Mobile terrestrial lidar data to detect traffic sign and light pole/ Dados de laser terrestre móvel para detectar placas de sinalização e postes de iluminação

• Autor(a) principal: Peixoto, Elizabete Bugalski de Andrade
• Data de Publicação: 2020
• Outros Autores: Centeno, Jorge Antonio Silva
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Revista Veras
• Texto Completo: https://ojs.brazilianjournals.com.br/ojs/index.php/BRJD/article/view/12012
• topic: Mobile mapping; point cloud processing; pole mapping

This paper introduces a method for detecting and classifying vertical objects from a mobile terrestrial laser scanner point cloud. The paper concentrates on the classification of the top of the poles, where shields or lamps are installed. First, the variance-covariance matrix of each segmented object is computed. Then the eigenvalues and eigenvector of this matrix are derived. The 3D coordinates of each point are then transformed using the principal components transform in order to compute new features in this new space. In the second step the distribution of the three eigenvalues of the different classes in the eigenvalues space is analysed. Is it deduced that similar objects align in this space, allowing proposing a classification rule based on the distance to the lines. An experiment was performed to verify the approach performance. In the classification of different objects, the global accuracy reached 75%. When the classification was more general, separating just flat from three-dimensional objects the accuracy reached 94%. From the obtained results it can be concluded that the proposed method is feasible and allows separating objects according to its shape