| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.98 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Os quebra-mares desempenham um papel crucial na proteção costeira, garantindo a segurança
das embarcações em portos e canais de navegação. Para avaliar o seu comportamento e planear a
construção de novas estruturas, o Laboratório Nacional de Engenharia Civil (LNEC) utiliza
modelos físicos 3D compostos por blocos de diferentes formas (cubos, antiferes e tetrápodes),
tamanhos e cores, que simulam as condições reais de ondulação e correntes. Atualmente, a
deteção de anomalias ou pontos fracos nessas estruturas, é feita com recurso a imagens óticas (IO)
e imagens de profundidade (IP), capturadas antes e após o impacto das ondas. As alterações na
superfície são identificadas por subtração de imagens ou seleção manual das faces dos blocos a
serem avaliados. Esta dissertação propõe uma solução de segmentação semi-automática das faces
dos cubos de forma a automatizar e melhorar a qualidade/velocidade do processo. Para obtenção
dos dados foi utilizada a câmara RGB-D Kinect-V2, de modo a minimizar o custo. A metodologia
utiliza os algoritmos Canny Edge Detector (CED) e Hough Transform (HT) para obter as arestas
dos cubos. As arestas são posteriormente processadas para formar uma máscara que delimita as
faces de cada bloco. Através da segmentação das IP com a máscara resultante da aplicação da
metodologia às IO, é possível calcular o plano 3D que melhor se ajusta aos pontos de cada face e
inferir o centro de massa de cada bloco. Os resultados obtidos demonstram os impactos positivos
da segmentação semi-automática, utilizando técnicas de processamento digital de imagem, na
melhoria da precisão e velocidade dos processos de monitorização de modelos físicos de quebramar. Esta abordagem contribui para aprimorar a eficiência e a precisão na avaliação/projeção de
estruturas essenciais na proteção costeira.
Breakwaters play a crucial role in coastal protection, ensuring the safety of vessels in ports and navigation channels. To assess their behaviour and plan the construction of new structures, the National Civil Engineering Laboratory (LNEC) uses 3D physical models made up of blocks of different shapes (cubes, antiphers and tetrapods), sizes and colours, which simulate the real conditions of waves and currents. Currently, anomalies or weak points in these structures are detected using optical images (IO) and depth images (IP), captured before and after wave impact. Changes to the surface are identified by subtracting images or manually selecting the faces of the blocks to be assessed. This dissertation proposes a solution for semi-automatic segmentation of cube faces to automate and improve the quality/speed of the process. The RGB-D Kinect-V2 camera was used to obtain the data, to minimize the cost. The methodology uses the Canny Edge Detector (CED) and Hough Transform (HT) algorithms to obtain the edges of the cubes. The edges are then processed to form a mask that delimits the faces of each block. By segmenting the IP with the mask resulting from applying the methodology to the IO, it is possible to calculate the 3D plane that best fits the points of each face and infer the centre of mass of each block. The results obtained demonstrate the positive impact of semi-automatic segmentation, using digital image processing techniques, in improving the accuracy and speed of monitoring processes for physical models of breakwaters. This approach contributes to improving efficiency and accuracy in the assessment/projection of essential coastal protection structures.
Breakwaters play a crucial role in coastal protection, ensuring the safety of vessels in ports and navigation channels. To assess their behaviour and plan the construction of new structures, the National Civil Engineering Laboratory (LNEC) uses 3D physical models made up of blocks of different shapes (cubes, antiphers and tetrapods), sizes and colours, which simulate the real conditions of waves and currents. Currently, anomalies or weak points in these structures are detected using optical images (IO) and depth images (IP), captured before and after wave impact. Changes to the surface are identified by subtracting images or manually selecting the faces of the blocks to be assessed. This dissertation proposes a solution for semi-automatic segmentation of cube faces to automate and improve the quality/speed of the process. The RGB-D Kinect-V2 camera was used to obtain the data, to minimize the cost. The methodology uses the Canny Edge Detector (CED) and Hough Transform (HT) algorithms to obtain the edges of the cubes. The edges are then processed to form a mask that delimits the faces of each block. By segmenting the IP with the mask resulting from applying the methodology to the IO, it is possible to calculate the 3D plane that best fits the points of each face and infer the centre of mass of each block. The results obtained demonstrate the positive impact of semi-automatic segmentation, using digital image processing techniques, in improving the accuracy and speed of monitoring processes for physical models of breakwaters. This approach contributes to improving efficiency and accuracy in the assessment/projection of essential coastal protection structures.
Descrição
Tese de Mestrado, Engenharia Geoespacial, 2024, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Quebra-mar Monitorização Processamento digital de imagem Hough Transform Canny Edge Detector Teses de mestrado - 2024
