Vol. 11, Issue 2 (2022)

Author(s):
SK Mundotiya, Jagdish Goyanka, Ravi Kumar Mundotiya and Sunil Ramling Swami

Abstract:
India is one of the centers of origin for rice, where large number of native varieties and land races are available and grown in different agro-climatic zones having unique characteristics and great adaptability. Variety identification is of prime importance under the PPV&FR Act, rice varieties are registered on the basis of Novelty, Distinctiveness, Uniformity and Stability (NDUS) characters which are taken as per the stage of observations listed in DUS guidelines. The scope of morphological differences between the varieties is less due to narrow genetic base, and requires skilled human power which is subjective in nature. Also this process is time, labour and cost intensive. Hence new techniques are needed, thus the present study was proposed with the objectives of establishing distinctiveness in phenotypically similar rice (Oryza sativa L.) varieties using image analysis. The study grouped the 28 varieties into five different clusters based on these DUS parameters. An exhaustive image library for various plant parts viz. seed, leaf, stem and panicle was also developed for the material under study. Two different types of softwares were used for extraction of features from the images viz. Grain Analysis Software (for size and shape features) and MATLAB software (for textural features). The softwares for extracting features from stem and panicle were still under construction by the end of study, hence the stem and panicle images were not used for feature extraction. Based on the comparative clustering patterns on four different kinds of features viz. DUS parameters, seed imaging, leaf imaging and DUS in combination with imaging; the study revealed that image features extracted from seed were most helpful for distinguishing the varieties.