Skip to main content
ARS Home » Southeast Area » Stoneville, Mississippi » Crop Production Systems Research » Research » Publications at this Location » Publication #335059

Research Project: Application Technologies to Improve the Effectiveness of Chemical and Biological Crop Protection Materials

Location: Crop Production Systems Research

Title: Noise-resistant spectral features for retrieving foliar chemical parameters

item ZHANG, JINGCHENG - Hangzhou Dianzi University
item Huang, Yanbo
item LIU, PENG - Hangzhou Dianzi University
item YUAN, LIN - Zhejiang University
item WU, KAIHUA - Hangzhou Dianzi University

Submitted to: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/28/2017
Publication Date: 12/1/2017
Citation: Zhang, J., Huang, Y., Liu, P., Yuan, L., Wu, K. 2017. Noise-resistant spectral features for retrieving foliar chemical parameters. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 10(12):5368-5380.

Interpretive Summary: To understand plant growing status the chemical components of the leaves are important to be monitored. Scientists at Hangzhou Dianzi University, Hangzhou, China, USDA-ARS Crop Production Systems Research Unit, Stoneville, Mississippi; and Zhejiang University of Water Resources and Electric Power, Zhejiang, China collaboratively developed a hyperspectral data measuring and analysis method to indicate and interpret the chemical components of the plant leaves, including chlorophyll, carotenoids, and leaf water content. In processing and evaluating the data the noise resistant capability of the data is critical for information extraction and interpretation. Wavelet analysis was used in comparison to conventional vegetation indices in processing and analysis of the data. The results indicated that the information extracted from wavelet analysis and vegetation indices exhibited strong resistance to spectral noise and the wavelet models achieved the best in analyzing and interpreting these leaf chemical parameters. This study illustrates that the leaf chemical components could be analyzed and interpreted well through wavelet analysis of the measured leaf hyperspectral data with strong noise resistance.

Technical Abstract: Foliar chemical constituents are important indicators for understanding vegetation growing status and ecosystem functionality. Provided the noncontact and nondestructive traits, the hyperspectral analysis is a superior and efficient method for deriving these parameters. In practical implementation of hyperspectral sensing, some extent of noise is a common issue which might influence the performance of the retrieving system. Therefore, the noise resistant capability is critical for spectral features and retrieving models. In this study, by introducing varying levels of noise to spectral signals, a systematic assessment on noise resistant capability of the retrieving models was conducted, to unveil the real-world performance of hyperspectral methods in retrieving the concentrations of chlorophyll (Chl), carotenoids (Car), and leaf water content (LWC). Two forms of spectral features were tested, including 14 wavelet features that were derived from the continuous wavelet analysis, and 13 selected classic vegetation indices. Two datasets were used for analysis and modeling, including a LOPEX dataset (n=330) for retrieving LWC, and a CORNSPEC dataset (n=213) for retrieving Chl and Car. The results suggested that the wavelet features (WFs) behaved stronger response to all leaf chemical parameters than the vegetation indices as conventional features (CFs). According to an evaluation by decay rate of retrieving error among different noise levels, both WFs and CFs exhibited strong resistance to spectral noise. Particularly for WFs, the noise resistant capability is relevant to the scale of the features. Based on the identified spectral features, both univariate and multivariate retrieving models were established and achieved satisfactory accuracies. Synthesized the retrieving accuracy, noise resistivity, and model’s complexity, the optimal univariate WF-models were recommended in practice for retrieving leaf chemical parameters.