Search

International audience; Special Issue InformationDear Colleagues,Remote and proximal sensing are the two most common techniques concerning the acquisition of information about an object or any phenomenon without physical contact with the object. Remote sensing is widely tied to the use of satellite, airborne or UAV platforms using multi- or hyperspectral imagery. In terms of proximal sensing, the sensor is close to the object and is installed on platforms ranging from handheld, fixed installations, or robotics and tractor-embedded sensors. The types of sensors range from simple RGB or grey-level-cameras to multispectral and hyperspectral high resoluted imaging systems or even thermographic camera.Associated with plant growth conditions and phenotyping techniques, remote and proximal sensing are able to provide information on nutrient deficiency, biotic stress such as pests and diseases as well as abiotic stresses, allowing Precision Agriculture and Viticulture practices.We invite thus papers on both fundamental and applied research relating on Remote and Proximal Sensing for Precision Agriculture and Viticulture, combining spectral, spatial and temporal information based on multi- and hyperspectral imagery with the capabilities of management-oriented crop simulation models. We also invite papers dedicated to new sensors able to be used in Agriculture; aiming at a better management of the crops, and methods for better crop management and more respectful of the environment.Dr. Frédéric CointaultGuest Editor

prod 2019-108 équipe BAP équipe EA GEAPSI AGROSUP

International audience; This paper presents the use of a mobile manipulator, a robotic arm translating on a guide, to achieve autonomous data acquisition on crops in a field. The redundancy imposed by this platform is managed using a collection of concurrent control laws, weighted in real time to increase the end-effector workspace, and optimize the point of view for data acquisition. This opens the way to full mobile manipulation using an autonomous robot holding the system.

International audience

National audience; Despite an increasing interest for the use of biostimulants (BS) in agriculture, methods allowing a precise description of their effects on plants remain rather limited. In the IRIS+ FUI project, two major and highly different worldwide crops, wheat (annual, monocotyledon) and grapevine (perennial, broadleaf) were chosen to deepen our knowledge of such compounds and explore their potential additional interest. One objective was to develop a tool adapted to the screening and study of the impact of a series of BS on the development and physiology of these crops in controlled conditions. We managed to develop such a tool adapted to grapevine herbaceous cuttings. It allows a quite rapid and non-destructive phenotyping of BS effects on the aerial and root parts of plantlets over time. At the end of the experiment, plant material can be sampled for biomass measurements (fresh and dry weight) and analysis of gene expression or metabolites. This tool was validated with some commercial BS prior being used to study new putative ones. Using this tool, we have identified a new BS provided by Arysta-Goëmar. The positive effects of this BS was dose dependent. We have observed that it transiently accelerates the shoot development and increases the number of roots, and also shoot and root biomass. Moreover, analysis of the root and shoot metabolomes highlighted clear fingerprints associated to BS effects. This tool will be useful to screen new BS potentially active in grapevine. Moreover, it will be useful to study and compare the effects of diverse BS on grapevine development, metabolism and physiology ; and to decipher their mode of action. In this context, work is still in progress to further develop this tool to make it suitable to the highlighting of diverse effects related to BS and other crops, such as wheat.

EAGEAPSIAGROSUPINRA; Le phénotypage d’une plante est classiquement basé sur la caractérisation de traits aériens tels que la surface foliaire, la biomasse ou in fine le rendement. Le système racinaire qui joue pourtant un rôle clé dans le fonctionnement de la plante, n’est que peu étudié de manière non-destructive et dynamique du fait de difficultés techniques d’accessibilité aux racines. Pour ce faire l’outil RhizoTube a été développé. Il consiste en un rhizotron cylindrique qui permet la visualisation 2D du système racinaire d’une ou plusieurs plantes de manière simultanée. Afin de visualiser les racines, une coque protégeant de la lumière s’ouvre permettant une acquisition d’image à moyen ou haut débit au sein de la chambre de photographie appelée RhizoCab. La RhizoCab a été conçue pour prendre des images de l’intégralité du système racinaire de chaque plante à une définition variant de 42 μm à 7 μm. Parmi les différents systèmes d’acquisition d’images existants, la solution retenue est celle de l’imagerie visible (400-700 nm) à l’aide d’une caméra monochrome linéaire de 12000 pixels et d’une rampe de LEDs permettant l’éclairage dans les trois longueurs d’onde primaires. L’image acquise nous permet alors d’accéder à des données diverses telles que la surface projetée, la profondeur de prospection ou des traits architecturaux.

