Color Optic Flow: A Computer Vision Approach for Object Detection on UAVs

Development and Initial Testing of the Aero Testbed: A Large-Scale Unmanned Electric Aerobatic Aircraft for Aerodynamics Research

“…18 Work at the University of Illinois using a 6.5-ft wingspan aerobatic unmanned aircraft involved risk mitigation in the form of failure mode analysis of subsystems and components, 19 particularly fuel management, 20 system fault detection, 21 and object and terrain detection using optical flow. 22,23 The aircraft used by the University of Illinois was also used by Boeing to test distributed communication networks. 24 The University of Kansas has performed extensive testing with their large, 10-ft wingspan, aerobatic unmanned aircraft: researchers have evaluated a COTS autopilot, 25, 26 tested a new flight control system, 27, 28 performed aircraft and avionics system identification, [29][30][31] used it as the base aircraft from which a test pilot transitioned to a new unfamiliar airframe, 32 evaluated flight loads using strain gauge measurements 33 and compared moment of inertia estimation methods with experimental measurements.…”

A Conceptual Framework and a Review of Conflict Sensing, Detection, Awareness and Escape Maneuvering Methods for UAVs

“…As the advances of new powerful processing units, cameras can be used as a passive sensor to detect the obstacles around UAV. Many efforts are already being conducted to use camera in CAS systems such as the research found in (Matthies et al, 1998;Oh, 2004;Boon Kiat et al, 2004;Muratet et al, 2005;Mehra et al, 2005;De Wagter & Mulder, 2005;Zhihai et al, 2006;Ortiz & Neogi, 2006;Prazenica et al, 2006;Frew et al, 2006;Subong et al, 2008;Moore et al, 2009;Zufferey et al, 2010). Video cameras are light and inexpensive and thereby fit to the UAV requirements especially the small one.…”

Convolutional neural network-based vision systems for unmanned aerial vehicles

“…• Chapter 4: Proposes an obstacle detection processing pipeline which uses stereoscopic vi sion in order to detect and map in 3D potential obstacles at or above the horizon line • Chapter 5: Provides conclusions in regards to the concepts presented in the thesis and suggests recommendations for future research. 14…”

“…In the case where a stereoscopic pair of cameras is not feasible, such as for MAVs, monocular cameras have also been used both for obstacle detection [14] and also for computing depth from motion [14], [13], [16]. Depth from motion computes depth by tracking image feature points be tween successive image frames and inferring the 3D motion of the features based on their movement across an image.…”