Skip to main content
ARS Home » Southeast Area » Stoneville, Mississippi » Crop Production Systems Research » Research » Publications at this Location » Publication #365789
Title: Machine learning-based spectral library for crop classification and status monitoring

Author
item ZHANG, JINGCHENG - Hangzhou Dianzi University
item HE, YUHANG - Hangzhou Dianzi University
item YUAN, LIN - Zhejiang University
item ZHOU, XIANFENG - Hangzhou Dianzi University
item Huang, Yanbo

Submitted to: Agronomy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/26/2019
Publication Date: 8/29/2019
Citation: Zhang, J., He, Y., Yuan, L., Zhou, X., Huang, Y. 2019. Development of a plant spectral library for crop classification and status monitoring. Agronomy. 9, 496. https://doi.org/10.3390/agronomy9090496.
DOI: https://doi.org/10.3390/agronomy9090496

Interpretive Summary: It is important to create a spectral library as a standard reference for agricultural remote sensing analysis and mapping. Scientists from Hangzhou Danzi University, Zhejiang University of Water Resources and Electric Power and USDA ARS Crop Production Systems Research Unit at Stoneville, Mississippi have collected spectral data of crop plants, tea and rice, and evaluated machine learning algorithms for feature extraction from these data to classify crops by characterizing crop growth. The results indicated that these algorithms could classify the crops well with crop growth characterization and proved that the creation of a standard spectral library for crops can be very useful for improving agricultural remote sensing analysis and mapping.

Technical Abstract: The establishment and application of a spectral library is critical in the standardization and automation of remote sensing interpretation and mapping. Currently, most spectral libraries are designed to support the classification of land cover types, whereas few are dedicated to agricultural remote sensing monitoring. Here, we gathered spectral observation data on plants in multiple experimental scenarios into a spectral database to investigate methods for crop classification (16 crop species) and status monitoring (tea plant and rice growth). We proposed a set of screening methods for spectral features related to plant classification and status monitoring (band reflectance, vegetation index, spectral differentiation, spectral continuum characteristics) that are based on ISODATA and JM distance. Next, we investigated the performance of different classifiers in the spectral library application, including K-nearest neighbor (KNN), Random Forest (RF), and a genetic algorithm coupling with support vector machine (GA-SVM). The optimal combination of spectral features and the classifier with the highest classification accuracy were selected for both crop classification and status monitoring scenarios. The GA-SVM classifier was found to perform the best, which produced an accuracy of OAA=0.94, Kappa=0.93 for crop classification in complex scenario (coexisted with 76 non-crop plant species); and promising accuracies for tea plant growth monitoring (OAA=0.98, Kappa=0.97) and monitoring of rice growth period (OAA=0.92, Kappa=0.90). Therefore, the establishment of a plant spectral library combined with relevant feature extraction and a classification algorithm effectively supports agricultural monitoring by remote sensing.