EAGEAPSIAgrosupCT DOCT; The development of the concepts of precision agriculture and viticulture since the last three decades has shown the need to use first 2D image acquisition techniques and dedicated image processing. More and more needs concern now 3D images and information. The main ideas of this chapter is thus to present some innovations of the 3D tools and methods in the agronomic domain. This chapter will particularly focus on two main subjects such as the 3D characterization of crop using Shape from Focus or Structure from Motion techniques and the 3D use for root phenotyping using rhizotron system. Results presented show that 3D information allows to better characterize crucial crop morphometric parameters using proxy-detection or phenotyping methods. résumé du livre : Sensor technologies play a large part in modern life as they are present in security systems, digital cameras, smartphones, and motion sensors. While these devices are always evolving, research is being done to further develop this technology to help detect and analyze threats, perform in-depth inspections, and perform tracking services. Developing and Applying Optoelectronics in Machine Vision evaluates emergent research and theoretical concepts in scanning devices and 3D reconstruction technologies being used to measure their environment. Examining the development of the utilization of machine vision practices and research, optoelectronic devices, and sensor technologies, this book is ideally suited for academics, researchers, students, engineers, and technology developers.

AgrosupINRAEAGEAPSI; High resolution imagery for phenotypic root traits detection. CIRG-AgEng 2016, International Conférence on Agricultural Engineering, Automation, Environment and Food Safety

AgrosupBAPGEAPSI; Micro and macro spray characterisation based on high speed camera techniques Presenting Author David Nuyttens - ILVO, Technology and Food Science Unit, Belgium Description Macro (spray angle, liquid sheet length, spray distribution) and micro spray characteristics (droplet size and velocity) influence the efficiency of the spray application. Consequently, more information on these spray characteristics can lead to a more efficient pesticide usage. This study focuses on measuring spray droplet characteristics using a high speed imaging technique developed by Vulgarakis et al. (2015, Trans ASABE, 58: 27-37). In the first part, the macro spray characteristics from different types of hydraulic spray nozzles were measured using a developed backlighted image acquisition system and image processing technique. Tests included five different nozzles (Albuz ATR orange and red, TeeJet XR 110 01, XR 110 04 and Al 110 04) (Vulgarakis et al., 2014, Asp. App. Biol. 122: 353-361). In the second part, micro spray characteristics were determined for the same nozzle pressure combinations. An in-focus droplet criterion was established to decide whether a droplet is in focus and can be measured in an accurate way (Vulgarakis et al., 2016, Sensors, 2: 218). A calibration system was devised using different sized droplets generated with a piezoelectric droplet generator (Vulgarakis et al., 2015, Crop Prot, 69: 18-27) at different distances from the focal plane. This enabled measurement of the grey level gradient and the in-focus parameter for every droplet size at various positions. From here, a critical in-focus parameter was established for every droplet size and an in-focus droplet criterion was deduced to decide whether a droplet is in focus or not depending on its diameter and in-focus parameter. The in-focus droplet criterion was applied to spray images and droplet characteristics were calculated. The effects of the nozzle type, nozzle size and measuring position on spray droplet size characteristics were studied. All imaging results were compared and related well with existing measuring techniques like the PDPA laser. Contributors Sofija Vulgarakis Minov (ILVO, Technology and Food Science Unit, Agricultural Engineering, 9820 Merelbeke, Belgium), Frédéric Cointault (AgroSup Dijon, Dijon cedex, France), Jürgen Vangeyte (ILVO, Technology and Food Science Unit, Agricultural Engineering, 9820 Merelbeke, Belgium), Jan G Pieters (Department of Biosystems Engineering, Faculty of Bioscience Engineering, Ghent University, Belgium), David Nuyttens (ILVO, Technology and Food Science Unit, Agricultural Engineering, 9820 Merelbeke, Belgium)

