Plant Leaf Recognition Using Texture, Colour, and Vein Density Features

Plants are essential in our ecosystem. There is various kind of plants in our ecosystem. Hence recognizing and categorizing them is a crucial task for humankind. In a nutshell, the proposed study discusses how to identify plant species using image processing techniques, the proper feature extraction of plant species recognition, and how classification can improve the accuracy of plant leaf classification. Image input, image pre-processing, feature extraction, and classification using Support Vector Machine (SVM) are the four main stages used in the proposed recognition method. In this study, the vein density feature extraction uses parameters like the count of vein pixels and the area of the leaf and colour feature is extracted through red, green and blue channels respectively, while the texture feature extraction uses contrast, correlation, inverse different moments, and entropy parameters. Support Vector Machine classification was applied for all features, and the results for separated texture, colour and vein density accuracy were 65.44%, 74.17% and 38.91%, respectively by using the Flavia dataset. The accuracy for extracting all features at once was 84.46%. As per the result of the study, the best way of identifying plant species is the combination of all selected three features.