Aerial Images from an UAV system: 3D modeling and tree species classification in a park area

The use of aerial imagery, acquired by the so-called UAVs (Unmanned Aerial Vehicles), is becoming increasingly popular for both geometric and descriptive purposes. The employment of such type of images is, for instance, scheduled within the FoGLIE project (Fruition of Goods Landscape in Interactive Environment) that was recently funded by Regione Lombardia to give value at the Cultural Heritage by industrial research and experimental development projects. The project starts from the need to enhance the natural, artistic and cultural heritage, widespread on the Lombardy's territory, and to produce a better usability of it by employing, along predetermined routes, audiovisual movable systems of 3D reconstruction. At the same time, the project comes also from the need to improve monitoring procedures by making use of new media for integrating the fruition phase with the documentation and preservation one. FoGLIE, which began in October 2010 and will be completed in the following two years, involves some Lombardy firms and Politecnico di Milano, for scientific and technical support in the photogrammetric surveys: these are indeed necessary to produce augmented reality scenes and bird's eye views, besides the widely used 3D visualization. By keeping in mind these aims, the UAVs systems surely represent an innovative and flexible technology, able to collect a big amount of very high resolution information, regarding forest landscape, critical environmental situations and not accessible or not properly enhanced areas. The pilot project is assessed on a single area, "Parco Adda Nord", suitable due to the presence of various goods' types inside it. Currently, some tests have already been performed to analyze the UAV images' quality with both photogrammetric and photo-interpretation purposes, to validate the vector-sensor system and to optimize the survey planning for 3D modeling and vegetation classification. A Microdrone MD4-200 was flown, by keeping an altitude of roughly 50 m, over a 200x100 m2 area, chosen because it's partially built-up but also covered by an agricultural field as well as by different tree species and bushes. Thus, multispectral high resolution images were taken by a Pentax OPTIO A40 (a low cost camera), a Sigma DP1 and a Tetracam ADC Lite: the first one was equipped with a CCD sensor with the classic RGB channels whereas the second one had a Foveon specifically modified to acquire the near infra-red band. Instead the last camera had a CMOS sensor (3.2 Mpixel) sensitive to the NIR, the red and the green components thereby yielding directly the so-called "Color Infra-Red" (CIR) images. In fact the IR spectral radiation, besides the visible one, is essential to define the vegetative cover of the territory through classification processes and also to distinguish among different tree species. Lastly, a big number of pre-signalized Ground Control Points has been surveyed through GPS (Trimble 5700 in RTK survey) to allow subsequent accuracy analysis. Indeed, Aerial Triangulation (AT) and DSMs productions were carried out with two photogrammetric commercial software, Leica Photogrammetry Suite (LPS) and Photomodeler, to pick out pros and cons of each package in managing non conventional aerial imagery as well as the differences in the modeling approach. Then, cross-comparisons of the results were performed even within the same software by changing the employed set of GCPs and Check Points: this was helpful for studying how much the choice of them could affect the ATs and the E.O. parameters up to the DSMs and the CPs positions. Further analysis were done on the differences between the E.O. parameters and the corresponding data coming from the on board UAV navigation system. The results of this test phase will be used as guidelines for the following applications on the whole park and for the monitoring of natural and partially urbanized areas through UAVs.