AgrosupEAGEAPSICT1; Presenting Author Jürgen Vangeyte - ILVO, Agricultural Engineering, Belgium Description Precision fertilization requires accurate techniques for determining the spread pattern of centrifugal fertilizer spreaders. Traditionally, the spread pattern was determined by measuring the fertilizer distribution on the ground. Because such measurements are time consuming and complex, a high speed stereovision setup and dedicated image processing algorithms were developed for position and motion estimation of fertilizer grains leaving a centrifugal spreader. The measurements are combined with a ballistic flight model to predict the landing points of the individual grains and the actual spread pattern. This presentation will show the different development steps taken and the problems encountered the last 10 years to come to this setup. In a first approach, a two-dimensional imaging technique was used with a small field of view (0.33 m on 0.25 m) only measuring horizontal outlet angle and speed. In a second approach, this 2D approach was further refined using a larger field of view (1 m on 1 m) and using motion estimation techniques inspired from Particle Image Velocimetry (PIV). Finally, a small field of view 3D stereovision system using a high irradiance LED system was developed, able to measure 3D motion and position, and which can therefore be used for inclined spreaders or spreaders with conical disks. Together with the improvement of the measuring system, a 3D ballistic flight model was developed not only including gravitational and drag force, but also Magnus force and drag torque. Simulations using this combined experimental/modelling approach were successfully validated against experimental data. Contributors Simon Cool (ILVO, Technology and Food Science Unit, Agricultural Engineering, 9820 Merelbeke, Belgium), Jan Pieters (Department of Biosystems Engineering, Faculty of Bioscience Engineering, Ghent University, Belgium), Koen Mertens (ILVO, Technology and Food Science Unit, Agricultural Engineering, 9820 Merelbeke, Belgium), David Nuyttens (ILVO, Technology and Food Science Unit, Agricultural Engineering, 9820 Merelbeke, Belgium), Bilal Hijazi (ILVO, Technology and Food Science Unit, Agricultural Engineering, 9820 Merelbeke, Belgium), Julien Dubois (AgroSup Dijon, France), Frederic Cointault (AgroSup Dijon, France), Jürgen Vangeyte (ILVO, Technology and Food Science Unit, Agricultural Engineering, 9820 Merelbeke, Belgium)

Presented at Conference International Advances in Pesticide Application (IAPA), Barcelone, ESP (2016-01-13 - 2016-01-15).; International audience; Spray droplet characteristics are important features of an agricultural spray. The objective of this study is to measure the droplet size for different types of hydraulic spray nozzles using a developed backlighted image acquisition system and image processing technique. An in-focus droplet criterion was established to decide whether a droplet is in focus and can be measured in an accurate way. Tests included five different nozzles (Albuz ATR orange and red, TeeJet XR 110 01, XR 110 04 and Al 110 04).

Biosystems Engineering (BE) term is a reference one in the USA and the UE, with sometimes very different definitions, used since more than a decade. Its contours and contents, for L and M degrees, have been studied in a European program for standardization inside the ERABEE network, allowing to designate this domain as a field of engineering which integrates engineering science and design with applied biological, environmental and agricultural sciences. Therefore, Biosystems Engineering is 'the branch of engineering that applies engineering sciences to solve problems involving biological systems ». This definition appeals to many skills and it is useful to mention that in France, the L1 and L2 levels for BE is essentially supported by the agricultural colleges and the L3 level by co-accreditation partnership between the universities, higher education institutes of agronomy and agricultural colleges. When considering only Master and Doctorate levels, Biosystems Engineering is only taught in three higher educational institutes of Agronomy (Beauvais, Dijon and Montpellier), contrary to the other European countries which get specialized universities. This French complex situation is evolving towards a European integration through ERASMUS exchanges, adoption of semesters for the studies . The need of correspondence between level of skills in France and Europe is crucial, to ensure students and staff exchanges. This specific work has been done (in terms of levels and domains of BE) as part of the Joint Technological Network (JTN) AgroETICA (Agricultural engineering and ICT for Agroecology) dedicated to agricultural engineering development. This JTN includes manufacturers, dealers, agricultural organisms, research and teaching organisms, and Ministry representatives. The conclusions are that the French system needs to be positioned at a European scale for the BE studies and skills, to allow exchanges and creation of new courses, in relation with industrial partners.

