Hyperspectral detection and determination of troublesome agricultural weeds


Weeds can create major problems for farmers. If not properly managed, they reduce yields, cause infestations while toxic weeds might even end up in the food chain. The goal of this research is to detect and identify troublesome weeds with the use of hyperspectral techniques. Research has proven that multi- and hyperspectral sensors are able to distinguish between certain similar species. Site-specific weed management leads to economic and environmental benefits as it is strongly linked to the quantity of herbicides that are applied. Weed identification can, in addition, aid in finding the proper management and, if necessary, the most appropriate herbicide type. In a first step, hyperspectral signatures of selected weeds will be collected with a spectrometer and investigated using different classification algorithms. On the one hand, weeds and their associated crops will be sampled with respect to site-specific management. On the other hand, co-occurring weeds that are hard to distinguish based on their morphology will be identified and signatures compared. In a next stage, weeds will be detected using a hyperspectral camera. 


Project administration

Researcher: Marlies Lauwers

Project duration: 01/06/2017-31/05/2023