TCC - Bacharelado em Ciência da Computação (Sede)
URI permanente para esta coleçãohttps://arandu.ufrpe.br/handle/123456789/415
Navegar
1 resultados
Resultados da Pesquisa
Item Detecção de linhas que separam o mar da areia e o mar do céu em imagens de praia(2018) Silva, Jeremias Leite da; Macário Filho, Valmir; http://lattes.cnpq.br/4346898674852080The border of the metropolitan region of Pernambuco has been presented several incidents from sharks and some of these cases were fatal. In order to reduce the incidents, this work proposes an approach for the segmentation of the sea as part of a monitoring system for bathers through cameras. Once one or more persons are identified within the risk zone, the system will issue an alert to central monitoring, and the nearest lifeguard would be alerted to move to the location. In order for the system to identify the people in the image, the system must be able to identify the beach region in the image in order to segment them and identify the bathers. The strip of water is formed by two borders, one with the sky and the other with sand. The boundary with the sky is a straight horizontal line called the horizon line, and the boundary with the sand is a contour formed by the boundary of the water with the sand that is called the shoreline. This work aims to propose algorithms for the detection of the horizon and coastlines for sea segmentation, which represents one of the main steps for the monitoring system of beach bather images. In this work four horizon-line detection algorithms were analyzed to evaluate which of these obtains the best detection result. Two state-of-the-art algorithms were analyzed: Lie et al. and that of Ahmad et al. Both are works for the detection of the horizon line in mountain images. Two other algorithms were contributions of this work: the Canny Edge Detection and Multi-Stage Graph Detection (DLHCGME) and Detection of Horizon Line with Sobel and Hough Transform (DLHSTH). Two new algorithms were proposed for Coastline Detection: Coastline Detection with Canny Edge Detection and Multi-Stage Graphs (DLCCGME) and Coastline Detection based on Contour of the Hue Channel (DLCCCH). In the detection of the horizon line in images without occlusions the experiments show that the DLHCGME obtained the best result with an error rate of 0.47 and the second was the DLHSTH with 1.11 and for images with occlusions the DLHSTH obtained the best result with an error rate of 1.98 and the DLHCGME was the second best result with 2.62.