National audience

Evaluation of carcass color at the slaughterhouse Expert notation is the reference method for the evaluation of bovine meat color, notably for quality. A total of 140 Charolais carcasses were scored at the slaughterhouse on a color scale going from 1 (too light) to 5 (too dark). These expert scores were compared to three types of instrumental data (L*a*b / RVB referential; image shooting with image processing / chromameter). Amongst the instrumental measurements, the values obtained after image processing of the whole muscle seems to be better associated with the scores obtained for color by experts than the scores obtained with the chromameter for which the zone of the muscle analyzed is more limited. Light / dark carcasses, which are excluded from the AOC Beef of Charolles appellation, can be distinguished by extreme instrumental variations. It is more difficult to differentiate the intermediary carcasses (light, bright red and/or dark) on the basis of their RVB or L*a*b values since the distribution of points is superimposed. Nevertheless, it would be interesting to evaluate the possibility of adding a third layer to the different expert scores in order to homogenize the colors within each layer and thus be able to associate them more easily to the instrumental values of color. To do this, complementary analyses should be performed on a larger number of carcasses representing a large panel of colors., La présente étude vise à proposer des éléments d’objectivation de la notation experte de la couleur, qui constitue en général la méthode de référence. Elle met en évidence que les notes extrêmes de couleur correspondent à des valeurs instrumentales significativement différentes et qu’il est donc possible par méthode instrumentale (chromamètre, prise de vue) d’éliminer les carcasses dont la couleur n’est pas conforme au cahi er des charges AOC Bœuf de Charolles. Résumé : La notation experte est la méthode de référence pour évaluer la couleur des viandes bovines, notamment dans les filières de qualité. Au total, 140 carcasses Charolaises ont été notées à l’abattoir sur une échelle de couleur allant de 1 (trop clair) à 5 (trop foncé). Ces notes expertes ont été comparées à trois types de données instrumentales (référentiels L*a*b* / RVB ; prise de vue photographique avec tra itement d’image / chromamètre). Au sein des mesures instrumentales, les valeurs obtenues après traitement d’image sur le muscle entier semblent mieux reliées aux notes expertes de couleur que les valeurs obtenues avec le chromamètre, dont la zone d’étude représente une partie plus limitée du muscle. Les carcasses trop claires / trop foncées, qui sont de fait exclues de l’agrément AOC Bœuf de Charolles, se distinguent par des v aleurs instrumentales extrêmes. Il est plus difficile de différencier les carcasses intermédiaires (claires, rouge vif et/ou foncées) sur la base de leurs valeurs RVB ou L*a* b* car les nuages de points se superposent. Néanmoins, il serait intéressant d’évaluer la possibilité d’ajouter des tiers de classe aux différentes notes expertes de façon à homogénéiser les couleurs au sein de c haque tiers de classe et ainsi d’envisager les lier plus aisément aux valeurs instrumentales de couleur. Pour ce faire, des analyses complémentaires devront être menées sur un effectif plus important de carcasses, représentant un grand panel de couleurs

AgrosupBAPGEAPSI; DAMAV project for vineyard disease detection by UAV imagery . CIRG-AgEng 2016, International Conférence on Agricultural Engineering, Automation, Environment and Food Safety

ARTICLE A NE PAS DIFFUSEREAGEAPSIAgrosupCT1; Solution de détection des maladies de la vigne par imagerie de drone. Diagnostic et réduction des pesticides à la parcelle

