No abstract

This paper details the development of a geometrical-based displacement extraction framework capable of automatically extracting critical infrastructure measurements in one sequence. The framework is a novel rail viaduct bearing inspection pipeline implemented on BINOv2: Bearing Inspector for Narrow-space Observation Version 2. BINOv2 is a tethered custom Unmanned Aerial Vehicle (UAV) system utilized to supplant labor-intensive pipelines and enhance inspection accuracy of infrastructure conditions in confined remote locations. The algorithm accepts stereoscopic images taken from a single monocular camera on a bi-directional cascaded linear actuator system in a rack-and-pinion configuration. A point cloud model generated from the image sets then runs through a hierarchical neural network for 3D segmentation to extract targeted regions of interest. Our training pipeline generates and forms the full model's training data set using only a small sample of real point clouds. The point cloud generated is inadequate to form the full bearing geometry profile. Therefore, the proposed framework projects best-fit circles based on the point cloud curvature to form the full bearing geometry profile so that the required displacement measurement is available for extraction. Several experiments were conducted on a mock-up and actual operational site to validate the proposed framework's accuracy, its robustness and comparison with other state-of-the-art alternatives.

In this paper, we propose a ground-based monocular UAV localisation system that detects and localises an LED marker attached to the underside of a UAV. Our system removes the need for extensive infrastructure and calibration unlike existing technologies such as UWB, radio frequency and multi-camera systems often used for localisation in GPS-denied environment.To improve deployablity for real-world applications without the need to collect extensive real dataset, we train a CNN on synthetic binary images as opposed to using real images in existing monocular UAV localisation methods, and factor in the camera's zoom to allow tracking of UAVs flying at further distances. We propose NoisyCutout algorithm for augmenting synthetic binary images to simulate binary images processed from real images and show that it improves localisation accuracy as compared to using existing salt-and-pepper and Cutout augmentation methods. We also leverage uncertainty propagation to modify the CNN's loss function and show that this also improves localisation accuracy. Real-world experiments are conducted to evaluate our methods and we achieve an overall 3D RMSE of approximately 0.41m.

No abstract

No abstract