WEED DETECTION AND CLASSIFICATION

Abstract

In the agricultural sector, weed detection and categorization are of critical technical and financial relevance. It is possible to automate various processes, such as shape, colour, and texture, in weed control systems. Building a machine vision weed control system that can locate weeds in real-time is the aim of this article. The system is designed to categorise photos into wide and narrow classes for on-the-fly spraying of selective herbicides. The developed Edge Link Detector-based algorithm has been put to the test on weeds in various areas, and the results have revealed how good the algorithm is at identifying weeds. Additionally, the outcomes demonstrate a very trustworthy performance on weeds under various field circumstances. Over 240 sample photos (wide, narrow, and no or little weeds) including 100 samples from broad weeds, 100 samples from narrow weeds, and the remaining 40 samples from no or little weeds were classified with over 93 percent accuracy, according to the results analysis.