Linear discriminant analysis in red sorghum using artificial intelligence

Red Sorghum landraces are predominantly cultivated in southern districts of Tamil Nadu, India and are noted for its anthocyanin (3-deoxyanthocyanin) content, high antioxidant activity, low glycemic index, dietary fiber, astringency, fodder value, high protein, biotic and abiotic stress resistance. In this study, the seeds of 72 sorghum landraces collected from Tamil Nadu were analyzed to measure morpho-colourimetric data using an image analyser. Artificial Intelligence is widely used in genetic research for its uncompromised accuracy and speed. The observed phenotypic data is used to create a statistical database and analysed by linear discriminate analysis (LDA) to identify and discriminate landraces for utilization in breeding projects. The classification is based on sorghum descriptors established by Indian Institute of Millet Research, Hyderabad. The LDA for the trait seed size had discriminated the landraces into three groups with an accuracy of 91.83%. Five groups were observed based on grain color Red Green–Blue colour space Model and CIELAB (Commission International de I’Elcairage Lightness Red/Green values Blue/Yellow values) colour space with accuracy 53.06%. The result showed that discrimination of landraces was possible with great accuracy by image analyser. The image analysis of seeds is found to be more effective, easy, non-destructive and fast method to classify landraces. This new phenomic approach is proved to be an important tool in the early selection of genotypes and facilitates speed breeding in colored sorghum.