Olive's trees growing, especially intended for olive oil production, is a central agricultural activity in the Mediterranean area and increasing efforts are directed toward reducing its production costs. Under this perspective, mechanical harvesting is a crucial task to improve the profitability of olive groves. In medium and high density, i.e. up to 600 trees per hectare, olive orchards trunk shaking is the most common practice, where harvesting efficiency greatly depends on several parameters, such as tree dimensional traits, cultivar, fruit ripeness indexes. In this contest, the ratio between the fruit removal force and fruit mass has been proposed as a predicting index for the harvesting efficiency, but at the same time has been recognized that such index it's not independent from the other above mentioned crop parameters (e.g. cultivar or canopy volume). So, the harvesting efficiency can be considered the output of a complex system with nonlinear relationships among input data, and where the exact nature of the interactions is unknown. In this situation, artificial neural network (ANN) it is a potentially suitable system to tackle the problem. This approach has been followed in the present work by gathering data from literature about olive mechanical harvesting efficiency and as much as possible of the underlying crop parameters. The latter have been used as input variables of an ANN model built to predict the olive harvest efficiency of trunk shakers. The model (10 input, 1 hidden layer, 2 nodes, learning rate 0.3, momentum 0.4) allows a satisfactory prediction of the harvest efficiency, as stated by a correctness of the internal validation of about 86% (within 5% range of validating internal data set), a determination coefficient of 0.88 between desired and predicted data, and a determination coefficient of 0.93 between actual and predicted data of an external validating data set. The results of this first ANN approach based on metadata indicate that this computing system could be effectively implemented at single orchard level as a tool for olive mechanical harvest optimization.

An artificial neural network model to predict olive mechanical harvesting: A first approach based on metadata / piernicola masella, giulia angeloni, agnese spadi, lorenzo guerrini, alessio cappelli, alessandro parenti, fabio baldi, enrico cini. - In: ACTA HORTICULTURAE. - ISSN 0567-7572. - ELETTRONICO. - 1311:(2021), pp. 0-0. ((Intervento presentato al convegno VI International Symposium on Applications of Modelling as an Innovative Technology in the Horticultural Supply Chain Model-IT 2019 tenutosi a Molfetta, Italy nel 09-12 giugno 2019 [10.17660/ActaHortic.2021.1311.45].

An artificial neural network model to predict olive mechanical harvesting: A first approach based on metadata

piernicola masella
;
giulia angeloni;agnese spadi;lorenzo guerrini;alessio cappelli;alessandro parenti;fabio baldi;enrico cini
2021

Abstract

Olive's trees growing, especially intended for olive oil production, is a central agricultural activity in the Mediterranean area and increasing efforts are directed toward reducing its production costs. Under this perspective, mechanical harvesting is a crucial task to improve the profitability of olive groves. In medium and high density, i.e. up to 600 trees per hectare, olive orchards trunk shaking is the most common practice, where harvesting efficiency greatly depends on several parameters, such as tree dimensional traits, cultivar, fruit ripeness indexes. In this contest, the ratio between the fruit removal force and fruit mass has been proposed as a predicting index for the harvesting efficiency, but at the same time has been recognized that such index it's not independent from the other above mentioned crop parameters (e.g. cultivar or canopy volume). So, the harvesting efficiency can be considered the output of a complex system with nonlinear relationships among input data, and where the exact nature of the interactions is unknown. In this situation, artificial neural network (ANN) it is a potentially suitable system to tackle the problem. This approach has been followed in the present work by gathering data from literature about olive mechanical harvesting efficiency and as much as possible of the underlying crop parameters. The latter have been used as input variables of an ANN model built to predict the olive harvest efficiency of trunk shakers. The model (10 input, 1 hidden layer, 2 nodes, learning rate 0.3, momentum 0.4) allows a satisfactory prediction of the harvest efficiency, as stated by a correctness of the internal validation of about 86% (within 5% range of validating internal data set), a determination coefficient of 0.88 between desired and predicted data, and a determination coefficient of 0.93 between actual and predicted data of an external validating data set. The results of this first ANN approach based on metadata indicate that this computing system could be effectively implemented at single orchard level as a tool for olive mechanical harvest optimization.
ISHS Acta Horticulturae 1311: VI International Symposium on Applications of Modelling as an Innovative Technology in the Horticultural Supply Chain Model-IT 2019
VI International Symposium on Applications of Modelling as an Innovative Technology in the Horticultural Supply Chain Model-IT 2019
Molfetta, Italy
09-12 giugno 2019
Goal 11: Sustainable cities and communities
Goal 12: Responsible consumption and production
piernicola masella, giulia angeloni, agnese spadi, lorenzo guerrini, alessio cappelli, alessandro parenti, fabio baldi, enrico cini
File in questo prodotto:
File Dimensione Formato  
2021 An artificial neural network model to predict olive.pdf

Accesso chiuso

Tipologia: Pdf editoriale (Version of record)
Licenza: DRM non definito
Dimensione 534.38 kB
Formato Adobe PDF
534.38 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2158/1238546
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? ND
social impact