Fixed wing UAV target geolocation estimation from camera images

In this thesis, a target geolocation measurement model for a fixed wing unmanned aerial vehicle which equipped with an image processing system and standard sensors has been created. The measurement model begins to calculate the position of the relevant stationary target when the moment the target of interest is selected in the image. The Extended Kalman filter, which uses the pixel position of the target in the image plane, position, and angular position of the aircraft, is developed for the measurement model. Different from previous studies, the deviation of camera placement angles has been taken into account in the designed filter. In the flight tests, the target position was calculated with an accuracy of 5 meters, from a distance around 400 meters, in a time interval of 15-20 seconds. Also, camera placement angle errors were calculated with a precision of 1 degree. In addition, a loitering maneuver control algorithm has been designed that uses the pixel position of the target on the image plane. This control algorithm holds the target in the camera field of view and desired distance between UAV and target. Unlike the common loitering control algorithms such as waypoint navigation, this loitering maneuver control algorithm is designed independently from any GPS measurement. Hence, it will work regardless of GPS, as long as the angular positions of the aircraft are measured correctly. The control algorithm is designed as modular, so it can be used in any fixed-wing aircraft which has an image processing system and standard sensors. It can be easily implemented as an outer loop for roll attitude controllers. Keywords: Target geolocation, Image based GPS denied loitering, Image based target line of sight estimation