Vision Based Landing Sensor System For UAV

Mini Unmanned Aerial Vehicle (UAV), which possesses special features, has broad applications in various fields such as geography mapping, traffic management, military reconnaissance, etc. As an indispensable part of Mini UAV, vision system has been paid great attention by colleges and research organizations in the world. Especially, vision-based landing system plays an important role in UAV autonomous landing process. In this paper, we focus on landing vision system design which can assist UAV land automatically.The design of landing vision system is discussed from both hardware and software aspects. Our system chooses TMS320DM642 DSP as the core processing chip, and is augmented by the capture module, communication module, power supply module, etc. The visual driver is based on TI GIO/FVID model. A series of algorithms including average filter, binarization, Hu invariant moments detection, mean shift following, SUSAN corner detection, etc. are embedded on the hardware platform. The vision system could recognize and track landing-pad, tracking and measure the relative distance based on binocular parallax.Compare to the traditional sensors such as GPS, sonar and gyroscope, our vision landing system is able to offer positional and height information, as a result, it can assist UAV to land on a certain place. Besides, the DSP vision system designed in this paper which gets rid of the restriction of link's precision for analog transmission is more flexible and is advanced for higher Cost-Performance Ratio. The vision system is also easy to incorporate new capabilities in the future research.