(Georgia Institute of Technology, 2010-04)
Shah, Syed Irtiza Ali; Wu, Allen D.; Johnson, Eric N.
This paper develops a vision-based algorithm to detect a visual marker in real time and
in the presence of excessive colored noise for Unmanned Aerial Vehicles. After using various
image analysis techniques, including color histograms, filtering techniques and color space
analyses, typical pixel-based characteristics of the visual marker were established. It was
found that not only various color space based characteristics were significant, but also
relationships between various channels across different color spaces were of great
consequence. A block based search algorithm was then used to search for those established
characteristics in real-time image data stream from a colored camera. A low cost noise and
interference filter was also devised to handle excessive noise that was encountered during
flight tests. The specific implementation scenario is that of detection of a Blue LED for
GeorgiaTech's participating aircraft into the International Aerial Robotics competition.
The final algorithm that was implemented on GTAR lama aircraft, used both multiple
thresholding and linear confidence level calculations and was successfully used in the
competition in 2009.