Spectral response to early detection of stressed oil palm seedlings using near-infrared reflectance spectra at region 900-1000 nm
A method was developed based on spectral analysis and classification models for early detection of water stress level in the leaves of oil palm seedlings. The healthy (well-watered: D0) and water-stressed (subjected to water stress for five days: D1-D5) leaves of oil palm seedlings were investigated...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English English |
| Published: |
Elsevier
2023
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/41791/1/Spectral%20response%20to%20early%20detection%20of%20stressed%20oil%20palm%20seedlings_ABST.pdf http://umpir.ump.edu.my/id/eprint/41791/2/Spectral%20response%20to%20early%20detection%20of%20stressed%20oil%20palm%20seedlings.pdf |
| Summary: | A method was developed based on spectral analysis and classification models for early detection of water stress level in the leaves of oil palm seedlings. The healthy (well-watered: D0) and water-stressed (subjected to water stress for five days: D1-D5) leaves of oil palm seedlings were investigated to identify and classify the stress levels. The stress levels were grouped as light, moderate, and severe. The region 900–1000 nm was selected because it is highly correlated with water content, particularly in terms of first and second derivatives. The measured reflectance spectra at 900–1000 nm were pre-processed using smoothing, standard normal variate (SNV), and first and second Savitzky-Golay (SG) derivatives. Principal component analysis (PCA) was performed on several transformed datasets to reduce the reflectance spectral dimension and derive the principal components (PCs). Support vector machine (SVM) and linear discriminant analysis (LDA) classification models were employed to the scores of PCs to achieve six classification levels of water stress. Classification accuracy was assessed using the overall accuracy and confusion matrix of testing datasets. The SVM and PCA-LDA classification models predicted the water stress levels with high average overall classification accuracy of 92 % and 94 % using the smoothed + SNV + first derivative and smoothed + SNV spectral dataset, respectively. The findings confirmed the potential of 900–1000 nm region to distinguish the different levels of water stress in oil palm seedlings. |
|---|