Hyperspectral plant sensing for differentiating Glyphosate-resistant and Glyphosate-susceptible Johnsongrass through machine learning

Authors: Huang, Yanbo; ZHAO, XIAHU - Hangzhou Dianzi University; PAN, ZENG - Hangzhou Dianzi University; Reddy, Krishna; ZHANG, JINGCHENG - Hangzhou Dianzi University

Submitted to: Pest Management Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/12/2022
Publication Date: 3/7/2022
Citation: Huang, Y., Zhao, X., Pan, Z., Reddy, K.N., Zhang, J. 2022. Hyperspectral plant sensing for differentiating Glyphosate-resistant and Glyphosate-susceptible Johnsongrass through machine learning. Pest Management Science. 78:2370-2377. https://doi.org/10.1002/ps.6864.
DOI: https://doi.org/10.1002/ps.6864

Interpretive Summary: Johnsongrass is a agronomically troublesome weed in the southern United States. Roundup (glyphosate) has been used to kill it in the crop fields for years but the weed has been evolved resistance to the herbicide at the same time. The scientists of USDA ARS and Hangzhou Danzi University collaboratively developed a hyperspectral plant sensing approach to differentiate glyphosate-resistant (GR) and glyphosate-sensitive (GS) plants. This research is a continuation of previous studies to develop the hyperspectral plant sensing approach to differentiate other GR and GS weeds for Palmer amaranth and Italian ryegrass, which integrated advanced machine learning algorithms to enhance the plant differentiation. The results show that the developed hyperspectral plant sensing approach with machine learning algorithms could have an accuracy of 77% in differentiating GR Johnsongrass from GS ones. With good accuracies of previous Palmer amaranth and Italian ryegrass GR and GS plant differentiations it can conclude that the similar hyperspectral approach could be used and transferred across differentiation of these different GR and GS weed species, and it is highly possible for differentiation of more other GR and GS weed species as well

Technical Abstract: BACKGROUND: Johnsongrass (Sorghum halepense) is one of the weeds that evolves resistance to glyphosate [N-(phosphonomethyl)-glycine], the most widely used herbicide, and the weed may cause agronomic troublesome in the southern United States. This paper reports a study on developing a hyperspectral plant sensing approach to explore the spectral features of glyphosate-resistant (GR) and glyphosate-sensitive (GS) plants to evaluate this approach wrapped with machine learning algorithms to differentiate between GR and GS plants. RESULTS: In average GR plants have higher spectral reflectance compared with GS plants. The sensitive spectral bands were optimally selected using the successive projections algorithm respectively wrapped with the machine learning algorithms of k-nearest neighbors (KNN), random forest (RF), and support vector machine (SVM) with Fisher linear discriminant analysis (FLDA) to classify between GS and GS plants. At 3 WAP (weeks after planting) KNN and SVM could not acceptably classify the GR and GS plants but they improved significantly with the stages to have their overall accuracies reached 73% and 77% respectively at 5 WAP. RF and FLDA had a better ability to classify the plants at 3 WAP but RF was low in accuracy at 2 WAP while FLDA dropped the accuracy to 50% at 4 WAP from 57% at 3 WAP and bumped up to 73% at 5 WAP. CONCLUSIONS: Previous studies were conducted developing the hyperspectral imaging approach to differentiation of GR Palmer amaranth from GS Palmer amaranth and GR Italian ryegrass from GS Italian ryegrass with classification accuracies of 90% and 80%, respectively. This study demonstrated that the similar hyperspectral plant sensing approach could be developed to differentiate GR Johnsongrass from GS ones with the highest classification accuracy of 77%, and indicated that the similar hyperspectral approach could be used and transferred from classification of these GR and GS weed species, Palmer amaranth, Italian ryegrass and Johnsongrass, and highly possible for classification of other GR and GS weed species as well. With classic pattern recognition approaches the process of plant classification can be enhanced by modeling using machine learning algorithms.