Abstract
Hyperspectral sensor systems play a key role in the automation of work processes in the farming industry. Non-invasive measurements of plants allow for an assessment of the vitality and health state and can also be used to classify weeds or infected parts of a plant. However, one major downside of hyperspectral cameras is that they are not very cost-effective. In this paper, we show, that for specific tasks, multispectral systems with only a fraction of the wavelength bands and costs of a hyperspectral system can lead to promising results for regression and classification tasks. We conclude that for the ongoing automation efforts in the context of cognitive agriculture reduced multispectral systems are a viable alternative.
Zusammenfassung
Hyperspektrale Sensorsysteme spielen eine Schlüsselrolle bei der Automatisierung von Arbeitsprozessen in der Landwirtschaft. Nicht-invasive Messungen von Pflanzen ermöglichen eine Beurteilung des Vitalitäts- und Gesundheitszustands und können auch zur Klassifizierung von Unkraut oder infizierten Pflanzenteilen verwendet werden. Ein großer Nachteil von Hyperspektralkameras ist jedoch, dass sie nicht sehr kosteneffektiv sind. In diesem Beitrag zeigen wir, dass für bestimmte Aufgaben multispektrale Systeme mit nur einem Bruchteil der Wellenlängenbänder und Kosten eines Hyperspektralsystems zu vielversprechenden Ergebnissen bei Regressions- und Klassifikationsaufgaben führen können. Wir kommen zu dem Schluss, dass für die laufenden Automatisierungsbemühungen im Rahmen der kognitiven Landwirtschaft reduzierte multispektrale Systeme eine praktikable Alternative sind.
About the authors
Florian Becker studied cognitive and computer science at Eberhard Karls University of Tübingen and Karlsruhe Institute of Technology (KIT). In 2018, he joined the Vision and Fusion Laboratory at KIT. His research activities are in close cooperation with the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB. In his research, he focuses on machine learning for hyperspectral image analysis.
Dr. Andreas Backhaus is a senior research engineer and has been associated with the Fraunhofer Institute for Factory Operation and Automation IFF in Magdeburg (Germany) since 2009. He holds a diploma degree in Technical Computer Science from the University of Ilmenau (Germany) and a Ph.D. in Computational Neuroscience from the University of Birmingham (United Kingdom). He has specialized on the use of machine learning techniques for sensor data analysis and fusion with an expertise in the acquisition and processing of hyperspectral sensor data. He is currently using his expertise in application areas like smart farming, plant breeding or food quality control. He has been leading or participating a numerous research projects funded by industry as well as the public sector.
Felix Johrden is a research engineer and has been associated with Fraunhofer Institute for Factory Operation and Automation IFF in Magdeburg (Germany) since 2018. He holds a bachelor and master degree in information technology with specialization in image processing and machine learning from the University of Magdeburg.
Merle Flitter is a mechatronics and research engineer at the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB in Karlsruhe, Germany. She holds a bachelor degree in engineering from the Technical University of Berlin and a masters degree in mechatronics from the Karlsruhe Institute of Technology.
Acknowledgment
We like to acknowledge the group of Prof. Klaus Pillen from the Plant Breeding department at the University of Halle-Wittenberg, Germany, and the group of Prof. Reinhard Töpfer at the Julius Kühn Institute, Siebeldingen, Germany, who were performing the field trials and measurements in earlier research projects in partnership with the Fraunhofer IFF, that generated the application datasets used in this study.
