A NOVEL MACHINE LEARNING BASED ALGORITHM TO DETECT WEEDS IN SOYBEAN CROP

Rameen Sohail; Qamar Nawaz; Isma Hamid; Humair Amin; Junaid Nawaz Chauhdary; Syed Mushhad Mustuhzar Gilani; Imran Mumtaz

doi:10.21162/PAKJAS/21.479

No Detail Found!

A NOVEL MACHINE LEARNING BASED ALGORITHM TO DETECT WEEDS IN SOYBEAN CROP

Author(s)

Rameen Sohail Department of Computer Science, University of Agriculture,Faisalabad,Pakistan

Qamar Nawaz Department of Computer Science, University of Agriculture,Faisalabad,Pakistan

Isma Hamid Department of Computer Science, National Textile University Faisalabad,Pakistan

Humair Amin Dept of Sociology and Criminology, University of Sargodha,Sargodha, Pakistan

Junaid Nawaz Chauhdary Water Management Research Centre, University of Agriculture,Faisalabad,Pakistan

Syed Mushhad Mustuhzar Gilani University Institute of Information Technology PMAS-Arid Agriculture University Rawalpindi,Pakistan

Imran Mumtaz Department of Computer Science, University of Agriculture,Faisalabad,Pakistan

Abstract

The traditional ways of weed management including spraying of herbicides on the whole field and manually uprooting them, are still in practice in many agricultural farms. This leads to herbicide overuse that causes serious health issues due to food quality degradation and environmental pollution. Computer aided weed detection systems can help in smart utilization of herbicides by detecting weeds through images. The aim of this study is to propose a novel weed detection system that provides accurate results in recognizing crops and weed using Machine Learning and Image Processing techniques. The image dataset chosen for this work is comprised of four different classes including broadleaf weed, grass, soil, and soybean. The proposed algorithm extracts texture and color features from each image in dataset and uses Random Forest algorithm to train a model using extracted feature descriptors. The working of the model is evaluated by computing regression metrics, precision, recall and F1 scores. Results showed that the model achieved a correct classification accuracy of 91% for weed, 100% for soil, 90% for grass and 99% for the soybean crop. The complete program took only 80 sec to execute which is ideal for a real-time environment.

Publication Details

Page(s) 1007-1015

DOI 10.21162/PAKJAS/21.479

Published Journal: Pakistan Journal of Agricultural Sciences, Volume: 58, Issue: 3, Year: 2021

Request the full text from PASTIC

Keywords

machine learning, Machine Vision, Robotic Weed Control, Image processing techniques, automatic weed detection

Citations

Downloads

155

Views