Deep Learning for Multispectral Imaging based Predictions in Agriculture
Article Sidebar
Main Article Content
Abstract
Introduction. Artificial Intelligence has achieved immense success in recent years, and although commercially profitable applications currently compete with humans in terms of accuracy and efficiency, there are other areas that could benefit from these technologies and in which obstacles still exist. An important aspect of this paper is that these results allow us to better understand the limitations related to the use of uncommon data in AI models. This may enable the development of tools to implement smaller, faster, and more efficient models with applications in agriculture, and other areas that use multispectral images. Objective. This paper proposes a scheme in which data from non-conventional sources related to agriculture are analyzed by custom AI models to generate predictions about variables measured in the fields, and that can eventually help the understanding of the underlying physical and biological phenomena. Methodology. This work summarizes the results obtained throughout the implementation of a project that has used multi- and hyperspectral image data of agricultural crops, as well as information taken in the field. The datasets include multispectral images of wheat crops, and hyperspectral images of canola and wheat, and includes manual measurements of certain variables. When it comes to AI models, these are closely related to addressing the problem of data processing. In both cases, simple deep learning models have been chosen, but with differences in the type of data that they are optimized to process. Results. The main result of this work is the demonstration of the use of AI/DL models for unconventional data analysis. In the first case, using 3D convolutional networks, we have managed to implement a model that can predict the yield of the wheat crops under analysis; and in the second case, using a dual scheme, with sequential and spatial models, we have managed to predict the moisture content. Conclusion. Primarily, this work demonstrates that a DL model can find useful features within an MSI dataset for yield prediction; in addition to finding an accurate DL model for the prediction of moisture content of canola and wheat crops, based on HSI. These results demonstrate the versatility of ML models and the possibility of extending the results obtained in other applications.
Downloads
Article Details
References
Chollet, F. (2017). Deep Learning with Python. https://www.manning.com/books/deep-learning-with-python
Dumoulin, V., & Visin, F. (2018). A guide to convolution arithmetic for Deep Learning. arXiv. https://doi.org/10.48550/arXiv.1603.07285
Han, S., Mao, H., & Dally, W. J. (2015, October 1). Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding. arXiv.Org. https://arxiv.org/abs/1510.00149v5
Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735
Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Proceedings of the 31st International Conference on Neural Information Processing Systems, 4768–4777.
Manolakis, D. G., Lockwood, R. B., & Cooley, T. W. (2016). Hyperspectral imaging remote sensing: Physics, sensors, and algorithms. Cambridge University Press.
Montesinos-López, O. A., Montesinos-López, A., Crossa, J., De los Campos, G., Alvarado, G., Suchismita, M., Rutkoski, J., González-Pérez, L., & Burgueño, J. (2017). Predicting grain yield using canopy hyperspectral reflectance in wheat breeding data. Plant Methods, 13(1), 4. https://doi.org/10.1186/s13007-016-0154-2
Torres-Tello, J., Singh, K., Ko, S.-B., & Shirtliffe, S., (2020). Transfer learning from RGB to hyperspectral images by means of pointwise convolutions. 5th Annual P2IRC Symposium. https://doi:10.13140/RG.2.2.31919.5648054
Torres-Tello, J., & Ko, S.-B. (2021a). Identifying useful features in multispectral images with deep learning for optimizing wheat yield prediction. 2021 IEEE International Symposium on Circuits and Systems (ISCAS), 1–5. https://doi.org/10.1109/ISCAS51556.2021.9401360
Torres-Tello, J. W., & Ko, S. (2021b). A novel approach to identify the spectral bands that predict moisture content in canola and wheat. Biosystems Engineering, 210, 91–103. https://doi.org/10.1016/j.biosystemseng.2021.08.004
Torres-Tello, J., & Ko, S.-B. (2022). Optimizing a multispectral-images-based dl model, through feature selection, pruning and quantization. 2022 IEEE International Symposium on Circuits and Systems (ISCAS), 1352–1356. https://doi.org/10.1109/ISCAS48785.2022.9937940
Torres Tello, J. W., (2022). Optimization of AI models as the main component in prospective edge intelligence applications [Thesis, University of Saskatchewan]. https://harvest.usask.ca/handle/10388/14092
Virnodkar, S. S., Pachghare, V. K., Patil, V. C., & Jha, S. K. (2020). Remote Sensing and Machine Learning for crop water stress determination in various crops: A critical review. Precision Agriculture, 21(5), 1121–1155. https://doi.org/10.1007/s11119-020-09711-9
Xiong, Z., Yuan, Y., & Wang, Q. (2018). Ai-net: Attention inception neural networks for hyperspectral image classification. IGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium, 2647–2650. https://doi.org/10.1109/IGARSS.2018.8517365