Implementation and evaluation of using terrain tracking technology for UAV image acquisition

Abstract

Small unmanned aerial vehicles have been used for earth observation due to its low cost, high mobility and the ability to hover or access to areas that are dangerous for human. Observations are taken by image sensors through a sampling framework, ground features are represented by pixels in images. Measurements with different scales cause variations of the amount of details in an image. The scale of measurement is determined by several factors, for a study using the same image sensor, the distances between the sensor and the ground are always changing across terrain and they have the most significant effect on scales of measurements. The investigation of the performance of using terrain tracking technology to address the scale problem existing in images captured by UAV platforms is reported here. Four groups of experiments were conducted using a UAV platform with and without terrain tracking technology. 60 photos are used in each experiment to create a 3D model and related products; corresponding evaluations are done by measuring the change of ground sample distance (calibrated images and orthophotos) or altitude accuracy (DEMs). Results show that the utilization of terrain following decreases the influence of terrain on actual Earth surface areas represented by image pixels, and for orthophotos, less deformation of ground objects could be detected. More than twice the number of tie points are generated during photo alignment in the groups which use terrain tracking for image acquisition.