ACQUISITION AND AUTOMATED RECTIFICATION OF HIGH-RESOLUTION RGB AND NEAR-IR AERIAL PHOTOGRAPHS TO ESTIMATE PLANT BIOMASS AND SURFACE TOPOGRAPHY IN ARID AGRO-ECOSYSTEMS

Increasing image resolution and shrinking camera size facilitates easy mounting of digital cameras on Unmanned Aerial Vehicles (UAVs) to collect large amounts of high-resolution aerial photos for soil surface and vegetation monitoring. Major challenges remain geo-referencing of these images, reliable stitching (mosaicking), elimination of geometric image distortions and compensation of limited image quality and high cost of the equipment. In this study, we report upon the design and field-testing of a custom-made, cost-effective mini-UAV allowing the acquisition of RGB and near-IR images covering areas of 1-2 km(2) in each flight and the development of a software tool to automatically combine the geo-referenced images into a seamless image mosaic. Object-orientated image classification was used to estimate plant biomass. The images allowed to determine the distribution and biomass of selected plant species and other landscape features such as field borders and settlement patterns as well as to construct a simple 3D model of the topography of the surveyed area. The setup facilitates the cost-effective acquisition, mosaicking and classification of hundreds of RGB and near-IR images with a spatial resolution of 5-10 cm.