Weed Detection And Killing Using Laser Beam

Authors: Mr. Abhishek P. Bangar, Dr. Monika Deshmukh

Abstract: Manual weeding is labour-intensive, and chemical herbicides carry well-documented risks of soil contamination, herbicide-resistant weed populations, and residue in food crops. Laser-based weed control has emerged as a non-chemical alternative in which a focused beam irradiates a weed's apical meristem, using thermal energy to destroy the growth centre without disturbing the soil. This paper presents the design of an autonomous weed detection and laser-killing system that combines an RGB/NIR camera, a convolutional neural network (CNN) for real-time crop–weed discrimination, a galvanometer-steered laser module for meristem targeting, and a closed-loop verification step that re-images the treated area to confirm the outcome. The system is designed around two literature-grounded design choices: attribute-level targeting of the meristem, which the literature identifies as the most energy-efficient kill point, and a species- and stem-diameter-aware dose controller, since required laser energy varies substantially with plant morphology. Because this system is at the design stage and has not yet been field-deployed, its evaluation combines literature-reported figures from six recent laser-weeding studies with explicit design targets for detection accuracy, kill efficacy, and processing speed, so that a future prototype can be benchmarked directly against both.

DOI: http://doi.org/10.5281/zenodo.23188526