Localização automática de pontos de controle em imagens aéreas baseada em cenas terrestres verticais

• Ano de defesa: 2014
• Autor(a) principal: Berveglieri, Adilson [UNESP]
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Cartografia; Fotogrametria; Imagens digitais; Sistemas de navegação global por satélite; Cartography
• Link de acesso: http://hdl.handle.net/11449/122121

Control field is essential in image orientation procedures and is based on the determination of ground coordinates from distinguishable features. Aerial survey techniques have evolved with the integration of Global Navigation Satellite System (GNSS) and inertial systems, enabling Direct Sensor Orientation (DSO). However, even with the development and consolidation of these techniques, Ground Control Points (GCPs) are still needed to correct for systematic errors and to assess the reliability and accuracy of the orientation data. The location of GCPs over images is still a task performed by an operator, which is an additional source of errors due to the uncertainties in the visual identification. Thus, an automatic flow is proposed in this work to obtain GCP coordinates and locate them in aerial images with sub-pixel accuracy. The approach is based on vertical ground scenes from distinguishable feature areas with a point surveyed by GNSS receiver (like a GCP). These ground scenes are matched with aerial image patches to transfer surveyed points. Feature- and area-based matching algorithms are applied for this task as well as a subsequent adaptive refinement by least squares. To perform the automatic flow, a technique to acquire ground images as required was designed. In addition, other techniques were developed to: prepare control scenes by rectification and orthorectification; locate GCPs into search window in aerial images, and minimize the differences between matched image patches, improving the sub-pixel location of the match point. Different experiments assessed the feasibility and accuracy achieved with the proposed techniques using bundle adjustment in stereo-models and image blocks. The analysis of results was based on discrepancies and residuals at GPCs and independent checkpoints...