Machine Learning Approach to Identify Tomato Bacterial Spot Disease before Symptoms Inclusion Using Hyperspectral Images

Document Type : Original Article

Authors

1 Department of Biosystems Engineering, Faculty of Agriculture, Shiraz University, Shiraz, Iran

2 Department of Plant Protection, Faculty of Agriculture, Shiraz University, Shiraz, Iran

3 Department of Plant Protection, Faculty of Agriculture, Isfahan University of Technology, Isfahn, Iran

Abstract

Introduction
Tomato (Solanum lycopersicum L.) is a self-pollinating plant from the Solanaceae family, which includes over 3,000 economically significant species. It is one of the most consumed vegetables globally and ranks seventh in importance after crops like maize and rice. Native to the coastal plains of South America, tomatoes were domesticated in Mexico and introduced to Iran in the 19th century. Rich in lycopene, beta-carotene, flavonoids, and vitamin C, tomatoes are renowned for their anti-cancer properties. Globally, tomato production for processing is around 41 million tons, with Asia accounting for about 60% of production. In Iran, about 4.9 million tons were produced in the 2017-2018 agricultural year. Tomato plants are hosts to over 200 pests and plant diseases, with bacterial spot disease being one of the most damaging. The bacterial spot disease can persist in seeds and plant residues for up to 16 months. Detection of bacterial spot disease in early stage is crucial for efficient disease management, however traditional methods like field scouting are inefficient and prone to human error. Recent advancements in spectral imaging, such as hyperspectral imaging, offer a non-destructive, efficient way to detect early stage of plant diseases. This research study explores the use of hyperspectral imagery and machine learning algorithms to detect bacterial spot in tomato leaves before symptoms inclusion, improving early intervention and disease management. The objective of this research study aimed on employing three machine learning algorithms to classify healthy plants versus infected ones using spectral reflectance in the range of 400-800 nm. Before employing machine learning algorithms, the preprocessing methods were employed to remove the noise and improve the classifier algorithm 's performance. In order to find the earliest time for identifying bacterial spot disease before symptoms inclusion, the combination of preprocessing methods and classifier algorithms were also evaluated through 7 to 19 days after inoculation.
Materials and Methods
In this study, tomato seedlings were sourced from a farm in Bavanat, Fars province, Iran, and transferred to a greenhouse in the Faculty of Agriculture Shiraz University for controlled growth conditions. The plants were inoculated with Xanthomonas euvesicatoria pv. perforans at the 4-5 leaf stage. The research aimed to detect bacterial spot disease before symptoms inclusion, using hyperspectral imaging. Hyperspectral images of 100 tomato samples were taken using hyperspectral camera (Partov Afzar Sanat, Zanjan, Iran). Spectral reflectance of tomato plants was collected in the range of 400–800 nm before inoculation and over the seventh to the nineteenth day after inoculation bacterial spot disease. Data analysis was conducted using Python, where spectral data was preprocessed to remove noise using four methods, including standard normal variate (SNV), multiplicative scatter correction (MSC), first and second derivatives (FD, SD). Principal Component Analysis (PCA) was applied in order to feature reduction. Three machine learning algorithms including random forest (RF), gradient boosting machine (GBM), and support vector machine (SVM) were employed to classify healthy and infected tomato plants. The model's performance was evaluated based on confusion matrix using accuracy, precision, sensitivity, specificity, and F-measure. Accuracy is defined as the ability of the classifier algorithm to detect the healthy and the infected plants correctly. Sensitivity and specificity are defined as the proportion of the healthy or infected plants correctly classified. In the precision calculation, the number of actual predicted infected is divided by the total number of predicted infected plants that were classified as true or false. F-measure defines the harmonic mean of sensitivity and precision where it reaches its best value at 1.0 (perfect precision and sensitivity) and worst value at 0.0. The classification models were validated using a 70-30 split of training and testing data, and the training process was conducted through ten- fold cross validation to ensure reliable results.
Results and Discussion
The analysis of spectral reflectance of infected and healthy plants revealed that the bacterial spot disease has a significant effect on the spectral signature infected plants. The most changes on the spectral reflectance of infected plants were happened in range of 740 to 800 nm, which is part of the NIR area and is related to changes from the structure of the leaf tissue. The results showed that FD, SNV, MSC preprocessing methods significantly improved the classification accuracy of healthy and infected plants over the seventh to the nineteenth day after inoculation bacterial spot disease. The FD preprocessing on the 7 and 10 days after inoculation resulted in the highest accuracy (98%), while MSC and SD methods performed best after 14 and 19 days. The RF, SVM, and GBM classification algorithms could classify the infected plants versus healthy plants with 98%, 100% and 100% accuracy respectively.  The results indicated that the 7th day after inoculation was the most reliable and earliest time before symptoms inclusion for classifying infected and healthy tomato plants. The highest classification accuracy was achieved with SVM and GBM algorithms and FD preprocessing method on the 7th day (100%), and SVM with MSC on the 19th day (98%).
Conclusion
The results of this research indicated the ability of machine learning algorithms and hyperspectral imagery for classifying healthy plants versus infected ones with bacterial spot disease in tomato plants before symptoms inclusion. Three classification machine learning algorithms including RF, SVM and RGB could classify infected plants on the 7th day after bacterial spot inoculation with more than 97% accuracy. Therefore, spectral reflectance of potato plant leaves in the range of 400 to 900 nm can be a potential way to identify bacterial spot disease in potato plants in early stage of disease for efficient crop management.

