Hyperspectral reflectance sensing for quantifying leaf chlorophyll content in wasabi leaves using spectral pre-processing techniques and machine learning algorithms

摘要

Changes in chlorophyll content can be a good indicator of disease as well as nutritional and environmental stresses on plants. Several pre-processing techniques have been proposed for reducing noise from spectral data to identify vegetation properties such as chlorophyll content. Machine learning algorithms have also been applied to assess biochemical properties; however, an approach integrating pre-processing techniques and machine learning algorithms has not been fully evaluated. Therefore, this study evaluates the effectiveness of five pre-processing techniques used in conjunction with five machine learning algorithms for estimating chlorophyll content in two wasabi cultivars. Overall, incorporating pre-processing techniques was effective for obtaining estimated values with high accuracy. Analyses utilizing both pre-processing and machine learning performed best in 88 of 100 repetitions. The kernel-based extreme learning machine (KELM) and Cubist algorithms yielded the highest performance and achieved the highest accuracies in 54 and 26 of 100 repetitions, respectively.