Sistema de rastreamento da mão humana utilizando visão artificial para aplicações embarcadas

• Ano de defesa: 2011
• Autor(a) principal: Freitas, Rodrigo Fernandes
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Teleinformática; Visão por computador; Processamento de imagens
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/10863

In recent years pro cessing capacity of mobile devices have greatly increased, allowing them to pro cess applications, such as games, previously only p ossible on platforms with greater computing p ower. However, the user interface has not accompanied this evolution of computing p ower, still b eing held using non-ergonomic keyb oards. This dissertation prop oses a interaction system for mobile devices based on Computer Vision. This system tracks the user's hand and recognizes six p ossible gestures: p ointing, emph zo om in, emph zo om out, clo ckwise rotation, counterclo ckwise rotation and drag. Initially this system captures images of the user's hand, applies pre-pro cessing lters on the images and segments the region of the skin by thresholding. Then, the contour of the hand is extracted and represented in a vector by a chain co de algorithm. The ngertips are lo cated from the b oundary and represented and, by a set of rules, the gesture made by the user is recognized. The system prop osed is simulated using the Simulink platform and implemented in C ANSI language. In addition, this system is compared with three other systems describ ed in literature based on four evaluation criteria: computacional cost, rotation invariance for the gesture of p ointing, robustness to regions at the background of image with skin-like color and robustness to o cclusion with regions with skin-like color. The results indicate that this system meets the requirements of evaluation criteria, which makes p ossible its use in mobile devices