| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 39.78 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Existem inúmeras minas desativadas que apresentam um elevado valor patrimonial e histórico. Contudo a maioria destas minas acaba por não ser documentadas com o devido cuidado, existindo o risco de deterioração do património com o passar do tempo, pelo que se torna necessária a existência de técnicas que efetuem a reconstrução e o mapeamento 3D de modo a preservar estes locais no mundo virtual. O projeto MineHeritage surge em resposta a essa necessidade, com o objetivo de documentar o património geo-mineiro e disponibilizar todos os dados para consultas futuras, utilizando para isso sistemas robóticos capazes de realizar tarefas de mapeamento e reconstrução dos ambientes subterrâneos das minas e da sua área envolvente. A dissertação endereça o desenvolvimento de uma plataforma móvel, composta por um conjunto de sensores e equipamentos necessários para a recolha dos dados (par câmaras stereo, Light Detection And Ranging (LiDAR), sensor inercial), de forma a construir, em pós-processamento, um mapa e respectiva reconstrução 3D das áreas subterrâneas exploradas. A solução desenvolvida ´e capaz de obter dados que permitam calcular a sua localização relativa num cenário GPS-denied (ambientes subterrâneos onde não é possível obter localização GPS) recorrendo ao uso de odometria visual-inercial. O desenvolvimento do sistema passou pelo levantamento de todos os requisitos de hardware e software necessários para o funcionamento pretendido, pelo desenho de todos os componentes e peças, e pela implementação do sistema. Numa primeira fase, foram realizados vários testes em ambiente controlado. Após a validação de todos as funcionalidades do GeoTec, procedeu-se `a realização dos primeiros testes em ambiente real na mina das Aveleiras situada no recinto do mosteiro de Tibães, que se encontra localizado em Braga. Com os dados obtidos nestes testes, foram realizados vários testes com dois programas que realizam reconstrução 3D: o programa Meshroom, que usa a framework de fotogrametria da Alice Vision, e em Robot Operating System (ROS), o package RTAB-Map, sendo que os resultados obtidos mostraram-se promissores. Foi também comparada a eficácia de cada um dos algoritmos na estimação da posição através da comparação da trajetória de um ground truth, obtida usando os dados do recetor GNSS RTK fundidos com o sensor inercial e as trajetórias obtidas usando ambos os programas.
There are countless deactivated mines that have a high heritage and historical value. However, most of these mines end up not being documented with due care, with the risk of loss of that heritage over time, making it necessary to have techniques that carry out reconstruction and mapping in order to preserve these sites in the virtual world. The MineHeritage project arises from this need, with the aim of documenting the geo-mining heritage and making all data available for future consultations, using robotic systems capable of mapping and reconstructing the underground environments of the mines and their surrounding areas. As such, it is essential to develop a mobile platform, consisting of a set of sensors and equipment necessary for data collection (stereo pair, LiDAR, inertial sensor), in order to build, in post-processing, a map and its respective 3D reconstruction of the underground areas to explore. It should be able to obtain its relative location in a GPS-denied scenario (underground environments where it’s not possible to obtain GPS signal) using visual-inertial odometry. The development of the system involved the survey of all the hardware and software requirements necessary for the intended operation, the design and implementation of the same. In a first phase, several tests were carried out in a controlled environment. After the validation of all GeoTec functionalities, the first missions were carried out at the Aveleiras mine located in the Tibães monastery, which is located in Braga. With the data obtained in this field mission, several tests were carried out with two programs that perform 3D reconstruction: Meshroom program, that uses the Alice Vision photogrammetry framework, and in ROS the RTAB-Map package. The results obtained revealed themselves promising. The effectiveness of each of the algorithms in estimating the position was also compared by comparing the trajectory of a ground truth, obtained using the data from the GNSS RTK receiver merged with the inertial sensor and the trajectory estimated by each program.
There are countless deactivated mines that have a high heritage and historical value. However, most of these mines end up not being documented with due care, with the risk of loss of that heritage over time, making it necessary to have techniques that carry out reconstruction and mapping in order to preserve these sites in the virtual world. The MineHeritage project arises from this need, with the aim of documenting the geo-mining heritage and making all data available for future consultations, using robotic systems capable of mapping and reconstructing the underground environments of the mines and their surrounding areas. As such, it is essential to develop a mobile platform, consisting of a set of sensors and equipment necessary for data collection (stereo pair, LiDAR, inertial sensor), in order to build, in post-processing, a map and its respective 3D reconstruction of the underground areas to explore. It should be able to obtain its relative location in a GPS-denied scenario (underground environments where it’s not possible to obtain GPS signal) using visual-inertial odometry. The development of the system involved the survey of all the hardware and software requirements necessary for the intended operation, the design and implementation of the same. In a first phase, several tests were carried out in a controlled environment. After the validation of all GeoTec functionalities, the first missions were carried out at the Aveleiras mine located in the Tibães monastery, which is located in Braga. With the data obtained in this field mission, several tests were carried out with two programs that perform 3D reconstruction: Meshroom program, that uses the Alice Vision photogrammetry framework, and in ROS the RTAB-Map package. The results obtained revealed themselves promising. The effectiveness of each of the algorithms in estimating the position was also compared by comparing the trajectory of a ground truth, obtained using the data from the GNSS RTK receiver merged with the inertial sensor and the trajectory estimated by each program.
Description
Keywords
Reconstrução 3D Cenários GPS-denied Visão stereo Odometria visual-inercial Extração de features Meshroom RTAB-Map 3D reconstruction GPS-denied scenes Stereo vision Visual inertial odometry Feature extraction