The Precision Crop Farming group is generating a lot of results lately (March-August 2018):
First of all, Irene Borra-Serrano and her co-authors published a paper in Euphytica in August 2018. The link can be found here. In short, Perrenial ryegrass (Lolium perenne) is a crop often used as forage. Important in the breeding programs of this ryegrass is persistency, unfortunately the human visual scoring of this trait is biased. In this study, the correlation between the scoring done by different breeders was only 0.243 (max. 1 = perfect correlation). The Precision Crop Farming group however developed a methodology to score the ryegrass breeding materials via a drone (Unmanned Aerial Vehicle (UAV)) and image analysis! The correlation between the automatically determined indices and the breeders decisions varied between breeders. In contrast, agreement between decisions based on different UAV flights was very high (91.6%). In summary, the results demonstrate that persistency selection in L. perenne breeding programs from UAV imagery is likely to be more objective in comparison to the currently-used visual scoring method.
Read the full paper or contact the authors for more info!
Secondly, a poster was presented at the Plant Biology Conference in Copenhagen, June 2018. The poster is titled Field based phenotyping of forage crops using close remote sensing tools.
Finally, another poster was presented at the COST FA1306 Meeting in Leuven, March 2018. This poster is titled Phenotyping grass canopy height evolution based on SfM and DEM estimation using a drone.
ISense was present at AgEng 2018!
Detection of Alternaria solani with proximal sensing
ISense attended the conference for agricultural engineering, AgEng 2018 in Wageningen, the Netherlands. We showed our latest findings concerning proximal sensing of A. solani with a poster presentation and a 2 min pitch. Image based disease detection is an important step toward localized application of crop protection products in precision agriculture and receives more attention lately in following up crop health. With this poster we show our first results of the hyperspectral changes caused by A. solani and share our findings of the statistical analysis. The poster presentation can be found here.
Alternaria solani detection model - Ruben Van De Vijver - update May 2018
The first field experiment in which healthy potato plant were inoculated with notorious diseases such as Alternaria solani (see below) allowed us to capture high quality hyperspectral images thanks to a fully automated hypercart constructed at ILVO (see below and here). Based on these images, we could investigate the potential of hyperspectral imaging for the detection of the infamous potato disease Alternaria solani which causes lesions on the leaves. In hyperspectral imaging, both spatial and spectral information (e.g. the ‘color’ info) are captured, which can be linked to disease development much more accurately. Figure 1 shows some of the lesions and buffered regions outside the lesion. On the right graph in Figure 1 the transition from healthy (green) to infected (red) can be observed. Three interesting regions can clearly be noted from this graph: the near-infrared region (NIR, 750nm+), the red region (around 680nm) and the green region (around 550nm) as can be seen in Figure 2. It starts being interesting when these regions are linked to plant behavior and physiology. Changes in the green region is probably caused by anthocyanin, a stress related pigment, which is responsible for yellowing of the leaves. In the red region, changes caused by chlorophyll destruction can be seen while the variations in the NIR region are initiated by the destruction of cell structure, e.g. the collapse of cell walls upon intrusion by the pathogen. The differences in strength of the response can also be seen on the right of Figure 2, where the disease gets most clearly visible in the NIR region and later on also develops in the visible regions (green and red). This clearly accentuates the potential for multi- or hyperspectral imaging for disease detection as it might be a game changer in crop health monitoring.
Figure 1: left: infected potato leaf with some A. solani spots highlighted and buffered, right: comparison between the spectral signal of the healthy outer buffer of the lesions and the heavily infected inner region of the A. solani spot
Figure 2: left: spectral information captured in hyperspectral images of potato plants infected with A. solani, right: images at the three most important regions for A. solani detection
Start greenhouse experiment - Marlies Lauwers - update May 2018
In a first experiment, hyperspectral signatures will be taken of yellow nutsedge (Cyperus esculentus var. microstachyus L.), bayonet grass (Bolboschoenus maritimus (L.) Palla) and hairy sedge (Carex hirta L.) in order to allow weed species classification based on spectral properties. C. esculentus is an invasive troublesome neophyte in many arable crops in Belgium. It co-occurs with some other members of the Poales in particular C. hirta and B. maritimus from which it is hard to distinguish in young vegetative stage. In another experiment hyperspectral signatures of black nightshade (Solanum nigrum ssp. nigrum L.), jimson weed (Datura stramonium L.) and volunteer potatoes (Solanum tuberosum L.) - some of them generate toxic berries - will be compared against those of spinach (Spinacia oleracea L.) and beans (Phaseolus vulgaris L. ‘fresano’); crops in which forementioned weeds pose a high risk of food contamination. Selected weeds are toxic and pose a great problem in some specialty crops as parts of these plants can end up in the food chain.
Plants of both experiments were placed in two separate greenhouses. For the species from the first experiment, vegetative propagules were used . In the second experiment, seeds were sown in plugs, except for the potato seed tubers. For most plants, 15L-pots will be used in order to obtain a stress-free environment. This setup was introduced to make a uniform environment so that variations in hyperspectral signature will be due to plant characteristics. The measurements will be made with the FLAME-VIS-NIR-ES (350-1000 nm; Ocean Optics B.V.) and spectroclip (Ocean Optics B.V.).