International audience

Initially developed for technical industrial sectors such as medicine or aeronautics, imaging technics are more and more used since 30 years in agriculture and viticulture. The development of acquisition tool and the decreasing of the calculation time allowed using imagery in laboratory under controlled conditions. At the beginning of the 90’s, the concept of Precision Farming has been developed in the USA, considering a field as a heterogeneous area needing different input in terms of fertiliser or protection product. In the same time, the aperture of the GPS system for civil applications has allowed the development of remote sensing domain. Combining GPS information and imagery conducted also to the emergence of proxy-detection applications, in agriculture and viticulture domains, in order to optimize crop management. A localized crop management needs the use of new technologies such as computing, electronics and imaging, and the conception of a proxy-detection system is motivated by the need of better resolution, precision, temporality and lower cost, compared to remote sensing. The use of computer vision techniques allows obtaining this information automatically with objective measurements compared with visual or manual acquisition. The main applications covered by the computer vision in agriculture are tied to the crop characterization (biomass estimation, leaf area, volume, height of the crop, disease determination etc), the aerial or root phenotyping in the fields or in specific platforms and the understanding of spraying and spreading processes. This presentation will explain the different imaging systems used to characterize the previous parameters, in 2D or 3D. It will also give some details on the dedicated image processing methods developed, related to motion estimation, focus information, pattern recognition and multi –hyper spectral data.

Au-delà des techniques d’acquisition d’images en proxy-détection et de la télédétection classique, l’imagerie aérienne revient en force dans le domaine agricole, et se développe fortement pour tout type d'applications : détection de rendement, détection de maladies, suivi de croissance … Le présent projet vise ainsi à utiliser l’imagerie aérienne à partir de drones, couplée à du traitement et de l’analyse d’images basés sur des informations multiples, pour détecter les foyers infectieux liés à la flavescence dorée dans les vignes bourguignonnes. Les enjeux sont à la fois économiques, pour cibler les traitements phytosanitaires et réduire les doses apportées, et sanitaires pour essayer de circonscrire les foyers détectés, la flavescence dorée touchant environ 55% du vignoble français. Les algorithmes développés, basés sur des approches couplant des informations texture/couleur/spectrales, et combinés à des approches neuro-inspirées, ont permis de localiser des zones infestées dans plusieurs parcelles du sud de la France et en Bourgogne, et ce quel que soit le degré d'infestation. Les objectifs finaux sont alors de fournir à la fois une base de données d’images de vignes infestées (avec l'ensemble des informations pertinentes), mais aussi des algorithmes spécifiques déterminant les zones infestées correspondant à la flavescence dorée. Ces résultats sont en cours de confrontation à des analyses en laboratoire confirmant la présence de la flavescence dorée, dont les symptômes peuvent visuellement se rapprocher de symptômes d’autres maladies. En parallèle à ce projet, et dans l’objectif d’exploiter les images de mêmes parcelles prises à des altitudes et résolutions différentes, nous développons actuellement de nouvelles méthodes mathématiques de fusion d’images.

National audience; Une pulvérisation de précision en viticulture implique une maîtrise conjointe du matériel de pulvérisation, des produits et de la répartition de ces produits sur le feuillage. Dans le contexte de viticulture de précision, nombreuses sont les recherches menées sur l’optimisation d’utilisation des produits phytosanitaires. L’objectif final étant de réduire de manière significative la quantité d’intrants dans les cultures. Néanmoins, dans ce cadre, peu de travaux ont été effectués pour tenter de déterminer les comportements des produits directement sur le feuillage. Les travaux présentés dans cette communication s’intéressent ainsi particulièrement à l’analyse de l’état de surface foliaire qui présente une part essentielle dans le processus d’adhésion du produit pulvérisé sur la feuille. L’analyse de surface de la feuille est réalisée en déterminant les caractéristiques texturales extraites d’images microscopiques. Des paramètres fréquentiels et spatiaux, basés sur les Descripteurs Généralisés de Fourier, ont été utilisés pour estimer la rugosité de la feuille. Ces paramètres permettent la caractérisation de l’homogénéité de la surface et la détection des nervures, des poils et des champignons au niveau de la surface de la feuille. Ce premier travail portant sur une étude fondamentale et technologique pour mieux comprendre le comportement des gouttelettes (impact et déplacement) à l’approche et sur le feuillage de vignes a été approfondi dans un objectif de définition et de conception d’un modèle statistique d’estimation de la quantité du produit restant sur la surface de la feuille. Ce modèle prend en considération différents paramètres de pulvérisation, tels que le type de buse et la vitesse de pulvérisation, la tension superficielle du produit, l’angle d’inclinaison et la rugosité de la surface de la feuille. Ce modèle pourra ainsi être vu comme un outil de décision commun pour optimiser la quantité du produit pulvérisé et l’estimation du volume du produit restant sur la feuille, ce dernier travail étant en cours de finalisation.

