Automatic Detection of Building Contours using a small format digital camera and LIDAR data

• Autor(a) principal: Machado, Alvaro Muriel Lima
• Data de Publicação: 2006
• Outros Autores: Mitishita, Edson A.
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Boletim de Ciências Geodésicas
• Texto Completo: https://revistas.ufpr.br/bcg/article/view/7240
• topic: Monorestituição; erros sistemáticos; processamento digital de imagens; Monoplotting; systematic errors; digital image processing

This work presents a methodology to automatically extract 2D building contours from the integration between a small format digital camera image and laser scanning raw data. The external image orientation and internal camera orientation parameters are assumed to be known. As a common space to integrate  information from both sources it is used the image space without distortion. To remove systematic errors in the original image, the Newton-Raphson iterative method for simultaneous equation is used in a resampling process to this space. At this step, allthe pixels have color and altitude associated information. The RGB color values are transformed to (almost isotropic) CIELUV color space and the image is segmented according to these values, by mean shift algorithm. The forward steps filter the gotten segments, using colors (to remove green vegetation), medium altitude of a region comparing it with the altitude of its neighbours (to preserve high regions), and polygonal simplification degree (Douglas-Peucker), attempting to detect linear features. Several experiments realized over a Centro Politécnico - UFPR area, inCuritiba/PR, are discussed, with good results for isolated buildings.