Selection of optimal hyperspectral wavebands for detection of discolored, diseased rice seeds

Authors: BAEK, INSUCK - University Of Maryland; Kim, Moon; CHO, BYUNG-KWAN - Chungnam National University; MO, CHANGYEUN - Korean Rural Development Administration; Barnaby, Jinyoung; McClung, Anna; OH, MIRAE - Us Forest Service (FS)

Submitted to: Applied Sciences
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/5/2019
Publication Date: 3/12/2019
Citation: Baek, I., Kim, M.S., Cho, B., Mo, C., Barnaby, J.Y., McClung, A.M., Oh, M. 2019. Selection of optimal hyperspectral wavebands for detection of discolored, diseased rice seeds. Applied Sciences. 9:1027. https://doi.org/10.3390/app9051027.
DOI: https://doi.org/10.3390/app9051027

Interpretive Summary: Bacterial panicle blight (BPM) is a disease found worldwide that greatly damages rice production yields due to rotting or sterilization of the rice plants and their developing grains. Because this bacterial disease is seed-transmitted, detection and removal of infected seeds prior to planting is desirable. This study analyzed visible/near-infrared hyperspectral images of sound and infected rice seeds to select optimal wavelengths of light for BPM detection and to test their use in classification models for discriminating between sound and BPM-infected seeds. The results showed that using two or three optimally selected wavelengths, particularly including selections in red and violet regions of visible light, could effectively classify the seeds at over 90% accuracies. Rice inspection using two or three wavelengths can be implemented via multispectral imaging technologies at much lower cost compared to using more expensive hyperspectral imaging technologies that provide an excessive volume of unnecessary spectral data. The results of this study show that multispectral imaging inspection can be effective and is feasible for BPM detection in rice seeds, which would greatly benefit rice producers in the U.S. and overseas.

Technical Abstract: The inspection of rice grain that may be infected by seedborne disease is important for ensuring uniform plant stands in production fields as well as preventing proliferation of some seedborne diseases. The goal of this study was to use hyperspectral imaging (HSI) technique to find optimal wavelengths and develop a model for detecting discolored, diseased rice seed infected by bacterial panicle blight (Burkholderia glumae), a seedborne pathogen. For this purpose, the HSI data spanning the visible/near-infrared wavelength region between 400 nm and 1000 nm were collected for 500 sound and discolored rice seeds. For selecting optimal wavelengths to use for detecting diseased seed, a sequential forward selection (SFS) method combined with various spectral pretreatments was employed. To evaluate performance based on optimal wavelengths, support vector machine (SVM) and linear discriminant analysis (LDA & QDA) models were developed for detection of discolored seeds. As a result, the violet and red regions of the visible spectrum were selected as key wavelengths reflecting the characteristics of the discolored rice seeds. When using only two or only three selected wavelengths, all of the classification methods achieved high classification accuracies over 90% for both the calibration and validation sample sets. The results of the study showed that only two to three wavelengths are needed to differentiate between discolored, diseased rice and sound rice, instead of using the entire HSI wavelength regions. This demonstrates the feasibility of developing a low cost multispectral imaging technology based on these selected wavelengths for non-destructive and high-throughput screening of diseased rice seed.