Extending quad-rotor UAV autonomy with onboard image processing

Abstract

One of the most dynamic technological advances of the last decade is the development of unmanned and autonomous vehicles. For the military, these vehicles represent a safer and more efficient way of fighting wars in aerial, ground, maritime, and underwater domains. Public and private companies have also vigorously researched these vehicles and used them for a wide range of tasks, from search-and-rescue operations to building inspections. Navigating these vehicles typically involves the use of GPS or other external cues to follow a path, detecting for and correcting errors along the way. The purpose of this research is to investigate the feasibility of tracking a ground target using a quadrotor that navigates solely based on relative position to the target. To achieve this goal, the quadrotor, a Quanser Qball-X4, is fitted with a small camera. By processing the camera’s image and utilizing pitch, roll, and altitude data from other onboard sensors, a targeting solution can be derived. To track the target, the tracking vehicle defines error as any deviation from the desired angular offset from that target, continuously correcting that error to maintain its desired offset. By using relative position, the tracking vehicle can continue to follow the target using its onboard camera.