Machine learning approaches for classification of indian leaf species using smartphone images

Abstract

Leaf detection and classification is fundamental to agriculture forestry rural medicine and other commercial applications Precision agriculture demands plant leaf disease diagnosis for automatic weed identification Leaf species identification using pattern recognition in environment forestry rural medicinal plants biodiversity areas etc need solutions for automatic tree species identification The problems in all the above areas have challenges in the resolution and extent of professionalism in which the leaf images are captured To address the needs of a common man and to disseminate the uses of Indian traditional herbs designing fast and robust machine learning algorithms could be applied for automatic leaf detection that are capable of leaf feature extraction from smartphone images for simple and easy identification Most of the existing literature on leaf classification focused largely on shape texture and color based features In spite of the presence of various big datasets on leaf classification research ensembling the learning over high dimensional features of smartphone leaf image data is less addressed To overcome this about 14 Indian herbal leaf species with 15 20 images per species are captured in smartphone The images are rotated for various leaf positions and a strong dataset is created The leaf species are subjected to various edge detection approaches to examine the shape vein and other morphological features of the leaves For identification and automatic recognition the method applied explores the possibility of k NN SVM in pre training with ANN followed by CNN based approaches newline