Rapid and non-destructive seed viability prediction using near-infrared hyperspectral imaging coupled with a deep learning approach

https://doi.org/10.1016/j.compag.2020.105683Get rights and content

Highlights:

  • NIR-HSI can cognize chemical component differences and its spatial distribution.

  • Enhancing the interpretability of NIR spectral images via PCA and SVM analysis.

  • CNN deep learning was utilized to “cognize” viable and non-viable seeds.

  • Approximately 90% viability prediction accuracy for naturally aged seeds.

  • The quality of individual seeds could be assessed nondestructively and rapidly.

Abstract

Seeds are the basis of the agricultural food industry, greater insights into seed viability before sowing could improve storage management and field performance. In the present study, we aimed to address this issue by using highly cost-efficient near-infrared hyperspectral imaging (NIR-HSI) and a convolutional neural network (CNN) deep learning approach. An NIR-HSI camera was used because it can recognize both molecular vibration information (i.e. chemical component differences) and its spatial distribution in each seed sample; this camera is much more informative than a regular RGB digital camera. Using this technology, the emphasis of this study was firstly to provide a methodology for enhancing the interpretability of viable and non-viable seeds via principal component analysis (PCA) and support vector machine (SVM) viability classification analysis of NIR-HSI data. A CNN was then constructed to“cognize” the differences in viable and non-inviable seeds and classify them automatically. Experimental results indicate that the methodology produces a ~90% classification accuracy for both a five-fold cross-validation set and a test set of naturally aged Japanese mustard spinach seeds. Therefore, this study provides a new strategy for effective and practical seed viability prediction.

Section snippets

Introduction:

Along with air and water, food is a basic requirement for human survival. However, food shortages become more likely as agricultural land is removed due to urbanization (Sahdev et al., 2017) and world’s population increases (Esper and Miihlbauer, 1998). Undoubtedly, seeds are the basis of the agricultural food industry (Huang et al., 2016); thus, investigating seed viability could yield useful information on how to predict the correct time for regeneration of accession and how to grow seedlings

Sample preparation

In total, 1200 Japanese mustard spinach (Brassica rapa var. perviridis) seeds were prepared, and the seed sample size was approximately 1.0–1.5 mm in diameter. Prior to experimentation, the seeds were stored at room temperature in an open package until approximately one year after their expiry date. These harsh storage conditions were expected to increase the negative effects on seed viability. It should be noted that no other artificially pretreatments (e.g. heating) were carried out on the

NIR spectral data of viable and non-viable seeds

Fig. 2 shows the averaged NIR reflectance spectra with their stand deviations, separately collected from 646 viable seeds and 554 non-viable seeds. The light absorptions around 1200 nm and 1780 nm are related to Csingle bondH second and first overtone stretching due to absorption by the CH3 functional group, respectively. Water content differences could also be considered as an important factor in seed viability (Mukasa et al., 2019). The wavelengths around 1450 nm and 1930 nm correspond to Osingle bondH stretching

Conclusion

The experimental results from this study demonstrate that NIR-HSI combined with a CNN approach, is an effective method for classifying seed viability. In naturally-aged Japanese mustard spinach seeds, the seed viability classification accuracies for the training set (five-fold cross-validation) and the test set were approximately 90% and 83%, respectively. The test accuracy could be improved to 90% by using the averaged prediction results of each sample but at different surfaces. The next steps

CRediT authorship contribution statement

Te Ma: Data curation, Validation, Writing - original draft. Satoru Tsuchikawa: Supervision, Writing - review & editing. Tetsuya Inagaki: Conceptualization, Methodology.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

The authors gratefully thank the financial support provided by JSPS (KAKENHI, No. 19H03015 and 19K15886).

References (38)

  • C. Wakholi et al.

    Rapid assessment of corn seed viability using short wave infrared line-scan hyperspectral imaging and chemometrics

    Sens. Actuators, B Chem.

    (2018)
  • Y. Xia et al.

    Recent advances in emerging techniques for non-destructive detection of seed viability: A review

    Artif. Intell. Agric.

    (2019)
  • Q. Zhang et al.

    Recent advances in convolutional neural network acceleration

    Neurocomputing

    (2019)
  • Z. Zuo et al.

    Exemplar based deep discriminative and shareable feature learning for scene image classification

    Pattern Recognit.

    (2015)
  • O. Abdel-Hamid et al.

    Applying convolutional neural networks concepts to hybrid NN-HMM model for speech recognition

  • C. Cortes et al.

    Support-Vector Networks

    Mach. Learn.

    (1995)
  • Y.N. Dauphin et al.

    Language modeling with gated convolutional networks

  • J.C. Delouche et al.

    Accelerated aging techniques for predicting the relative storability of seed lots 1

    Seed Sci. Technol.

    (1973)
  • C. Gao et al.

    Enhancement of ELDA tracker based on CNN features and adaptive model update

    Sensors (Switzerland)

    (2016)
  • Cited by (58)

    View all citing articles on Scopus
    View full text