Keywords

Main Subjects

References

Abdulridha, J., Ampatzidis, Y., Kakarla, S. C., & Roberts, P. (2020). Detection of target spot and bacterial spot diseases in tomato using UAV-based and benchtop-based hyperspectral imaging techniques. Precision Agriculture, 21(5), 955–978. https://doi.org/10.1007/s11119-019-09703-4.
Al-Saddik, H., Simon, J.-C., & Cointault, F. (2017). Development of spectral disease indices for ‘Flavescence Dorée’grapevine disease identification. Sensors, 17(12), 2772. https://doi.org/10.3390/s17122772.
 
Barbin, D. F., G. ElMasry, D. W. Sun, and P. Allen. (2012). Predicting quality and sensory attributes of pork using near-infrared hyperspectral imaging. Anal. Chim. Acta 719: 30-42. https://doi.org/10.1016/j.aca.2012.01.004.
 
Bergougnoux, V. (2014). The history of tomato: From domestication to biopharming. Biotechnology Advances, 32(1), 170–189. https://doi.org/10.1016/j.biotechadv.2013.11.003.
 
Bock, C. H., Poole, G. H., Parker, P. E., & Gottwald, T. R. (2010). Plant disease severity estimated visually, by digital photography and image analysis, and by hyperspectral imaging. Critical reviews in plant sciences, 29(2), 59-107. https://doi.org/10.1080/07352681003617285.
 
Castillejo-González, I. L., Pena-Barragán, J. M., Jurado-Expósito, M., Mesas-Carrascosa, F. J., & López-Granados, F. (2014). Evaluation of pixel-and object-based approaches for mapping wild oat (Avena sterilis) weed patches in wheat fields using QuickBird imagery for site-specific management. European Journal of Agronomy, 59, 57-66. https://doi.org/10.1016/j.eja.2014.05.009.
 
Cordon, G., Andrade, C., Barbara, L., & Romero, A. M. (2022). Early detection of tomato bacterial canker by reflectance indices. Information Processing in Agriculture, 9(2), 184–194. https://doi.org/10.1016/j.inpa.2021.06.004.
 
FAO. 2015. Statistical database. Advailable on Faostat.Fao.org
 
 Farber, C., Mahnke, M., Sanchez, L., & Kurouski, D. (2019). Advanced spectroscopic techniques for plant disease diagnostics. A review. TrAC Trends in Analytical Chemistry, 118, 43–49. https://doi.org/10.1016/j.trac.2019.05.022.
 
Fu, Xiaping, Yibin Ying, Huishan Lu, Huirong Xu, and Haiyan Yu (2007). FT-NIR diffuse reflectance spectroscopy for kiwifruit firmness detection. Sensing and Instrumentation for Food Quality and Safety, 1, 29-35. https://doi.org/10.1007/s11694-007-9004-2.
 
Gao, Z., Khot, L. R., Naidu, R. A., & Zhang, Q. (2020). Early detection of grapevine leafroll disease in a red-berried wine grape cultivar using hyperspectral imaging. Computers and Electronics in Agriculture, 179(April), 105807. https://doi.org/10.1016/j.compag.2020.105807. https://doi.org/10.1016/j.compag.2020.105807.
 
Gerszberg, A., Hnatuszko-Konka, K., Kowalczyk, T., & Kononowicz, A. K. (2015). Tomato (Solanum lycopersicum L.) in the service of biotechnology. Plant Cell, Tissue and Organ Culture, 120(3), 881–902. https://doi.org/10.1007/s11240-014-0664-4.
 
Kazemi, F. (2021). Feasibility of using visible and near-infrared spectroscopy to identify squash yellow mosaic virus infection before symptoms appear on leaves, Master's Thesis, Faculty of Agriculture, Shiraz University, Shiraz, Iran. (In Persisan).
 
Madden, L.V., Hughes, G., & Bosch, F. (2007). The Study of Plant Disease Epidemics. American Phytopathological Society (APS Press), University of Minnesota.
 
Mahlein, A.-K., Oerke, E.-C., Steiner, U., & Dehne, H.-W. (2012). Recent advances in sensing plant diseases for precision crop protection. European Journal of Plant Pathology, 133(1), 197–209. https://doi.org/10.1007/s10658-011-9878-z.
 
Miller, S. A., Beed, F. D., & Harmon, C. L. (2009). Plant disease diagnostic capabilities and networks. Annual Review of Phytopathology, 47, 15–38. https://doi.org/10.1146/annurev-phyto-080508-081743.
 
