Click on “Download PDF” for the PDF version or on the title for the HTML version.


If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Determination of Leaf Water Content with a Portable NIRS System Based on Deep Learning and Information Fusion Analysis

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  Transactions of the ASABE. 64(1): 127-135. (doi: 10.13031/trans.13989) @2021
Authors:   Lei Zhou, Chu Zhang, Mohamed Farag Taha, Zhengjun Qiu, Yong He
Keywords:   Decision fusion, Deep learning, Leaf water content, Local computing, Portable NIRS system.

Highlights

A portable NIRS system with local computing hardware was developed for leaf water content determination.

The proposed convolutional neural network for regression showed a satisfactory performance.

Decision fusion of multiple regression models achieved a higher precision than single models.

All of the devices and machine intelligence algorithms were integrated into the system.

Software was developed for system control and user interface.

Abstract. Spectroscopy has been widely used as a valid non-destructive technique for the determination of crop physiological parameters. In this study, a portable near-infrared spectroscopy (NIRS) system was developed for rapid measurement of rape (Brassica campestris) leaf water content. An integrated spectrometer (900 to 1700 nm) was used to collect the spectra. A Wi-Fi module was adopted for driving the spectrometer and realizing data communication. The NVIDIA Jetson Nano developer kit was employed to handle the received spectra and perform computing tasks. Three embedded spectral analysis models, including support vector regression (SVR), partial least square regression (PLSR), and deep convolutional neural network for regression (CNN-R), and decision fusions of these methods were built and compared. The results demonstrated that the separate models produced satisfactory predictions. The proposed system achieved the highest precision based on the fusion of PLSR and CNN-R. The hardware devices and analytical algorithms were all integrated into the proposed portable system, and the tested samples were collected from an actual field environment, which shows great potential of the system for outdoor applications.

(Download PDF)    (Export to EndNotes)