Identification of Some Citrus Genotypes using Leaf Shape Analysis Based on Elliptical Fourier Descriptors

Author: Maryam Abdolalipour*, Fariborz Zaare Nahandi*, Mohammad Reza Dadpour* and Zahra Sadighzadeh**

Download PDF

Abstract

Computer-based identification of plant species has been the subject to numerous investigations in the last few decades. Image analysis of plants organs have been performed using different features such as shape, color, texture and by employing various mathematical models. In this work, identification of eight citrus genotypes has been carried out through image analysis of their leaves. Elliptical Fourier Descriptors (EFDs) have been used to model the outline of the leaves and subsequent statistical analyses have been performed on these EFDs. The results from the discriminant analysis indicate success rates of greater than 95% in correct identification of different genotypes, which is comparable with the results in the existing literature. Furthermore, principal component and hierarchical clustering analyses have been proven to be efficient methods for classification of these genotypes. The results show that the image-processing method is effective in identification of the studied genot

Keywords

Image analysis, Elliptical Fourier Descriptors, Leaf Image, Shape Software, Chain code

Conclusion

Digital image analysis of eight citrus genotypes has been performed to evaluate the functionality of the computer-based identification to classify these genotypes and to obtain the level of similarities between them. Outline of 430 leaves belonging to above genotypes were subjected to image processing by EFDs. Subsequently, three different statistical methods have been applied to interpret the results from various aspects. CDA revealed promising results for accurate classification of the genotypes by achieving a success rate of greater than 95%. Furthermore, principal component analysis illustrated the prominent features which could be considered in the classification process. It was shown that using the first five principal components, at least 90% of the variances among each genotype could be retained. Finally, different genotypes of leaves were subjected to HCA which identified the overall similarity between the genotypes by clustering them into multiple groups based on the ecolo

References

-

How to cite this article

Maryam Abdolalipour, Fariborz Zaare Nahandi, Mohammad Reza Dadpour and Zahra Sadighzadeh (2016). Identification of Some Citrus Genotypes using Leaf Shape Analysis Based on Elliptical Fourier Descriptors , Biological Forum – An International Journal 8(1