UAVs for spatial data acquisition. Sensors evaluation, flight design and planning, multi-temporal solutions

During past years, UAVs (Unmanned Aerial Vehicles) have become very popular and, for this reason, my research activities were concentrated on the use of these systems as geomatic tools for measurements taken in monitoring and change-detection scenarios. The interest in these aerial platforms is due to their capability of performing aerial surveys in an easy and flexible way using low-cost sensors. The improvements in hardware and software frameworks allow for their use in different environmental fields and by non-expert users. In fact, the most significant difference between common aerial platforms and UAVs is that the latter have no pilot on-board. This issue allows for a reduction in the cost of the entire platform, but it is even more advantageous for safety reasons. UAVs can be big, like common manned platforms, or as small as insects. Obviously, also their uses can be very different: bigger systems are mainly used for military applications, while smaller systems are preferable for surveillance or games. In the civil field, the most investigated categories are mini and micro UAVs (payload <30 Kg) both fixed wings and multi-rotor. The advantages are their easy use and also the possibility of easily transporting the systems to perform surveys in different areas. On the other hand, their small size requires specifically designed sensors that have to be both reliable and lightweight enough to respect the limits of the payload. For this reason, three fundamental topics related to the use of UAVs for monitoring application are investigated along this thesis: - evaluation of navigation sensors; - photogrammetric flight design and planning; - proposal of an automatic procedure for the usage of multi-temporal data. All the addressed issues used applications based on different geomatics techniques related to photogrammetry, positioning, navigation and topography. The first part of the thesis was devoted to the analyses of on-board and external (installable on-board) sensors to evaluate their accuracy in relation to well-known topographic measurements. The navigation platform of a UAV was analysed to investigate the reliability of the internal GNSS (Global Navigation Satellite System) receiver and the IMU (Inertial Measurement Unit) platform. Then, an external GNSS receiver and a IMU were housed on-board and analysed to assess if it is possible to improve positioning and attitude information of the UAV through external sensors. In particular, strategies for their installation on-board, for the flight and for the data acquisition and processing were evaluated and tested to obtain a reliable solution. Performed tests showed that internal navigation sensors have metrical accuracy and, for this reason, another fundamental issue in the use of UAVs for data acquisition is the flight plan. UAVs can be programmed to fly along predefined trajectories, but in most cases, real flights are different to the planned ones due to the untrustworthiness of the navigation platform. For this reason, the second part of the thesis was related to the investigation of the flight planning for photogrammetric purposes. Different features of the flight (such as overlaps, cross and border stripes) were evaluated by analysing the final block orientation in relation to different parameter configurations. This was undertaken following a simulation approach which implemented the image orientation through collinearity equations in a Matlab code. The results were then evaluated through different real case studies (two of which were reported in the thesis) to assess if outcomes corresponded to reality. Finally, if we are able to calibrate different flight features to the expected final result, it is easily possible to repeat the same flight plan in different times. This is the case of monitoring scenarios where the availability of time-series information can be a useful instrument of investigation. In this context, the open issue was to relate (co-register) different 3D models derived from this time information. Currently this problem is solved with the introduction of Ground Control Points (GCPs) whose use is time consuming, costly and sometimes not possible at all. For this reason, the last part of my thesis was related to the development of an automatic technique to co-register multi-temporal high-resolution image blocks from UAVs without the introduction of GCPs. The proposed strategy is instead based on the use of a reference epoch (considered fixed) and the registration of the others according to it. The methodology (developed in collaboration with the University of Twente, The Netherlands) was finally evaluated using two real cases, one for testing and the other for validation.