Particular Leaf Contour-Based Feature Extraction Technique to Identify the Species when the Leaf is shrouded

Abstract

A critical and challenging in pattern recognition is the identification of plant species from obstructed leaf photographs. The biggest issue at that time is to accurately identify the species of leaf when all of the leaves are identical in appearance and obscured. Shape is one of the most important visual elements and is also recognized as a fundamental attribute for conveying the content of pictures. Since it can be difficult to gauge how similar distinct forms are to one another, as well as describe the content of shapes. The two main categories of shape descriptors are region-based and contour-based shape descriptors (CBSD) techniques. Region-based approaches use the complete area of an item for shape description as opposed to contour-based approaches that only use the information contained in an image's contour. In this study, we presented a shape description approach called Particular Contour-Based Shape Descriptors (PCBSD) for the identification of the plant leaves since the CBSD recovered the low level visual properties of the pictures. This method successfully captures the local and global characteristics of a leaf shape while preserving the translation, rotation, and scaling similarity transformations. This method is also quite compact and has a low processing complexity. To evaluate our experiments, we utilized Flavia datasets of typical plant leaves. We show that our technique created the best complete leaf match when high occlusion (around 50% occlusion) occurs. We may say that our method exceeds prior state-of-the-art shape-based plant leaf recognition algorithms and it generates accuracy of 76%.Picture processing methods are used to separate the leaf-based characteristics from the leaf image. Eventually, using machine learning methods, then leaf identification was accomplished.