Machine Learning Interventions for Weed Detection using Multispectral Imagery and Unmanned Aerial Vehicles -- A Systematic Review

TL;DR

This systematic review evaluates machine learning methods applied to multispectral imagery from UAVs for weed detection, highlighting current techniques, challenges, and future research directions to improve agricultural weed management.

Contribution

It provides a comprehensive survey of ML interventions for UAV-based multispectral weed detection, summarizing models, features, and performance analysis in this emerging field.

Findings

Various ML models used for weed detection identified

Key features from multispectral data analyzed for effectiveness

Challenges include data variability and model generalization

Abstract

The growth of weeds poses a significant challenge to agricultural productivity, necessitating efficient and accurate weed detection and management strategies. The combination of multispectral imaging and drone technology has emerged as a promising approach to tackle this issue, enabling rapid and cost-effective monitoring of large agricultural fields. This systematic review surveys and evaluates the state-of-the-art in machine learning interventions for weed detection that utilize multispectral images captured by unmanned aerial vehicles. The study describes the various models used for training, features extracted from multi-spectral data, their efficiency and effect on the results, the performance analysis parameters, and also the current challenges faced by researchers in this domain. The review was conducted in accordance with the PRISMA guidelines. Three sources were used to obtain the relevant material, and the screening and data extraction were done on the COVIDENCE platform. The search string resulted in 600 papers from all sources. The review also provides insights into potential research directions and opportunities for further advancements in the field. These insights would serve as a valuable guide for researchers, agricultural scientists, and practitioners in developing precise and sustainable weed management strategies to enhance agricultural productivity and minimize ecological impact.