Rapid and non-destructive seed viability prediction using near-infrared hyperspectral imaging coupled with a deep learning approach
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 CH 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 OH 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)
- et al.
X-ray CT image analysis for morphology of muskmelon seed in relation to germination
Biosyst. Eng.
(2018) - et al.
High speed measurement of corn seed viability using hyperspectral imaging
Infrared Phys. Technol.
(2016) - et al.
Comparative nondestructive measurement of corn seed viability using Fourier transform near-infrared (FT-NIR) and Raman spectroscopy
Sens. Actuators, B Chem.
(2016) - et al.
CNNTracker: Online discriminative object tracking via deep convolutional neural network
Appl. Soft Comput. J.
(2016) - et al.
Solar drying - an effective means of food preservation
Renew. Energy
(1998) - et al.
Non-destructive evaluation of watermelon seeds germination by using Delayed Luminescence
J. Photochem. Photobiol. B Biol.
(2018) - et al.
Classification of maize seeds of different years based on hyperspectral imaging and model updating
Comput. Electron. Agric.
(2016) - et al.
Near-infrared hyperspectral imaging system coupled with multivariate methods to predict viability and vigor in muskmelon seeds
Sens. Actuators, B Chem.
(2016) - et al.
Noncontact evaluation of soluble solids content in apples by near-infrared hyperspectral imaging
J. Food Eng.
(2018) - et al.
Determination of viability of Retinispora (Hinoki cypress) seeds using FT-NIR spectroscopy
Infrared Phys. Technol.
(2019)
Rapid assessment of corn seed viability using short wave infrared line-scan hyperspectral imaging and chemometrics
Sens. Actuators, B Chem.
Recent advances in emerging techniques for non-destructive detection of seed viability: A review
Artif. Intell. Agric.
Recent advances in convolutional neural network acceleration
Neurocomputing
Exemplar based deep discriminative and shareable feature learning for scene image classification
Pattern Recognit.
Applying convolutional neural networks concepts to hybrid NN-HMM model for speech recognition
Support-Vector Networks
Mach. Learn.
Language modeling with gated convolutional networks
Accelerated aging techniques for predicting the relative storability of seed lots 1
Seed Sci. Technol.
Enhancement of ELDA tracker based on CNN features and adaptive model update
Sensors (Switzerland)
Cited by (58)
Rapid and accurate identification of bakanae pathogens carried by rice seeds based on hyperspectral imaging and deep transfer learning
2024, Spectrochimica Acta - Part A: Molecular and Biomolecular SpectroscopyDetection of Insect-Damaged Maize Seed Using Hyperspectral Imaging and Hybrid 1D-CNN-BiLSTM Model
2024, Infrared Physics and TechnologyA research review on deep learning combined with hyperspectral Imaging in multiscale agricultural sensing
2024, Computers and Electronics in AgricultureDetection of fungal infection in apple using hyperspectral transformation of RGB images with kernel regression
2023, Postharvest Biology and TechnologyA review of deep learning techniques used in agriculture
2023, Ecological Informatics