A unified detection-tracking-estimation vision navigation system and algorithms framework for position and pose estimation of small aerial vehicle (UAV) is proposed in conjunction with our recent 3D dynamic localization technique. The major contribution of this work is a novel combination of deep learning for detection of moving object as the small UAV, follow by the vision tracking and estimation of the spatial location in global frame. The system can be multiplied and that the extended systems would provide 3D coordinates of the UAV. The demonstrations in both indoor and outdoor were conducted and that the system was able to detect and locate the small aerial vehicle in 3D space.

This paper describes the development of an automated dimensional extraction algorithm of specific regions, implemented on BINO: Bearing Inspector for Narrow-space Observation. BINO is a tethered Unmanned Aerial Vehicle (UAV) system designed to eliminate the labor-intensive process and to improve the efficiency of inspecting infrastructures in confined remote locations. The UAV platform provides an effective method to access these hard to reach areas, and the algorithm proposed provides a dimensional analysis of the targeted infrastructure. The algorithm takes in several stereo pair images captured from only a single monocular camera. The availability of the multiple stereo images is possible due to a customized linear actuator where the camera sits on a rack-and-pinion configuration. The algorithm adopts a convolution network for image segmentation to extract crucial areas and, at the same time, generate disparity maps to obtain depth information. A pinhole camera model using the depth information and a single reference object of the bearing enables the derivation of the measurements of the segmented areas. Through experiments, the algorithm has an accuracy of an average RMSE of 1.23mm and a max error of 2.23mm. The ease of conducting inspection using BINO and more accurate measurements from the algorithm ultimately leads to an increase in inspection quality, safety and efficiency.