Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 1
Encontrada(s) 1 página(s)

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Volatile Fingerprinting Empowers Salinity Monitoring in Peppermint Using MOS Sensors and Feature-Optimized Machine Learning

  • Ahmad, Ali|||0000-0001-5530-7374
  • Silva-Alvarado, Vinie Lee|||0009-0000-5857-3248
  • Sendra, Sandra|||0000-0001-9556-9088
  • Lloret, Jaime|||0000-0002-0862-0533
  • Heydari, Arman
[EN] Early detection of salinity stress is essential for precision agriculture, particularly in scalable, resource-constrained monitoring systems. This study presents a portable sensing module integrating a low-cost metal oxide semiconductor (MOS) sensor array with potential application for edge deployment to detect salinity stress in peppermint. Salinity significantly reduced plant biomass, confirming physiological stress induction. Volatile organic compound (VOC) fingerprints were collected over eleven consecutive days in a controlled enclosure. Sensor signals underwent outlier filtering, normalization, and smoothing, while treatment discrimination was verified using the Kruskal¿Wallis test. Thirty-three machine learning models were evaluated using a 75:25 train¿test split with five-fold cross-validation. Wide neural network models achieved the highest predictive performance, exceeding 98% test accuracy and a 97% macro F1 score. Feature adequacy analysis showed that six sensors captured the dominant variance required for reliable classification. Considering computational constraints, a bilayered neural network using only six features maintained over 97% accuracy with a memory footprint of 0.008 MB while remaining Pareto optimal. These findings support the feasibility of a compact, computationally efficient, and edge-compatible VOC sensing framework for salinity stress detection in precision agriculture and intelligent crop monitoring., Part of this work was funded by the Grant PRE2021-100809 funded by Agencia Estatal de Investigación (AEI) MICIU/AEI/10.13039/501100011033, by European Social Fund Plus (ESF+); by Universitat Politècnica de València through the Programa de Ayudas de Investigación y Desarrollo (PAID-01-24) ; and by Universidad Nacional José Faustino Sánchez Carrión through the Concurso de Proyectos de Investigación para Docentes Año 2026 ratified by RCU-0969-2026-CU-UNJFSC. This work has also received funding from the PRIMA Programme supported by the European Union, project PAVER, and it is supported by MCIN/AEI/10.13039/501100011033. Funders had no role in the design and execution of the study.