In-flight horizon line detection for airplanes using image processing

Abstract

The paper presents an image processing based horizon recognition method used for aircraft attitude determination. The results are presented using data in the form of captured images by a wide angle lens video camera mounted on the vertical stabilizer of a two-seater sailplane during flight. The pixel coordinates of the visible horizon line are firstly obtained using robust image recognition techniques, successfully managing problems posed by the varying exposure levels in captured images. Adaptive contrast detecting methods are employed, solving challenges caused by different ground patterns and varying color intensity of the cloud base. A means for optimal projection of the detected horizon line on the image plane is also implemented. Distortion correction caused by the wide angle lens is employed, using the a-priori identified internal camera parameters. A method for alignment of aircraft and camera coordinate frames is also proposed, based on image processing. The pitch and yaw angles of the aircraft are calculated based on a concept tested and validated both in the laboratory and in the field. The presented method can be used for unmanned aerial vehicle (UAV) control when supplemented with additional sensors.