THE EVALUATION OF VEGETATION CLASSIFICATION BY NARROW-BAND MULTI-SPECTRAL UAV IMAGES

ABSTRACT: Unmanned Aerial Vehicle (UAV) is a high mobility and low cost platform, which can obtain higher spatial and temporal resolution images. In this study, a Miniature Multispectral Camera Array (MiniMCA) is mounted on a fixed-wing UAV to acquire multispectral images for the purpose of various vegetation classifications. MiniMCA is a narrow-band and 12 different lenses composed camera, which can acquire spectrum ranges from blue to near infrared spectral response (450-950 nm). These 12 narrow bands can derive more vegetation indices (VIs) than broad band multi-spectral images, such as NDVI, CAR, GNDVI, OSAVI, and PRI, which has been applied on precision agriculture, environment monitoring, and water stress evaluation. Since the different viewpoints of each camera causes significant misregistration effect, it has first been corrected and aligned to one sensor viewpoint geometry, and secondly generate multi-layer ortho-image with 40 cm spatial resolution through aerial triangulation technic. In the end, the digital number of image is transferred to radiometric reflectance value for computing various VIs. The classification is performed through Object-based Image Analysis (OBIA), and combining the VIs and objects geometry to evaluate the vegetation classification ability