Skip to Main content Skip to Navigation

Potentialités de l’imagerie couleur embarquée pour la détection et la cartographie des maladies fongiques de la vigne

Abstract : The downy mildew of the vine is a phytopathology of fungal origin particularly worrying for the wine industry. The aim of this thesis is to study the potentialities of on-board color imaging to estimate the health status of vineyards affected by downy mildew on an intra-plot scale. The proposed solution aims at assisting epidemiological monitoring networks in the estimation of health risks and in the recommendation of chemical control plans. In practice, two image processing chains are proposed, one dedicated to the segmentation of vine organs, and the other to the detection, counting and measurement of symptomatic tissues of downy mildew. These two chains are designed on a common strategy and are aimed at images acquired directly at the plot under the conditions of viticultural work. The proposed strategy is based on structure-color representations and probabilistic models of the tissue classes present in the vines. It operates in three steps : Formulating descriptors to extract the characteristic and discriminating properties of each class ; Modelling the statistical distributions of these descriptors in each class ; Assigning each pixel to on of the classes according to its suitability to their models. The descriptors combine the Local Structure Tensors (LST) with colorimetric statistics calculated in the neighborhood of the pixel considered. To account for the specific nature of LSTs, the descriptors are transformed to be represented in the Log-Euclidean space. In this space, it becomes possible to model the classes of interest by distributions of multivariate Gaussian mixtures of structure-color representations. This strategy is first applied to healthy vine images. It involves the partitioning of an image into organ classes (foliage, bunches or inflorescences and stems). A pixel-wise MAP (Maximum A Posteriori) classification is carried out and regularized by stochastic relaxation and mathematical morphology operations. The results obtained for three phenological stages are very convincing. In addition, the processing chain is robust to the setting of the main hyperparameters. In a second step, the previous methodological chain is adapted to process images with symptoms of downy mildew as well as necrosis, discolorations, deficiencies, mechanical wounds, which are confounding factors. The decision method is based on a reconstruction of symptoms by propagation around germs. The criteria used are based on the previously defined color-structure representations and probabilistic models. The new processing chain reliably detects downy mildew symptoms and estimates the area of the affected tissues.
Document type :
Complete list of metadatas

Cited literature [174 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Thursday, March 5, 2020 - 11:30:11 AM
Last modification on : Tuesday, August 25, 2020 - 3:38:41 AM
Long-term archiving on: : Saturday, June 6, 2020 - 2:05:50 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02499420, version 1


Florent Abdelghafour. Potentialités de l’imagerie couleur embarquée pour la détection et la cartographie des maladies fongiques de la vigne. Physique [physics]. Université de Bordeaux, 2019. Français. ⟨NNT : 2019BORD0430⟩. ⟨tel-02499420⟩



Record views


Files downloads