EA SPE GEAPSI; International audience; A 3D imaging technique using a high speed binocular stereovision system was developed in combination with corresponding image processing algorithms for accurate determination of the parameters of particles leaving the spinning disks of centrifugal fertilizer spreaders. Validation of the stereo-matching algorithm using a virtual 3D stereovision simulator indicated an error of less than 2 pixels for 90% of the particles. The setup was validated using the cylindrical spread pattern of an experimental spreader. A 2D correlation coefficient of 90% and a Relative Error of 27% was found between the experimental results and the (simulated) spread pattern obtained with the developed setup. In combination with a ballistic flight model, the developed image acquisition and processing algorithms can enable fast determination and evaluation of the spread pattern which can be used as a tool for spreader design and precise machine calibration.

International audience; Nowadays farmers recognize the importance of a correct and precise fertiliser application: non-uniform spread patterns cause extra pressure on the environment and might result in economic losses for the farmer. In Europe most spreading is done by centrifugal fertilizer spreaders but their spreading process is not easy to monitor and to control. To perform a precise fertilising farmers need proper tools to determine and evaluate the spread pattern at farm level. Therefore the Flemish Institute for Agricultural and Fisheries Research (ILVO) is exploring and developing a fast and accurate technique for measuring the spread pattern of conventional centrifugal spreaders. This method aims to be low cost and applicable at the farm level. Also, the proposed solution has to be mobile to the extent that the device can be built up on site to test several machines. The device should enable the adjustment of the spreader in such a way that a uniform spread pattern is obtained. Three approaches for evaluating the spread pattern are currently available: the collector tray method, the full modelling approach like the Discrete Element Method (DEM), and the hybrid approach that combines measurements and modelling. In this research the hybrid approach was applied: the spread pattern was predicted with a ballistic flight model based on the measurement of the horizontal outlet angle, the vertical outlet angle, the grain diameter, the grain density and the initial velocities. In a first step, a 2-dimensional imaging techniques were used with a small field of view (0.33m x 0.25m) to measure the horizontal outlet angle and the speed of the grains at different camera positions at the circumference of the disk. The vertical outlet angle and the mass distribution were measured with a cylindrical collector. The grains flying under the measurement unit were imaged using two different techniques: the high speed technique and a newly developed stroboscopic imaging technique. For the high speed technique a camera, type MotionXtra HG 100K (Roper Scientific, New Jersey, USA), was used. The stroboscopic technique combined a specially designed LED stroboscope with the Nikon D 100 camera. Overall the stroboscopic technique and the high speed technique were capable of measuring the outlet angle and the outlet speed. Small differences between the measurements with both techniques existed, but ultimately we were interested in the resulting spread pattern in the field. When comparing the simulated spread pattern measured in a spread hall, relative errors amounted up to 30%. Therefore, in the next phase of the research the 2-dimensional imaging techniques were replaced with a 3D stereovision technique, with a much larger field of view ( 1m x 1m) and improved motion estimation algorithms, resulting in lower relative errors between the simulated and the measured cylindrical spread pattern. In the final phase the small field of view with strobe as used in the first approach will be combined with the 3D stereovision set-up (and related algorithms) to further improve the simulation of the spread pattern.

