Crop protection represents one of the highest costs for farm budgets. Farmers are under increasing social and economic pressure to dramatically reduce pesticide use. Although most weed and disease infestations occur in patches, the most widely used practice is to spray pesticides uniformly. In this project, a ground-based real-time remote sensing system will be conceived. This will make it possible to detect plant diseases automatically in arable crops at an early stage of disease development, even before the diseases are visibly detectable, during field operations. The methodology will use differences in reflectance and fluorescence properties, and leaf temperature variations between healthy and diseased plants. An intelligent multisensor fusion decision system based on neural networks will be used to decide on the presence of diseases or plant stresses, in order to treat the diseases in a spatially variable way.
The main objective is to develop a ground based, real-time, remote sensing system for detecting diseases in arable crops at an early stage of the development of the disease, before the disease can be visibly detected. The methodology will be based upon differences in reflectance and fluorescence properties, and leaf temperature variations between healthy and diseased plants.
Early detection of diseases can be used to reduce pesticide use through spatially variable treatment. The methodology must be developed so that it can be used under field conditions.
The optical system can be mounted on a field vehicle so that it can be applied in the mapping concept and real-time concept as well. The optical devices with the image processor and the classifier must respond fast enough to enable the vehicle to cross the field at normal speed
Progress to Date
During the first year, preparatory work and calibrations were performed to find the best hardware and software configurations for disease detection. After all the partners had obtained all the equipment needed to begin the actual project work, some preliminary experiments were performed. The feasibility of using all the sensors together to perform sensor fusion was proven. Test protocols for the experiments were constructed and a series of experiments were executed under greenhouse conditions. Reflectance and thermal infrared imaging techniques were developed for simultaneous acquisition in the same canopy. New fluorescence acquisition techniques were described and experimental platforms were constructed.
During the second year, extensive measurement campaigns were performed. Hyperspectral reflectance, fluorescence imaging, and multispectral reflectance imaging techniques were developed for simultaneous acquisition in the same canopy. New fluorescence acquisition techniques were described, experimental platforms were constructed, and the advantage of using sensor fusion was proven. Furthermore, mapping of diseases based on automated optical sensing and intelligent prediction gave a spatially variable recommendation for spraying.
During the third year, a robust multi-sensor platform integrating optical sensing, GPS and a data processing unit was constructed and calibrated. Furthermore, field tests were carried out to optimise the functioning of the multi-sensor disease-detection device.
Due to the multi-functionality of the multi-sensor optical prototype, strategies to enable further exploitation of the technologies developed in the period after the project's conclusion were planned.
The following apparatus and systems were developed:
1) a multi-sensor fusion optical disease-detection system for greenhouses, able to recognise types of stress and disease (for example, septoria from water stress)
2) an experimental field chlorophyll fluorescence apparatus that can be used for field measurements under ambient light conditions
3) a hyperspectral system that can recognise the type of stress and disease in-field on winter wheat (yellow rust from nitrogen stress)
4) a multi-spectral imaging system capable of discovering disease spots in the field and quantifying the infestation
5) a multi-sensor fusion optical disease-detection system that can recognise in-field winter wheat infestation. The severity can be quantified through neural networks
6) GPS-based mapping of disease severity based on a multi-sensor fusion optical system
7) a multi-sensor fusion disease-detection prototype equipped with GPS, a multi-fibre spectrograph, an infrared colour camera and built-in data processing unit.
ARABLE CROPS, CROP PESTS AND DISEASES
Scientist responsible for the project
Professor HERMAN RAMON
Kasteelpark Arenberg 30
Belgium - BE Region NUTS Flanders
Phone: +32 16 321446
Fax: +32 16 328590