Document Type : Original Article
Authors
1
Assistant Professor, Department of Plant Production, College of Agriculture Science and Natural Resources, Gonbad Kavous University, Gonbad, Golestan, Iran.
2
Professor, Department of Plant Production, College of Agriculture Science and Natural Resources, Gonbad Kavous University, Gonbad, Golestan, Iran.
3
MSc Graduate, Department of Plant Production, College of Agriculture Science and Natural Resources, Gonbad Kavous University, Gonbad, Golestan, Iran.
Abstract
Objective
This study was conducted to evaluate the potential of digital image processing and artificial neural networks in identifying and classifying rice cultivars based on grain morphological characteristics and also to introduce an applied software in this field. Rice cultivars included Gil3, IR362542/2, Restore50, Domsiah, MusaTarom, GHARIB3, Gharibsiahryhani, IR50maz, Line304, LINE229-2, Nemat, KMP41, DCL, Lebant, IR67017, DomsiahSolymandarab, Dashtisard, Hashemi, Dolar, IR24, IR50, line831, IR3441, AnbarboElam, CY, Mehr, Line213, Fujiminuri, Hasani, Ghasraldashti, TE, Sangtarom, Dasht, line216, Vad, IR662320, Canhopatra and Usen rice lines and 17 lines resulting from crossing Azucin and Bala. Evaluation of morphological diversity is considered an essential step in breeding programs and conservation of plant genetic resources.
Materials and Methods
Rice grains used in this study were cultivated during the 2015–2016 growing season at the research farm of Gonbad Kavous University. After harvest, digital images of paddy grains were captured using a standard imaging chamber equipped with a CCD camera. Using MATLAB R2015a software and image processing algorithms, grain characteristics such as length, width, perimeter, area, roundness, and aspect ratio were extracted. A Self-Organizing Map (SOM) neural network was then applied to cluster the grains based on these morphological features under both normal and drought stress conditions.
Results
The image processing results demonstrated the capability of this method to extract the geometric features of rice grains accurately. The developed software also performed effectively in measuring these characteristics. The SOM neural network successfully clustered grains into distinct groups based on morphological traits, particularly length and width. The SOM visualization maps illustrated both the morphological differences among samples and the influence of drought stress on these traits.
Conclusion
The combination of digital image processing and a Self-Organizing Map neural network provides an efficient, non-destructive approach for assessing morphological diversity and classifying rice cultivars. This method has valuable applications in breeding programs, germplasm management, and seed quality assessment. Moreover, the developed software serves as a valuable tool for researchers in the field. Future studies are recommended to include a broader range of morphological, color, and textural traits in SOM modeling and to explore their relationships with genetic and agronomic characteristics of diverse rice cultivars.
Keywords
)