Présentation orale/ Communication avec actes

Under conditions of low nitrogen in soil, legumes have the ability to fix nitrogen in nodules. Understanding these mechanisms is currently a major focus in the field of agronomy. In this context, we need to characterize the interactions between plants and environment by "pheno-methods". Our study consists in developing a morphological data processing tool to characterize the growth of organ using nondestructive and automated innovative techniques. The research is incorporated through the High-Throughput Phenotyping Platform (PPHD) of the INRA Centre of Dijon. The patented system installed on the PPHD platform, the Rhizocab, is able to monitor the development of plant root system without destruction. The rhizotron images acquired from this system are the input data for our research. Although there are several commercial tools on the market to treat rhizotron images, none is suitable with our problem. However, there is currently no automatic method for the detection of phenotypic traits in image processing. We present here a set of methods and a program for the quantitative analysis of root image. Once the rhizotron image was acquired, the program allows us to extract the root from the noisy background. For the first period of our research, we work on plants with a taproot system, so the next step is to extract the primary root. Step by step, we can finally calculate several parameters which biologists need such as: the lengths of the primary root, the local diameter of the primary root, the number of nodules and their locations, the number of lateral roots on the primary root etc. Spatial analysis of nodulated roots provides the possibility of comparative analysis between phenotypes and genotypes of legume species, also the possibility to study the dynamic interaction between genetic background and the physical world. This approach could also be used to analyze non taproots system and even other species than legume species.

International audience; Agricultural pesticide sprays are applied with different types of spray nozzles each with its own spray characteristics. The objective of this study was to measure the macro-spray characteristics (spray angle, liquid sheet length, spray shape) from different types of hydraulic spray nozzles using a developed backlighted image acquisition system and image processing technique. Tests included five different nozzles (Albuz ATR orange and red, TeeJet XR 110 01, XR 110 04 and Al 110 04). The results were compared and related well with the results from existing measuring techniques like a flow rate test bench, a spray distribution bench (horizontal patternator) and a Phase Doppler Particle Analyzer (PDPA) laser.

National audience; L'étape principale de la méthode de reconstruction 3D " Shape from Focus " est l'utilisation d'un opérateur de mesure de netteté de chaque pixel de la séquence d'image. Le choix de l'opérateur de mesure de netteté est une étape cruciale pour une reconstruction 3D de qualité. La précision de la mesure de netteté dépend de la taille du voisinage autour du pixel choisi et de la présence ou non de bruit additif dans la séquence d'images. Dans cet article, nous présentons deux nouveaux opérateurs de mesure de netteté basés sur les Descripteurs Généralisés de Fourier. Une nouvelle étude comparative des différents opérateurs est présentée. Cette comparaison est basée sur un plan d'expériences mettant en évidence les effets de la variation du voisinage utilisé et le bruit additif sur la précision de la reconstruction. Les résultats font office de recommandation pour le choix de l'opérateur à utiliser en fonction du voisinage choisi et du bruit présent. Les résultats de nos opérateurs montrent qu'ils sont à recommander lorsqu'on est en présence d'une scène bruitée et que l'on travaille sur un petit voisinage.

National audience

International audience

La viticulture est confrontée à des problèmes majeurs dont la lutte contre les maladies cryptogamiques telles que l’oïdium ou la pourriture grise. Elle utilise alors majoritairement des produits phytosanitaires de synthèse susceptibles de poser des problèmes environnementaux et sanitaires, et de favoriser la sélection de souches de microorganismes pathogènes résistantes. Dans le cadre du plan Ecophyto 2018, la recherche de méthodes visant à diminuer, voire à remplacer, les fongicides de synthèse au vignoble, a pris toute sa place. Deux approches peuvent alors être conduites pour répondre à cet objectif : 1) la réduction du nombre de traitements et des doses appliquées et 2) le développement de stratégies alternatives ou complémentaires. La réduction du nombre de traitements s’appuie sur des outils d’aide à la décision, dont la modélisation. Couplée à des données obtenues à partir d’acquisitions et de traitements d’images adaptés (thermographie IR, proche IR …), la modélisation doit ainsi permettre la prédiction de la cinétique de croissance, en fonction de paramètres environnementaux. Cette prédiction sera alors d’une grande utilité pour les prises de décision du viticulteur (période précise de traitement, prévision du tri de la vendange …). Les travaux qui sont menés au sein de l’UMR Agroécologie agrègent ainsi des compétences en pathosystèmes et en imagerie pour concevoir les outils adéquats. La lecture de la résistance des pathosystèmes est réalisée par imagerie de la sporulation, du brunissement ou de l’altération de la couleur. Les outils ainsi développés doivent permettre de caractériser les maladies d’abord sous serre (PPHD (Plateforme de Phénotypage Haut Débit)), puis de transposer ces techniques au vignoble. D’autres maladies (flavescence dorée) pourraient alors être déterminées.

