Leaf recognition for plant identification is very important in agriculture. Leaves of different plants have distinctive features to be used for categorization. Of the numerous features, one of the prominent biometric features is the leaf vein. The extraction and classification of leaf veins based on these characteristics contributes to more precise identification of plants. In practice, leaf vein extraction becomes difficult due to changes in different lighting conditions and orientations. This work focuses on the extraction of veins using ridge orientation and frequency estimation using an area mask that under various conditions produces a good quality vein structure. The vein structure thus obtained is used by the Harris corner detector to classify keypoints. Using the SURF feature extraction process, features are extracted from the keypoints and finally the qualified and query images are compared using FLANN matcher to classify the correct leaf species. The Flavia leaf image database is used for 32 different species, resulting in an accuracy of 98.75%. For the identification of plant leaves in the real world, the proposed technique may be used for the identification of medicinal plants and other plant categories. This technique can be used to classify dry leaf veins, which can further extract the characteristics and identify the organisms. This paper presents how, using frequency estimate and area mask, leaf veins are segmented and extracted.
Author (s) Details
Dr. Juby George
Department of Computer Applications, Marian College Kuttikkanam, Idukki District, Kerala, 685531, India.
Dr. S. Gladston Raj
Department of Computer Science, Govt. College Nedumangad, Thiruvananthapuram, Kerala, India.
View Book :- https://stm.bookpi.org/RDER-V11/issue/view/26