Selection of a Spectral Index for Detection of Orange Spotting Disease in Oil Palm (Elaeis guineensis Jacq.) Using Red Edge and Neural Network Techniques

Abstract

Spectral screening can play an important role in successful detection of viroid-infected oil palm seedlings from nursery stage prior to transplanting into the field. Coconut cadangcadang viroid (CCCVd) is the main causal agent of orange spotting (OS) disease. OS disease is an emerging disease in Malaysian plantation. In this study, a glasshouse experiment was conducted with fifteen CCCVd-inoculated and five healthy oil palm seedlings in the growing season of 2015. Spectral screening was performed using a hyperspectral spectroradiometer, Analytic Spectral Device HandHeld 2 (325-1075 nm).The red edge, a steep gradient in reflectance between red and near-infrared bands (680-780 nm), was used for selection of red edge bands. A maximum point (i.e., 700 nm) and minimum point (i.e., 768 nm) of red edge were selected from healthy and inoculated spectra. Shifts of red edge inflection point from healthy to inoculated spectra were also studied. Four well-known spectral indices, namely simple ratio, red edge normalized difference vegetation index, two-band enhanced vegetation index 2 (EVI2), and chlorophyll index red edge, were evaluated using selected red edge bands. The multilayer perceptron neural network model was used to establish a nonlinear relationship between selected spectral bands and each spectral index. EVI2 was selected as a best spectral index which resulted in zero errors at the training, testing, and validation datasets. The highest coefficient of correlation (r= 1) was recorded between spectral bands (input values) andEVI2 (target values).

Current State of Hyperspectral Remote Sensing for Early Plant Disease Detection: A Review

“…The reviewed works prove the possibility of detecting oil palm [36,[50][51][52][53][54][55][56][57][63][64][65][66], citrus [73][74][75][76][77][78], Solanaceae family crops [91][92][93][94][95][96][97][98][99][100][101][102][103] and wheat [24,[124][125][126][127][128][129][130][131][132][140][141][142][143][144][145] diseases using HRS.…”

“…On the subject of OS, the authors conducted and published a number of studies on SVI and ANN choice for disease determination, as well as chlorophyll content at the leaf scale of the diseased plants. The hyperspectral data of OS diseased and healthy oil palm seedlings was processed by five different ANNs for evaluation of four red-edge indices followed by the selection of spectral bands from the red edge (680-780 nm), with a result that a red-edge inflection point (at 700 nm) could serve as a good indicator of the plant stress caused by OS in oil palm seedlings [65,66]. A systematization of the reviewed materials is present in Table 1.…”

Problems and Prospects of Improving Abiotic Stress Tolerance and Pathogen Resistance of Oil Palm

“…By analyzing the reflectance spectra of oil palm seedlings and healthy oil palm seedlings inoculated with coconut cadang-cadang viroid (CCCVd), it was shown that the infrared inflection point (ReIP) at 700 nm could well reflect the condition of oil palm seedlings under CCCVd stress. The dual-band enhanced vegetation index (EVI2) is the best spectral index for detecting orange spot disease in mature oil palm forests [84]. Zhang et al [85] used a loop-mediated isothermal amplification technique to develop a rapid and sensitive detection method for oil palm damping-off bacteria and designed a total of 6 primers based on 8 loci in the ITS region gene.…”

Proximal Methods for Plant Stress Detection Using Optical Sensors and Machine Learning

“…ANNs have been used successfully in many studies for the identification and classification of various plant stresses. These include detecting powdery mildew and soft rot in zucchini [129], classifying biotic stresses in pomegranate [106], detecting orange spotting disease in oil palm [130], and identifying crown rot in wheat [131]. A major advantage of ANNs is their ability to be used without specialized knowledge on the data and its interpretation; however, disadvantages include being prone to overfitting and requiring greater amounts of computational resources [132].…”