International audience

International audience; In precision spraying, spray application efficiency depends on the pesticide application method, the phytosanitary product as well as the leaf surface properties. For environmental and economic reasons, the global trend is to reduce the pesticide application rate of the few approved active substances. Under these constraints, one of the challenges is to improve the efficiency of pesticide application. Different parameters can influence on pesticide application as nozzle types, liquid viscosity and leaf surface. Specific models have been developed showing that the predominant factor for the leaf is the leaf roughness, because it is related on adhesion mechanisms of liquid on surface and it is used to define nature of leaf (hydrophobic/hydrophylic). In this work, we focus on the leaf surface properties, in particular the vine leaves. We propose to discriminate between two kinds of vine leaves (Pinot and Chardonnay) for different stages of development to follow their roughness growth. Moreover, this discrimination has an impact on the product behavior, and allows adjusting the product viscosity and spraying parameters according to the roughness and the stage of vine leaf development. In the purpose of vine leaf discrimination, it is interesting to use texture analysis because of spatio frequential aspect of the features that can be extracted. In order to test the proposed leaf classification, experiments have been done on images acquired with a SEM microscope. These images represent various surfaces (above, below, with and without rib) of the vine leaves for different stages of development: "young" and "mature" leaves. Generally, the major issue with natural texture classification is the sensitivity to illumination, changes of scale and orientation. The overall performance of a texture classifier may be totally degraded if the unknown patterns to be classified are slightly rotated with respect to the training samples. In our case we consider the invariant features (scale, illumination and rotation) called Generalized Fourier Descriptors (GFD). The application of Generalized Fourier Descriptors allows the extraction of a vector robust texture features. However, this vector has high dimensionality which can cause erroneous classification due to the Hughes phenomenon. To avoid this constraint, it is interesting to apply dimensionality reduction (RD) techniques in order to obtain representative data with reduced size. Nonlinear RD technique as Kernel Discriminant Analysis (KDA) is used, which is the most commonly used and the most suitable for our application. A multi-layer perceptron will be used as classifier tool. The result shows that the combination of GFD and KDA appears to provide sufficient information to characterize vine leaves for different stages of development. The rate of classification for young leaves is 100%, and 94.44% for mature leaves of Chardonnay, 97.22% for young leaves and 98.55% for mature leaves of Pinot.

International audience; The application of plant protection product has an important role in agricultural production processes. With current pesticides management, a huge amount of them are applied to worldwide orchards. In precision spraying, spray application efficiency depends on the pesticide application method, the phytosanitary product as well as the leaf surface properties. For environmental and economic reasons, the global trend is to reduce the pesticide application rate of the few approved active substances. Under these constraints, one of the challenges is to improve the efficiency of pesticide application. Different parameters can influence pesticide application such as nozzle types, liquid viscosity and leaf surface. In this paper, we focus on the vine leaf surface properties and discrimination between two kinds of vine leaves (Pinot and Chardonnay) in different stages of development for following their roughness growth. This discrimination allows studying the impact of the product behavior, and allows to adjusting the product viscosity and spraying parameters according to the roughness and the stage of vine leaf development. In this context, we propose to explore the performance of combination of Generalized Fourier Descriptor with Kernel Discriminant Analysis methods using neural network. The result appears provide sufficient information to characterize vine leaves.