Estimation of grapes yield is critical for both farmers and winemakers in order to improve human labor and resources allocation and, finally, product pricing. This paper presents a novel grape yield estimation method based on Intervals’ Numbers (INs). The advantage of INs is their ability to process distributions of samples. The INs here are induced from successive grape clusters’ weight measurements during ripening. A parametric neural network (NN) IN-regressor architecture is trained with past INs in order to predict future INs. The training process involves the calculation of optimal parameters to minimize the estimation error in terms of the metric distance between real and estimated INs. The method is tested on real world weight data involving 30 grape clusters of three different grape cultivars, collected once in a week from veraison to harvest time. Preliminary results indicate a weight estimation accuracy of up to 96.33% and absolute estimation error of 16.22 gr per vine. The proposed method can be integrated in a crop growth model towards early grape yield estimation.


Christos Bazinas, Eleni Vrochidou, Chris Lytridis, and Vassilis G. Kaburlasos. 2021. Yield Estimation in Vineyards Using Intervals’ Numbers Techniques. In 25th Pan-Hellenic Conference on Informatics (PCI 2021). Association for Computing Machinery, New York, NY, USA, 454–459. DOI:https://doi.org/10.1145/3503823.3503906

View on ACM