Multi-Sensor 3D Model Reconstruction in Unknown Environments
Safe exploration and remote visualization for dangerous and unknown environments have been of a growing demand for several applications. Exploration of dangerous environments can be life threating if it directly involves a human being. Several approaches have been proposed and developed to allow human operators to explore unknown environments in a safe and efficient manner. 3D reconstruction utilizing vision systems is commonly used for object and environment reconstruction and visualization. Different single or multi-sensor vision systems can be utilized for object and environment 3D reconstruction. Single vision sensor systems are the most prone to failure due to environmental factors, that can cause reconstruction failure or degradation. The nature of the reconstructed scene also contributes to the success of the reconstruction process. On one side, Image-based reconstruction is widely utilized for 3D reconstruction which depends on RGB image footage. Despite the high accessibility of the technology and its relatively less expensive cost, it suffers from different failure scenarios. Low illumination and low textured scenes are among the factors that can lead to reconstruction failure. On the other side, point cloud-based reconstruction is another approach that can be utilized using range sensors such as LiDARs. This method is relativity less accessible and more expensive. However, it offers robust performance in different scenarios. Similar to the image-based approach, point cloud-based methods have their limitations that mostly appear in highly reflective environments. This thesis presents a multiplatform 3D reconstruction method operating from unmanned ground vehicles to increase system robustness and to introduce a safer reconstruction in unknown environments. The proposed system integrates low-cost RGB imaging camera with a 3D LiDAR for enhanced reconstruction. A novel toolchain is developed to reduce the processing time of the reconstruction operation. The robustness of the proposed method was examined over several test scenarios, which involves both indoor and outdoor reconstruction. The proposed methods have increased the reconstruction system robustness and reduced the computational complexity, resulting in increasing the overall system reliability.
URI for this recordhttp://hdl.handle.net/1974/26243
Request an alternative formatIf you require this document in an alternate, accessible format, please contact the Queen's Adaptive Technology Centre
The following license files are associated with this item: