Nova metodologia de localização de regiões candidatas em imagens digitais utilizando arquiteturas reconfiguráveis

• Ano de defesa: 2007
• Autor(a) principal: Pacheco, Márcio Alexandre
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria; BR; Engenharia Elétrica; UFSM; Programa de Pós-Graduação em Engenharia Elétrica
• Palavras-chave em Português: Regiões candidatas; FPGA; Processamento de imagem; Candidate regions; Image processing; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: http://repositorio.ufsm.br/handle/1/8590

This work involves the study and the implementation of different techniques of image processing in software and hardware , aiming the implementation of a new methodology of digital images location of candidate regions . Both implementations will be compared with the objective of verifying the obtained performance during the execution of the algorithms in PC and FPGA XILINX Spartan3 model3s400ft256.®. This task will be performed by applying some morphological operators, like erosion, dilatation and gradient to obtain only the borders of the processed image. After the conclusion of this stage location of candidate regions technique is applied aiming to restrict correctly only the x and y coordinates that may contain the vehicle plate or the traffic sign. The methodology of location of candidate regions was evaluated in software on a base of 823 images with varied dimensions. This technique is based on the analysis of the signature of the vehicle plate or traffic sign that are obtained through the process of border extraction. After concluding the validation of the above mentioned techniques in software the mapping out of these ones is done in a reconfigurable architecture. The results obtained from the execution of the implemented descriptions in hardware were compared with the ones obtained in software to evaluate the performance between the PC and FPGA types platforms. It is important to point out that the methodology of location of the candidate regions in digital images was developed is protected by the rights of intellectual property under the number 0000270607032603