Performance Analysis of Camera Rotation Estimation Algorithms in Multi-Sensor Fusion for Unmanned Aircraft Attitude Estimation

In this paper the possibility to use the camera of the onboard sense and avoid system as an aid in unmanned aircraft attitude estimation is examined. At first, the error analysis of four image processing algorithms targeting the reconstruction of camera orientation change is done. Camera spatial resolution, sampling frequency (temporal resolution), pixelization errors and image noises are all considered in the error analysis. After performing the error analysis, the camera rotation estimates from the two best algorithms are integrated into an IMU (inertial measurement unit)-GPS attitude estimator to obtain a Camera-IMU-GPS estimator. The attitude estimation results of the two algorithms are compared to each other considering hardware-in-the-loop simulated ’flight data’ and simulated camera images. After experiencing estimation error decrease with the augmented Camera-IMU-GPS algorithm a final proof of concept was conducted based-on real flight data. Off-line calculated IMU-GPS and Camera-IMU-GPS estimates were compared to the estimates of the onboard AHRS (attitude heading reference system) algorithm of the test aircraft. Estimation of errors relative to true orientation is done through the visualization of horizon line (obtained from the estimated Euler angles) in the camera image. This shows possible improvement in the precision by applying the Camera-IMU-GPS algorithm, but the considered very short (about 17sec) flight data section makes only the proof of concept possible. Final decision between the algorithms should be made considering long flight test results during the future development.