Related Experiment Video
Updated: Jan 3, 2026
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
973
Weed Detection in Perennial Ryegrass With Deep Learning Convolutional Neural Network
Jialin Yu1, Arnold W Schumann2, Zhe Cao3
1Co-Innovation Center for Sustainable Forestry in Southern China, Nanjing Forestry University, Nanjing, China.
Frontiers in Plant Science
|November 19, 2019
Summary
Deep convolutional neural networks (DCNNs) show promise for precision weed control in turfgrass. VGGNet and DetectNet models achieved high accuracy in identifying dandelion, ground ivy, and spotted spurge, paving the way for smart sprayers.
Area of Science:
- Agricultural Engineering
- Computer Vision
- Plant Science
Background:
- Precision herbicide application reduces costs and environmental impact in turfgrass management.
- Intelligent spot-spraying systems utilize machine vision for autonomous weed control.
Purpose of the Study:
- To evaluate the effectiveness of deep convolutional neural networks (DCNNs) for detecting specific weed species in perennial ryegrass.
- To compare the performance of different DCNN architectures (VGGNet, AlexNet, GoogleNet, DetectNet) for weed identification.
Main Methods:
- Trained multiple DCNNs on a dataset of 15,486 negative and 17,600 positive images of perennial ryegrass with and without target weeds.
- Evaluated network performance using F1 scores, recall, precision, and Matthews correlation coefficient (MCC).
Main Results:
- VGGNet achieved high F1 scores (≥0.9278) and recall (≥0.9952) for detecting dandelion, ground ivy, and spotted spurge.
- DetectNet also demonstrated high effectiveness with F1 scores (≥0.9843) for dandelion detection.
- AlexNet showed generally lower performance than VGGNet, while GoogleNet exhibited low precision.
Conclusions:
- DCNNs, particularly VGGNet and DetectNet, are effective tools for developing machine vision systems for precision weed control in perennial ryegrass.
- This approach supports the development of smart sprayers for targeted herbicide application.
