Content area
Full Text
The development, implementation, and validation of a simple algorithm for airplane attitude estimation, based on a video of the horizon obtained from a camera mounted on the plane, is presented. The method provides absolute attitude estimation, which makes it suitable for the calibration and testing of artificial horizon systems. Solutions for some of the difficulties encountered in the determination of roll and pitch angles are discussed and presented. Experimental results from actual flights in a light aircraftare used to compare estimates obtained by the proposed method with data obtained from a commercial attitude reference system.
Introduction: The research and development of electronic flight infor- mation systems (EFIS) have been a trend in the development of avionic systems. One of the main components of an EFIS is known as artificial horizon (AH), which is responsible for informing the pilot of the airplane attitude (roll and pitch). The present work was motivated by the need to calibrate and test in-house-built AH systems, which, in general, are based on the integration and combination of measurements from inertial sensors. An algorithm that estimates the absolute airplane attitude from a video of the horizon obtained by a camera fixed on the plane is presented. Since the proposed algorithm is efficient enough to be executed online, during the flight, it could also be directly used to navigate and control unmanned air vehicles (UAVs), as proposed in [1], In this application, the main advantage of visual systems over iner- tial based systems is that the visual information is not subject to inte- gration errors. On the other hand, the main limitations of systems are that they depend on good weather conditions and good images of the horizon.
The computation of aircraft attitude estimates from video was first pro- posed in [1] and [2], Similar to both works, the methodology proposed in this Letter classifies the pixels of each image into ground and sky, and uses the resultant binary image to compute the attitude. Actually, before the development of the method proposed here, the proposal of Comall et al. [2], which is very simple and efficient, was considered to constitute the benchmark against which other methods should be compared. However, the results from the application of the technique published in [2]...