Use of Information From the Digital Elevation Model in Autonomous Air Navigation Based on Image Odometry

Unmanned aerial vehicles (UAVs) have been used in several applications, such as surveillance and remote sensing, among others. This type of aircraft can have its location defined by an inertial navigation system (INS), by a satellite positioning system (SPS), by a vision system (VS), and other systems that employ sonar and laser, for altimetry. It is also possible to combine these systems to achieve greater navigation efficiency. This work proposes a methodology to estimate the position of the UAV, considering techniques based on image odometry, integrating with an estimation of image odometry, and matching with digital elevation models (DEMs) data, to increase the accuracy in estimating the position of a UAV. To test the methodology, two tests with real data were performed (Experiments 1 and 2). The UAV flight was carried out with a Phantom 4 Pro Plus, with a flight height of 80 m, a ground sample distance (GSD) of 2.3 cm, and an approximate altitude of 968 m. The base DEM was generated by a commercial Pix4D, for comparison. Based on the experiment results, it is concluded that it is possible to navigate a UAV using the proposed methodology, since the difference is in the centimeter range, considering that the altitude is in 100-m range.