Classification and modelling of urban environments from point clouds for physical accessibility diagnosis and pedestrian pathfinding

• Autores: Jesús Balado Frías
• Directores de la Tesis: Lucía Díaz Vilariño (dir. tes.), Pedro Arias Sánchez (dir. tes.)
• Lectura: En la Universidade de Vigo ( España ) en 2019
• Idioma: inglés
• Tribunal Calificador de la Tesis: Antonio Fernández Álvarez (presid.), María Flor Álvarez Taboada (secret.), José Alberto Gonçalves (voc.)
• Programa de doctorado: Programa de Doctorado en Geotecnologías Aplicadas a la Construcción, Energía e Industria por la Universidad de Salamanca y la Universidad de Vigo
• Materias:
  • Ciencias tecnológicas
    • Tecnología de la construcción
      • Ingeniería civil
• Enlaces
  • Tesis en acceso abierto en: Investigo
• Resumen

  • Mobility in urban environments is still a challenge for society today, despite the great advances that have been made in recent years to improve physical accessibility. Complexity of urban ground elements and specific needs of the different motor skills, produce some people cannot reach their destinations following a direct route easily. People with reduced mobility are especially vulnerable to this problem. Their limitations when travelling make them a group at risk of social exclusion.

    One of the main deficiencies in urban pedestrian mobility can be seen in the absence of an effective tool capable of generating safe pedestrian routes according to motor skills. The information available on the elements of urban ground, and their state, is non-existent or outdated. LiDAR technology allows the acquisition of urban areas in a fast and precise way, storing their geometry in point clouds that can be processed later. However, point clouds are a source of disordered data, with redundant information, occlusions and without uniform point density that makes them difficult to manage and visualize.

    In this thesis, automatic methodologies for point cloud processing to interpret as-built urban environment are presented, focusing specifically on ground elements conforming the navigable space. The main objective is to generate urban models to calculate realistic pedestrian routes for pedestrians according to motor skills. Geometric information contained in point clouds makes them an optimal source of information for this work, despite the limitations described in the previous paragraph. The point clouds have been classified, firstly at the lowest possible level of detail for urban areas detection by artificial intelligence techniques, and secondly with high level of detail to classify ground elements with geometric and topological information. The modelling has considered dynamic objects in the scene, corrected occlusions and distributed nodes on the ground surface, according to international accessibility standards and as-built environment.

    All the works of this thesis have been tested in carefully selected real case studies, obtaining results that contribute to the solution of the mentioned problem. The generated models have allowed the direct application of a pathfinding algorithm. The generated routes are safe and coherent with the classified ground elements and motor skills, they consider static but not dynamic objects, and they assure a minimum free unobstructed space for people to move comfortably.

    This thesis is structured as a compendium of seven scientific publications. Four are articles published in journal, indexed on the Journal Citation Report (JCR), and three are double-blind peer reviewed conference proceedings.