Three-dimensional quantitative characterization of grapes morphology and possible relation with grey mould susceptibility

Grey mould is one of the most important diseases of grapevine in the Mediterranean regions caused by the fungi Botrytis cinerea. Many factors are responsible for this disease among them, the morphology of grapes plays a crucial role in grey mould infection. The grapes with highly compact berries are the most susceptible to infection. The common methods applied to evaluate the compactness of grapes cannot apply to grapevine bunches from the same variety. Therefore, novel methods are used to detect compactness by image processing analyses such as photogrammetry for 3D model reconstruction. This study proposes an alternative analysis of bunch morphology and compaction assessment based on virtual 3D models. Seventeen Pinot Gris clones and six Pinot Noir clones were manually collected at harvest time, and the grey mould severity evaluation was carried out in the field. All the grapes were photographed at different angulations, and the 3D model reconstruction was performed by the photogrammetry technique. Several measures and indexes were extracted from each bunch. Principal component analysis (PCA) and two multiple linear regression models (MLR) were applied to identify the descriptors of the clones most related to grey mould infection. The first model assessed the correlation between the grey mould severity and the descriptors from the 2D analysis, while the second model analyzed both descriptors from the 2D and 3D analysis. The 3D MLR presented higher performances than the 2D MLR. The R-square value (R2) and the root mean square error (RMSE) were compared between models. For Pinot Gris, the R2 rose from 0.656 to 0.838, moving from the 2D to the 3D MLR, while the RMSE decreased from 1.713 to 1.175. In Pinot Noir, the 2D model did not provide sufficient robustness, while the proposed MLR estimated R2 with 0.936 value and RMSE with 0.29 value. Additional studies were performed by analyzing the data with graphs and statistics. Consequently, the most significant traits include the estimated empty volume, the width of the grape, weight, volume, shape, and the ratio between surface and height.