An effective identification between various plant species using shape descriptors and image processing technique

Abstract

A modern agricultural sector requires accurate crop identification and classification. A new computer vision system is presented here that successfully discriminates between various plant species in real time under uncontrolled lighting. Features are vital for image classification and shape descriptors are mainly considered in this study. This system consists of image processing delivering results in real-time and a pixel calculator with more accuracy. Using these components together results in an efficient, reliable system for achieving excellent results in many different situations. Tested on several leaf species taken from the UCI repository. The system successfully detects an average of 87% under different variety of species. Additionally, the system has shown to produce acceptable results even under extremely challenging conditions, such as disease infected leaf or irregular shape leaf. The leaf boundaries was determined and evaluated through Harris corner algorithm. Compared to other high-cost methods, it was observed high species classification and lower testing time for our approach. The researchers also discussed challenges and solutions related to leaf classification, including identifying different leaves, classes of leaf shapes, lighting conditions, and stages of growth.