UAV autonomous navigation with thermal infrared camera

Resumo

Image processing approach has been employed as an alternative to the global navigation satellite system
(GNSS) signal for the unmanned aerial vehicle (UAV) autonomous navigation. The UAV image is caught during
the fly. A supervised artificial neural network is configured to be an image edge extractor. Two images are
employed to the UAV positioning: a satellite georeferenced image, and image from the UAV. The image processing
procedure consists to extract edges from the images, and to apply a correlation operator on the two segmented
former images. The maximum correlation coordinate will be to indicate the UAV position. A low cost thermal
infrared camera of type FLIR (Forward Looking Infrared) Duo is coupled to two UAVs (DJI Phantom 3 Standard,
DJI Matrice 600 Pro) for two experiments, during daytime and nighttime. The algorithm is also implemented in
Raspberry Pi and its processing time was evaluated. The results have shown good results for the UAV autonomous
navigation using the low cost thermal infrared camera