from image and from ALS data. The energy function is optimized by the Genetic Algorithm (GA) method. The method proposed in this work was evaluated based on real data - high-resolution aerial images and ALS data. The results obtained in the experimental evaluation showed that the methodology works adequately in the task of extracting the contours of building roofs. The proposed energy function associated with the GA optimization method correctly differentiated the building roof contours from the other high objects present in the scenes. The extracted roof contours show good quality, which is evidenced by the indexes of completeness and correctness obtained by numerical evaluation. Based on the mean indexes obtained for each experiment, the average completeness and correctness for the experiments were equal to 90.96% and 98.99%, respectively. The maximum completeness and correctness values are 99.19% and 99.94%, respectively, and the minimum values are 78.08% and 97.46%, respectively. The lowest values of completeness are associated to the vegetation occlusion areas and presence of shadows.
This paper proposes a method for the automatic extraction of building roof contours through a combination of Airborne Laser Scanner (ALS) and photogrammetric data, and Markov Random Field (MRF). Initially, a normalized digital surface model (nDSM) is generated on the basis of the difference between the digital surface model and the digital terrain model, obtained from the LiDAR point cloud. Then the nDSM is segmented to obtain the polygons representing aboveground objects. These polygons are projected onto image to restrict the search space for image segmentation into regions. This process enables the extraction of polygons in the image representing aboveground objects. Building roof contours are identified from among the aboveground objects in the image by optimizing a Markov-random-field-based energy function that embodies roof contour specific properties. In the MRF model are used both polygons extracted from image and from ALS data. The energy function is optimized by the Genetic Algorithm (GA) method. The method proposed in this work was evaluated based on real data - high-resolution aerial images and ALS data. The results obtained in the experimental evaluation showed that the methodology works adequately in the task of extracting the contours of building roofs. The proposed energy function associated with the GA optimization method correctly differentiated the building roof contours from the other high objects present in the scenes. The extracted roof contours show good quality, which is evidenced by the indexes of completeness and correctness obtained by numerical evaluation. Based on the mean indexes obtained for each experiment, the average completeness and correctness for the experiments were equal to 90.96% and 98.99%, respectively. The maximum completeness and correctness values are 99.19% and 99.94%, respectively, and the minimum values are 78.08% and 97.46%, respectively. The lowest values of completeness are associated to the vegetation occlusion areas and presence of shadows.
Descrição
Palavras-chave
Fotogrametria, Extração de telhados, Dados VLA, Imagens aéreas, MRF, Photogrammetry, Roofs extraction, ALS Data, Aerial image