Vibration Compensation for Fisheye Lenses in UAV Applications

Low-cost aerial vision systems need to face the challenges of using low quality products to perform aerial photography. Such systems are widely used in remote controlled aircrafts and unmanned aerial vehicles (UAVs) to collect aerial imagery and are used for image acquisition, terrain mapping or remote sensing. A one-pixel shift in a 0.8 mega pixel resolution image captured from a UAV operating at 1000ft will correspond to about 2.5m measurement error on the ground. In our case, a vibrating fisheye lens moving relative to the camera added new uncertainties in the collected images and required compensation. This paper presents a vibration compensation approach using a modified Hough Transform utilizing the circle shaped image provided by a fisheye lens. We define the fisheye circle boundary by using a Canny edge detector. Our vibration compensation was tested using our collected aerial images with enhanced performance in more than 80% of the cases.