Automação do caminhamento fotogramétrico com arranjo triangular de câmeras em trajetória curva
Carregando...
Data
2017-12-16
Autores
Patuci, Tobias Rehder da Cunha
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Um sistema de mapeamento móvel (SMM) permite obter informações geográficas das vias e de seu entorno. O objetivo deste trabalho é controlar a propagação de erros na determinação dos POE no processo de caminhamento fotogramétrico, para isto, foi estudada a adição de uma terceira câmera ao sistema e a integração das imagens com dados dos sensores de posicionamento e navegação. As câmeras foram dispostas em um arranjo triangular, de modo que o terceiro raio homólogo de um ponto fosse não coplanar aos demais raios homólogos deste mesmo ponto. Isto permite melhorar a estimativa da posição geométrica dos pontos, principalmente no sentido transversal à base do levantamento. Para a realização do levantamento, o sistema foi calibrado no campo de calibração da FCT/Unesp. O trajeto percorrido foi no entorno de uma quadra urbana, de modo a percorrer trechos retos e curvos. Os algoritmos SIFT (Scale Invariant Feature Transform) e RANSAC (RANdom SAmple Consensus) foram aplicados nas imagens coletadas pelo sistema, a fim de identificar os pontos homólogos nos trios de imagens e nas imagens subsequentes. Os parâmetros de orientação exterior (POE) das câmeras foram calculados pelo processo de fototriangulação, o ajustamento foi pelo método dos mínimos quadrados (MMQ), modelo paramétrico. Os dados foram processados variandose a quantidade de câmeras, duas ou três, para analisar a contribuição da terceira câmera. Os experimentos foram realizados para trechos retos e curvos e variou-se o tamanho do bloco processado e o avanço de base. Além disso, foi analisada a diferença entre o processamento em lote e sequencial. O controle de qualidade foi realizado pela análise da raíz do erro médio quadrático, comparando-se as coordenadas obtidas na fototriangulação com as coordenadas medidas em campo, em um levantamento topográfico. Os experimentos mostraram que com a adição da terceira câmera foram obtidos melhores resultados, principalmente para a trajetória curva.
A mobile mapping system (MMS) provides geographical information of the roads and their surroundings. When carried by pedestrians, it can provide information of indoor areas or inaccessible to vehicles. The objective of this work was to control the spread of errors in determining the EOP in photogrammetric traversal process to this, the addition of a third camera to the system and the integration of images and positioning and navigation sensors data were studied. Cameras were mounted in a triangular arrangement, in a way that the third homologous beam of a point is non-colinear to the others homologous beams of the same point. This enables a point position estimative enhancement, mainly in the transverse direction of the base. For the survey, the system was calibrated in the FCT/Unesp calibration field. The route taken was in the vicinity of an urban block, passing by straight and curved sections. SIFT algorithm (Scale Invariant Feature Transform) and RANSAC (RANdom SAmple Consensus) were applied on the images collected by the system to identify the corresponding points in image trios and subsequent images. The exterior orientation parameters (EOP) of the cameras were calculated by the phototriangulation process, the adjustment method used was the least squares (LSM), parametric model. The data were processed by varying the number of cameras, two or three, to analyze the contribution of the third camera. The experiments were realized for straight and curved sections and was varied the block size and base advance. Moreover, it was analysed the difference between sequential and full block calculation. Quality control was held by root mean square error analysis, comparing coordinates obtained in phototriangulation with object coordinates obtained by topographic survey. Finally, the experiments presented better results with the third camera addition, mainly for the curved section.
A mobile mapping system (MMS) provides geographical information of the roads and their surroundings. When carried by pedestrians, it can provide information of indoor areas or inaccessible to vehicles. The objective of this work was to control the spread of errors in determining the EOP in photogrammetric traversal process to this, the addition of a third camera to the system and the integration of images and positioning and navigation sensors data were studied. Cameras were mounted in a triangular arrangement, in a way that the third homologous beam of a point is non-colinear to the others homologous beams of the same point. This enables a point position estimative enhancement, mainly in the transverse direction of the base. For the survey, the system was calibrated in the FCT/Unesp calibration field. The route taken was in the vicinity of an urban block, passing by straight and curved sections. SIFT algorithm (Scale Invariant Feature Transform) and RANSAC (RANdom SAmple Consensus) were applied on the images collected by the system to identify the corresponding points in image trios and subsequent images. The exterior orientation parameters (EOP) of the cameras were calculated by the phototriangulation process, the adjustment method used was the least squares (LSM), parametric model. The data were processed by varying the number of cameras, two or three, to analyze the contribution of the third camera. The experiments were realized for straight and curved sections and was varied the block size and base advance. Moreover, it was analysed the difference between sequential and full block calculation. Quality control was held by root mean square error analysis, comparing coordinates obtained in phototriangulation with object coordinates obtained by topographic survey. Finally, the experiments presented better results with the third camera addition, mainly for the curved section.
Descrição
Palavras-chave
Mapeamento móvel, Integração de sensores, Caminhamento fotogramétrico, Fototriangulação, Mobile mapping, Sensor integration, Photogrammetric bridging, Phototriangulation