A Study of Classification of Leaf Shape Categorization of two Stages Based on Fusion

Advancements in Leaf Shape Categorization Using Fusion Techniques

Abstract

The paper work only considered frontal and fresh leaves when creating leaf image databases.In future, CAP-LR can consider and analyze leaves that are wrinkled, occult and dry. Another difficultfield of research can be considered in similar ways is color discoloration or discolored leaves. Advancedoperations such as parallel workflow processing can be studied to increase the speed of the recognitionof the leaf disorder for plant identification. Parallel task processing can be used to group algorithmstogether during the different recognition stages. This study proposed techniques for improving theoperation of plant identification leaf recognition. Positive results from the different experiments showthat the model proposals discriminate effectively against the various leaves and identify the right plant tomatch the image of the inserted leaf. The botanists can therefore safely use this to increase theirefficiency in plant recognition and thus save valuable plants in order to improve the quality of human lifeand life on Earth.A system for improving the leaf image was proposed, called 'Enhanced wavelet-baseddemoizing with built-in edge enhancement and automatic contrast adjustment algorithm.The 197 methodcombines the wavelets, CLAHE (contrast adjustment), corner enhances and a relaxed middle filter (noiseremoval), with a single procedure to increase the visual quality of the leaf image. Texture-based colorsegmentation technique called 'Enhanced wavelet-based segmentation using the WCF method to extractthe leaf image from its background. Five types of features are extracted during extraction geometrical,texture, colour, fractal and leaf-like. These functions are combined to form the GLFS (Geometrically +Leaf), CLFS (Color + Leaf), TLFS (Texture + Leaf) and FLFS (Fractals + Leaf) functionalities. In addition toshared and merged operators to select optimal feature sets two selection algorithms for feature, thegenetic algorithm, and the Kernel main component analysis algorithm have been coupled.