Figure 1: Greenhouse with yellow nutsedge, bayonet grass and hairy sedge
Autonomous hypercart for crop sensing - Ruben Van De Vijver - update December 2017
To monitor crop status like disease development in potato crops, a motor driven controlled and steered hypercart equipped with different sensors was built by ILVO. This hypercart was constructed from an aluminium frame with a width of 2.25 m, a height of 2.3 m and a length of 3 m. Different sensors can be mounted on a moving beam in order to scan crops in a plot of 0.85 m by 3 m. The hypercart can be equipped with up to five sensors to gather RGB, multispectral, hyperspectral and height information about the canopy i.e. (1) a RGB camera (D90, Nikon, Japan), (2) a multispectral camera (Sequoia, Parrot, ), (3) a hyperspectral snapshot mosaic camera (made by 3D-one and based on an IMEC CMOS-chip, 41 bands from 470-975 nm), (4) a hyperspectral linescan sensor (Imspector V9, 430-900 nm) and (5) a LiDAR sensor (SICK, LMS111). During field measurements, the hypercart was covered with a black cloth to eliminate external light disturbance. A combination of 18 halogen spot lights (50W, OSRAM Decostar) mounted on the moving beam produced artificial light conditions in order to generate uniform and stable lighting conditions. The hypercart can operate completely autonomous thanks to the use of two lithium-ion accu packs (each 5kWh) and a highly efficient power management system. Propulsion and steering is realized with two brushless DC motors (Motenergy, ME0201013501, 2.4 kW) steered by a Curtis controller and a Controllino mini PLC module. Communication between user, sensors, the moving beam and the steering and propulsion system is done by an integrated shuttle pc (Shuttle, DH170). Thanks to this fully equipped and motorized hypercart the early detection and monitoring of potato diseases and other crop characteristics will take a rolling start!
In addition, thesis student Jordi Demarest (VIVES en ILVO) received a "Best thesis in electro-mechanics" award at VIVES for his hard work on the hypercart!! Well done Jordi!! More info can be found here.
Schematic of a hypercart
ECPA conference - Irene Borra-Serrano - update July 2017
Participation in the "11th European Conference on Precision Agriculture – ECPA 2017" (16-20/07/2017 in Edinburgh, UK).
with the poster entitled ‘Non-destructive monitoring of grassland canopy height using a UAV’. The poster can be found here.
During the conference different aspect of Precision Agriculture were presented and provided us with an overview of the current state of the research done in this field. The sessions covered topics such as Satellite applications, Precision Horticulture, Crop disease, UAVs applications, etc.
The congress was a great opportunity to know what other groups are working on and learn from them techniques or procedures to apply. Moreover new contacts for future collaborations were made.
Drones & Research - Irene Borra-Serrano - update May 2017
Participation in the "Drones and research at the university: state of play in Belgium" day the 31st May 2017 in Louvain-la-Neuve at the UCL (Université Catholique de Louvain). An oral presentation entitled; “Grasslands monitoring” allowed us to present the ISENSE project and in more detail the work carried out in the PhD.
This activity was equally aimed at researchers working in the development of drones and at researchers that use drones as a working tool. The aim was to identify research groups in the field of drones in Belgian universities and research institutes to encourage collaboration and to share information. It was an interesting day with many different topics.
Start field experiment - Ruben Van De Vijver - update April 2017
After weeks of preparation, finally the field experiment has started. Several potato diseases were inoculated and planted into the soil in the course of several months during which potatoes thrive. The field experiment included six different diseases caused by different types of pathogens such as viruses, bacteria, nematodes, fungi and oomycetes (pseudo-fungi). To encompass a broad range of physiological symptoms, both soil borne and airborne pathogens were chosen as soil borne pathogens primarily attack the rooting system, while airborne pathogens mainly infest leaf tissue.
The soil borne pathogens like Verticillium dahliae (Figure 1) and Globodera rostochiensis were blended into the soil to create a homogenous mixture of infected soil.
Figure 1: Verticillium dahliae inoculum in a concentrated (left) and a diluted suspension (middle)
For this blending process a concrete blender was used (Figure 2) as an efficient method to reach a homogeneous mixture. Other soil borne diseases such as Pectobacterium carotovorum were inoculated onto the tuber two days before the planting. Airborne diseases like Alternaria solani were inoculated during the growing season as the infection period was significantly shorter compared to that of soil borne diseases.
Figure 2: practical setup for the inoculum preparation
After blending the soil, the potatoes were planted by hand in previously drilled holes (Figure 3). These holes allowed the precise application of infected soil around the potatoes tuber, which is necessary to control the concentration of pathogens.
Figure 3: Drilled holes allow the precise application of infected soil around the potato tuber
After the planting the measurement campaign begins where we collect hyperspectral data by different types of sensors to create an extensive dataset for disease detection.
COST meeting - Irene Borra-Serrano - update March 2017
Oral presentation at the 3rd general COST Meeting in Oeiras (Portugal) entitled: “Evaluation of persistency of forage grasses in a breeding context using UAV imagery”. This meeting summarized the current state of phenotyping in plants and facilitated the expansion of our network.
Earth Observation Day - Irene Borra-Serrano - update Dec 2016
Attendance to the Belgian Earth Observation Day 2016 on the 8th December 2016 in Saintes. Interesting day with many different topics that allowed us to know what other research groups are working on in Remote Sensing and establish contact for future collaborations.