Usually, weed-infested crops require the use of herbicides as a useful manner of controlling the quality and quantity of crop production. Although there are weed-free areas, the most common approach is to broadcast herbicides entirely over crop fields, resulting in a potential reduction of profits. Recently, patch spraying has allowed the application of site-specific weed management by using precise and timely weed maps at very early phenological stage obtained either by ground sampling or proximal or remote sensing. Remote imagery from piloted planes and satellites are not suitable for this purpose given their low spatial and temporal resolutions; however, unmanned aerial vehicles (UAV) represent an excellent alternative. This paper presents a new classification framework for weed monitoring via UAV showing promising results and accurate generalisation in different scenarios.