Accuracy and crop line misalignment over high resolution ortho-mosaics from Unmanned Aerial Vehicles

Abstract

The Unmanned Aerial Vehicles (UAVs) are able to take multiple overlapped images of the study site at a very high spatial and temporal resolution. These image series have to be ortho-mosaiced to cover the whole target area. The UAV ortho-mosaics are turning into an important tool for the development of precision agriculture strategies, in particular for accurate prescription maps for site specific weed management (SSWM) at early grow stages, These stages require to discriminate small plants (crop and weeds) that can not be detected using other remote platforms with coarse spatial resolution. Recent studies on crop-weed discrimination at early stages are focused on crop line detection as the first step for a further discrimination of crop and weeds emerged between the crop rows due to plants which are not located in the crop lin e can be assumed as weeds. However, due to the instability of UAV and image distortion, some of the single images which compose the ortho-mosaic do not fit perfectly which would cause no correspondence between crop row alignment of two overlapped images. This study investigated the geometric accuracy differences among ortho-mosaics created from a UAV image series taken at three different flight altitudes (30, 60 and 100 m). The effect of crop line misalignment of the final mosaics was also evaluated according to these flight altitudes.