Version 1
: Received: 19 February 2024 / Approved: 20 February 2024 / Online: 20 February 2024 (06:17:02 CET)
How to cite:
An, A.; Wang, R. Use of HSI Coupled with Machine Learning for the Identification of Water Stress in Pear Seedling Leaves. Preprints2024, 2024021103. https://doi.org/10.20944/preprints202402.1103.v1
An, A.; Wang, R. Use of HSI Coupled with Machine Learning for the Identification of Water Stress in Pear Seedling Leaves. Preprints 2024, 2024021103. https://doi.org/10.20944/preprints202402.1103.v1
An, A.; Wang, R. Use of HSI Coupled with Machine Learning for the Identification of Water Stress in Pear Seedling Leaves. Preprints2024, 2024021103. https://doi.org/10.20944/preprints202402.1103.v1
APA Style
An, A., & Wang, R. (2024). Use of HSI Coupled with Machine Learning for the Identification of Water Stress in Pear Seedling Leaves. Preprints. https://doi.org/10.20944/preprints202402.1103.v1
Chicago/Turabian Style
An, A. and Ran Wang. 2024 "Use of HSI Coupled with Machine Learning for the Identification of Water Stress in Pear Seedling Leaves" Preprints. https://doi.org/10.20944/preprints202402.1103.v1
Abstract
Water plays a vital role in the healthy growth of pears. Early detection of water stress can play a significant role in the timely management of water deficiency for pear yield. Most current methods are labour-intensive, time-consuming, only provide point measurements, and could be destructive. In this research, a real-time non-destructive method using a push-broom hyperspectral system (400-1000nm) was used to collect hyperspectral image data and detect the water stress of the pear seedling leaves. To build a reliable prediction model, machine learning techniques were used. The Successive Projections Algorithm (SPA) was applied for optimal wavelength selection. In particular, CNN was applied to obtain the features of key wavelengths. Both the CNN features and key wavelengths were put into RR-MLR, BLR and ENN for analysis. The training accuracy of the three modellings all reach the accuracy above 70% after about 100 epochs, while combination of CNN features outperformed the mere main spectra analysis. This research demonstrated that hyperspectral imaging coupled with machine learning techniques could be applied to predict the water content of pear leaves predict non-destructively.
Keywords
hyperspectral imaging; plant water stress; machine learning; optimal wavelengths
Subject
Biology and Life Sciences, Horticulture
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
