Fuzzy reasoning for long cane pruning : a relaxation algorithm for complex vine configurations

Long cane pruning cannot be performed at present with machines which do not allow the selection and the saving of the desired wood. Research is being conducted to design a system enabling such an automatic selection. The system consists of an image processor which individualises and tracks the different woods in the image. A decision device simulates the pruner's reasoning by choosing the most appropriate long wood to save according to wood features measured on the images. This decision device is based on a fuzzy multi-criteria decision-making (FMCDM), the preference defined by th expert on the criteria values are represented by fuzzy sets. Results show that the decision system fitted with the pruner's decisions in the vineyard. The decision system which was developed is able to work an standard vine pruning cases. Sometimes, however, the vine configuration is not good (problems of frost, hail storm or mechanised operations can damage woods). In such cases, the proposed decision system is not able to choose the most appropriate long wood because it is not accurate enough to make a distinction among very bad woods. Nevertheless, the pruner always takes a decision. In such cases, the pruner's reasoning is not the same, the pruner becomes more laxist and makes a relaxation of his preferences. This paper presents the relaxation algorithm which was used to simulate the pruner's reasoning when the system faces non standard pruning cases. / Le travail a pour but de développer un système pour choisir automatiquement le long bois à préserver. Le système comprend un processeur d'images qui individualise les sarments. Le système de décision simule le raisonnement du tailleur en fonction des paramètres mesurés sur l'image : il est basé sur un système de logique floue multi-critères. Le système de décision fonctionne pour les cas de vigne standard. Dans le cas de vignes complexes, tous les sarments sont trop mauvais. Dans ce cas le raisonnement n'est plus le même. Le papier présente l'algorithme de relaxation utilisé pour simuler le raisonnement du tailleur dans ces cas