A non-destructive testing method for early detection of ginseng root diseases using machine learning technologies based on leaf hyperspectral reflectance

Guiping Zhao, Yifei Pei, Ruoqi Yang, Xiang Li, Zihan Fang, Ye Wang, Yin Dou, Jie Wu, Dan Gao, Dade Yu, Xiwen LiFrontiers in Plant Science 2022 · DOI

TL;DR

This paper introduces a novel non-destructive method for early detection of ginseng root diseases using leaf hyperspectral reflectance and machine learning. By analyzing spectral data from leaves that appear healthy, the researchers achieved high accuracy in identifying diseased roots, offering a crucial tool for preventing yield loss and reducing pesticide use.

What Problem Does This Paper Address?

Ginseng is a highly valued medicinal plant, but its yield is significantly threatened by root diseases. The challenge lies in the fact that visible symptoms of root disease often appear late, only after the infection has progressed significantly and potentially spread. Traditional detection methods rely on manual observation, which is time-consuming, labor-intensive, and often inaccurate, especially in the early stages. This delay means that by the time symptoms are visible, it's often too late for effective intervention, leading to substantial crop loss. The paper aims to bridge this gap by developing an efficient, non-destructive testing (NDT) method for early detection, allowing for timely treatment and prevention.

Methods

The core of this research lies in combining hyperspectral remote sensing with machine learning.

  • Hyperspectral Reflectance: The study collected hyperspectral reflectance data from ginseng leaves across a broad spectrum (325-1,075 nm). Crucially, this was done on plants that visually appeared healthy, targeting the early detection of root issues before they manifest above ground.
  • Machine Learning Classification: Several machine learning algorithms were employed to build classification models: Random Forest (RF), Extremely Randomized Trees (ET), Adaptive Boosting (ADA), and Gradient Boosting Decision Tree (GBDT). These models were trained to distinguish between healthy and diseased ginseng based on the spectral data.
  • Data Preprocessing: To enhance the spectral data and improve model performance, various preprocessing techniques were applied, including Savitzky-Golay (SG) filtering, standard normal variate (SNV), and multiple scattering correction (MSC). The SG method was identified as the most effective for this dataset.
  • Feature Selection: The study analyzed the importance of different spectral bands to identify specific regions most indicative of root disease. This involved analyzing the contribution of various bands to the classification accuracy of the ML models.
  • Model Optimization and Validation: Hyperparameter tuning was performed to optimize the chosen ML models. The best-performing model was then validated on an independent set of ginseng plants to assess its real-world applicability.

Datasets

The study utilized a dataset of 217 ginseng plants cultivated in Jingyu County, Jilin Province, China. These plants were grown for 3-5 years.

  • Training/Testing Data: The 217 plants were divided into training (3:1 ratio) and testing sets.
  • Validation Data: An additional 12 ginseng plants from a different plot were collected for model validation.

The dataset was categorized into "healthy" and "diseased" groups based on visual inspection of the roots after harvest.

Key Insights and Ideas

  • Leaf Spectra as a Proxy for Root Health: The most significant insight is that even when ginseng leaves show no visible symptoms, their hyperspectral reflectance can reveal underlying root diseases. This challenges the assumption that above-ground symptoms are the sole indicators of plant health.
  • The "Red Edge" is Key: The analysis pinpointed a specific spectral range (690-726 nm) as particularly important for detecting ginseng root diseases. This region, often referred to as the "red edge," is known to be sensitive to changes in plant structure and physiology.
  • Machine Learning's Power in Early Detection: The study demonstrates the effectiveness of ensemble learning algorithms (particularly ET) in processing complex hyperspectral data for early disease detection, achieving high accuracy where traditional methods fail.
  • Physiological Traits are Not Always Reliable Early Indicators: Contrary to expectations, basic physiological measurements like chlorophyll content (SPAD) and antioxidant enzyme activities (SOD, POD, CAT) did not show significant differences between healthy and diseased plants in the early stages, highlighting the need for more sensitive detection methods.

Key Results

  • High Classification Accuracy: The Extremely Randomized Trees (ET) model, after preprocessing with the Savitzky-Golay method, achieved an impressive accuracy of 98% for identifying diseased ginseng and 99% for healthy ginseng.
  • Optimal Spectral Band: The 690-726 nm range was identified as the most critical spectral band for early detection of ginseng root diseases.
  • Model Performance: The ET model outperformed other tested algorithms (RF, ADA, GBDT) in terms of accuracy, precision, recall, and F1-score, with an F1-score of 0.99 and recall of 1.0 for diseased plants.
  • Model Validation: The ET model demonstrated a practical accuracy of 83.3% when validated on a small, independent set of 12 ginseng plants.

Limitations and Future Work

  • Sample Size for Validation: The authors acknowledge that the validation set of 12 plants was limited, suggesting that a larger dataset would be beneficial for more robust validation.
  • Environmental Factors: Hyperspectral reflectance can be influenced by various factors beyond disease, such as ginseng age, developmental stage, disease severity, and environmental conditions. Future work could incorporate these variables into more comprehensive models.
  • Mechanism Understanding: While the study successfully detects diseases, a deeper understanding of the underlying physiological and biochemical mechanisms linking leaf spectral changes to root disease is needed. This could be achieved by combining hyperspectral analysis with pathological studies.
  • Database Development: The researchers envision building a comprehensive hyperspectral reflectance database for ginseng root diseases. This database, coupled with advanced ML algorithms, could enable more precise identification of disease types, onset, and severity.
  • Technological Advancement: The paper suggests future development of integrated algorithms and portable detection equipment, potentially incorporating intelligent robots for smart agriculture applications.

Summary updated

Abstract

Ginseng is an important medicinal plant benefiting human health for thousands of years. Root disease is the main cause of ginseng yield loss. It is difficult to detect ginseng root disease by manual observation on the changes of leaves, as it takes a long time until symptoms appear on leaves after the infection on roots. In order to detect root diseases at early stages and limit their further spread, an efficient and non-destructive testing (NDT) method is urgently needed. Hyperspectral remote sensing technology was performed in this study to discern whether ginseng roots were diseased. Hyperspectral reflectance of leaves at 325-1,075 nm were collected from the ginsengs with no symptoms on leaves at visual. These spectra were divided into healthy and diseased groups according to the symptoms on roots after harvest. The hyperspectral data were used to construct machine learning classification models including random forest, extreme random tree (ET), adaptive boosting and gradient boosting decision tree respectively to identify diseased ginsengs, while calculating the vegetation indices and analyzing the region of specific spectral bands. The precision rates of the ET model preprocessed by savitzky golay method for the identification of healthy and diseased ginsengs reached 99% and 98%, respectively. Combined with the preliminary analysis of band importance, vegetation indices and physiological characteristics, 690-726 nm was screened out as a specific band for early detection of ginseng root diseases. Therefore, underground root diseases can be effectively detected at an early stage by leaf hyperspectral reflectance. The NDT method for early detection of ginsengs root diseases is proposed in this study. The method is helpful in the prevention and control of root diseases of ginsengs to prevent the reduction of ginseng yield.
BibTeX
@inproceedings{zhao2022non,
  author = {Guiping Zhao and Yifei Pei and Ruoqi Yang and Xiang Li and Zihan Fang and Ye Wang and Yin Dou and Jie Wu and Dan Gao and Dade Yu and Xiwen Li},
  title = {{A non-destructive testing method for early detection of ginseng root diseases using machine learning technologies based on leaf hyperspectral reflectance}},
  booktitle = {Frontiers in Plant Science},
  year = {2022},
  doi = {10.3389/fpls.2022.1031030},
}
Figure