References
1. Andreas Backhaus, Praveen C. Ashok, Bavishna B. Praveen, Kishan Dholakia and Udo Seiffert. Classifying scotch whisky from near-infrared Raman spectra with a radial basis function network with relevance learning. In ESANN 2012 proceedings, 20th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, pages 411–416, 2012.Search in Google Scholar
2. G. Cybenko. Approximation by superpositions of a sigmoidal function. Mathematics of Control, Signals, and Systems (MCSS), 2(4):303–314, December 1989.10.1007/BF02551274Search in Google Scholar
3. Reza Dehghani and Nezam Mahdavi-Amiri. Scaled nonlinear conjugate gradient methods for nonlinear least squares problems. Numerical Algorithms, pages 1–20, 2018.10.1007/s11075-018-0591-2Search in Google Scholar
4. Ronald Aylmer Fisher. The use of multiple measurements in taxonomic problems. Annals of Eugenics, 7(2):179–188, 1936.10.1111/j.1469-1809.1936.tb02137.xSearch in Google Scholar
5. Baofeng Guo, Steve R Gunn, Robert I Damper and James DB Nelson. Band selection for hyperspectral image classification using mutual information. IEEE Geoscience and Remote Sensing Letters, 3(4):522–526, 2006.10.1109/LGRS.2006.878240Search in Google Scholar
6. Qing-Juan Han, Hai-Long Wu, Chen-Bo Cai, Lu Xu and Ru-Qin Yu. An ensemble of Monte Carlo uninformative variable elimination for wavelength selection. Analytica Chimica Acta, 612(2):121–125, 2008.10.1016/j.aca.2008.02.032Search in Google Scholar PubMed
7. Paul Herzig, Andreas Backhaus, Udo Seiffert, Nicolaus von Wirén, Klaus Pillen and Andreas Maurer. Genetic dissection of grain elements predicted by hyperspectral imaging associated with yield-related traits in a wild barley nam population. Plant Science, 285:151–164, 2019.10.1016/j.plantsci.2019.05.008Search in Google Scholar PubMed
8. Petr Kadlec, Bogdan Gabrys and Sibylle Strandt. Data-driven soft sensors in the process industry. Computers & Chemical Engineering, 33(4):795–814, 2009.10.1016/j.compchemeng.2008.12.012Search in Google Scholar
9. Anna Kicherer, Katja Herzog, Nele Bendel, Hans-Christian Klück, Andreas Backhaus, Markus Wieland, Johann Rose, Lasse Klingbeil, Thomas Läbe, Christian Hohl, et al. Phenoliner: A new field phenotyping platform for grapevine research. Sensors, 17(7):1625, Jul 2017.10.3390/s17071625Search in Google Scholar PubMed PubMed Central
10. Changhong Liu, Wei Liu, Wei Chen, Jianbo Yang and Lei Zheng. Feasibility in multispectral imaging for predicting the content of bioactive compounds in intact tomato fruit. Food Chemistry, 173:482–488, 2015.10.1016/j.foodchem.2014.10.052Search in Google Scholar PubMed
11. Dan Liu, Da-Wen Sun and Xin-An Zeng. Recent advances in wavelength selection techniques for hyperspectral image processing in the food industry. Food and Bioprocess Technology, 7(2):307–323, 2014.10.1007/s11947-013-1193-6Search in Google Scholar
12. T. M. Martinetz and K. J. Schulten. A Neural-Gas network learns topologies. In Artificial Neural Networks, pages 397–402. North-Holland, Amsterdam, 1991.Search in Google Scholar
13. John Moody and Christian J. Darken. Fast learning in networks of locally tuned processing units. Neural Computation, 1:281–294, 1989.10.1162/neco.1989.1.2.281Search in Google Scholar
14. Martin Fodslette Møller. A scaled conjugate gradient algorithm for fast supervised learning. Neural Networks, 6(4):525–533, 1993.10.1016/S0893-6080(05)80056-5Search in Google Scholar
15. Masateru Nagata, Jasper G Tallada and Taiichi Kobayashi. Bruise detection using NIR hyperspectral imaging for strawberry (fragaria × ananassa duch.). Environmental Control in Biology, 44(2):133–142, 2006.10.2525/ecb.44.133Search in Google Scholar
16. Giuseppe Palermo, Paolo Piraino and Hans-Dieter Zucht. Performance of PLS regression coefficients in selecting variables for each response of a multivariate pls for omics-type data. Advances and applications in bioinformatics and chemistry: AABC, 2:57, 2009.10.2147/AABC.S3619Search in Google Scholar
17. Gerrit Polder, Gerie WAM van der Heijden and Ian T Young. Spectral image analysis for measuring ripeness of tomatoes. Transactions of the ASAE, 45(4):1155, 2002.10.13031/2013.9924Search in Google Scholar
18. Clifford H Spiegelman, Michael J McShane, Marcel J Goetz, Massoud Motamedi, Qin Li Yue and Gerard L Coté. Theoretical justification of wavelength selection in PLS calibration: development of a new algorithm. Analytical Chemistry, 70(1):35–44, 1998.10.1021/ac9705733Search in Google Scholar
19. Svante Wold, Michael Sjöström and Lennart Eriksson. PLS-regression: a basic tool of chemometrics. Chemometrics and Intelligent Laboratory Systems, 58(2):109–130, 2001.10.1016/S0169-7439(01)00155-1Search in Google Scholar
20. Svante Wold, Michael Sjöström and Lennart Eriksson. PLS-regression: a basic tool of chemometrics. Chemometrics and Intelligent Laboratory Systems, 58(2):109–130, 2001. PLS Methods.10.1016/S0169-7439(01)00155-1Search in Google Scholar
© 2021 Walter de Gruyter GmbH, Berlin/Boston