Miller, P. W., & Bollen, W. B. (1946). Walnut bacteriosis and its control. Technical Bulletin Oregon Aricultural Experiment Station, 9, 107 pp.
 
Moffett,  M.  L.  (1973).  Bacterial  spot  of  stone  fruit  in  Queensland.  Australian Journal of Biological Sciences, 26, 171- 179.
 
Moghadam, P., Ward, D., Goan, E., Jayawardena, S., Sikka, P., & Hernandez, E. (2017). Plant disease detection using hyperspectral imaging. DICTA 2017 - 2017 International Conference on Digital Image Computing: Techniques and Applications, 2017-Decem, 1–8. https://doi.org/10.1109/DICTA.2017.8227476.
 
Nicolaï, Bart M, Beullens, Katrien, Bobelyn, Els, Peirs, Ann, Saeys, Wouter, Theron, Karen I, and Lammertyn, Jeroen. (2007). Nondestructive measurement of fruit and vegetable quality by means of NIR spectroscopy: A review. Postharvest Biology and Technology, 46(2): 99-118. https://doi.org/10.1016/j.postharvbio.2007.06.024.
 
Razdan, M. K., & Mattoo, A. K. (2006). Genetic improvement of Solanaceous crops: Tomato 2,  CRC Press.
 
Rumpf, T., Mahlein, A. K., Steiner, U., Oerke, E. C., Dehne, H. W., & Plümer, L. (2010). Early detection and classification of plant diseases with Support Vector Machines based on hyperspectral reflectance. Computers and Electronics in Agriculture, 74(1), 91–99. https://doi.org/10.1016/j.compag.2010.06.009.
 
Saffari, M. (2018). Early detection of cucumber mosaic virus by visible-near-infrared spectroscopy method, master's thesis, Isfahan University of Technology, Isfahan, Iran. (In Persian).
 
Shahin, M. A., & Symons, S. J. (2011). Detection of Fusarium damaged kernels in Canada Western Red Spring wheat using visible/near-infrared hyperspectral imaging and principal component analysis. Computers and Electronics in Agriculture, 75(1), 107–112. https://doi.org/10.1016/j.compag.2010.10.004.
 
Shirzadifar, A., Bajwa, S., Nowatzki, J., & Shojaeiarani, J. (2020). Development of spectral indices for identifying glyphosate-resistant weeds. Computers and Electronics in Agriculture, 170 (February), 105276. https://doi.org/10.1016/j.compag.2020.105276.
 
Shirzadifar, A., Bajwa, S., Mireei, S. A., Howatt, K., & Nowatzki, J. (2018). Weed species discrimination based on SIMCA analysis of plant canopy spectral data. Biosystems Engineering, 171, 143–154. https://doi.org/10.1016/j.biosystemseng.2018.04.019.
 
Siedliska, A., Baranowski, P., Zubik, M., Mazurek, W., & Sosnowska, B. (2018). Detection of fungal infections in strawberry fruit by VNIR/SWIR hyperspectral imaging. Postharvest Biology and Technology, 139(January), 115–126. https://doi.org/10.1016/j.postharvbio.2018.01.018.
 
Sivakumar, M. (2011). Long wavelength near-infrared hyperspectral imaging for classification and quality assessment of bulk samples of wheat from different growing locations and crop years. University of Manitoba (Canada).
 
Xie, C., Shao, Y., Li, X., & He, Y. (2015). Detection of early blight and late blight diseases on tomato leaves using hyperspectral imaging. Scientific Reports, 5, 1–11. https://doi.org/10.1038/srep16564.
 
Xu, H. R., Ying, Y. B., Fu, X. P., & Zhu, S. P. (2007). Near-infrared Spectroscopy in detecting Leaf Miner Damage on Tomato Leaf. Biosystems Engineering, 96(4), 447–454. https://doi.org/10.1016/j.biosystemseng.2007.01.008.
 
Xuan, G., Li, Q., Shao, Y., & Shi, Y. (2022). Early diagnosis and pathogenesis monitoring of wheat powdery mildew caused by blumeria graminis using hyperspectral imaging. Computers and Electronics in Agriculture, 197(April), 106921. https://doi.org/10.1016/j.compag.2022.106921.
 
Zhang, H., Paliwal, J., Jayas, D. S., & White, N. D. G. (2007). Classification of fungal infected wheat kernels using near-infrared reflectance hyperspectral imaging and support vector machine. Transactions of the ASABE, 50(5), 1779-1785.
 
Zehr, E. I., & Sheard, D. P. (1996). Bacterial spot of peach as influenced by water congestion, leaf wetness duration and temperature. Plant Disease, 80, 339- 340.
Journal of Agricultural Mechanization
Volume 10, Issue 2
June 2025
Pages 75-89

Files

History

  • Receive Date: 21 February 2025
  • Revise Date: 02 June 2025
  • Accept Date: 11 June 2025

Share

How to cite

Statistics