Related Experiment Video
Updated: Sep 3, 2025
10:06
Protocols for Quantifying Transferable Pesticide Residues in Turfgrass Systems
Published on: March 15, 2017
7.3K
Deep learning for detecting herbicide weed control spectrum in turfgrass
Xiaojun Jin1,2, Muthukumar Bagavathiannan3, Aniruddha Maity3
1College of Mechanical and Electronic Engineering, Nanjing Forestry University, Nanjing, 210037, Jiangsu, China.
Plant Methods
|July 25, 2022
Summary
Deep convolutional neural networks (DCNNs) can accurately identify weeds based on herbicide susceptibility. This enables precision spraying, significantly reducing herbicide use in turfgrass management.
Area of Science:
- Agricultural Science
- Computer Science
- Plant Science
Background:
- Precision spraying of herbicides can reduce overall input by targeting weeds based on their control spectrum.
- Evaluating deep convolutional neural networks (DCNNs) for weed detection and discrimination is crucial for efficient herbicide application.
Purpose of the Study:
- To assess the efficacy of DCNNs in identifying and differentiating weeds in turfgrass.
- To categorize weeds based on their susceptibility to ACCase-inhibiting and synthetic auxin herbicides.
Main Methods:
- Training GoogLeNet, MobileNet-v3, ShuffleNet-v2, and VGGNet models.
- Classifying vegetation into three categories: ACCase-susceptible weeds, synthetic auxin-susceptible weeds, and turfgrass (no weeds).
Main Results:
- ShuffleNet-v2 and VGGNet achieved high accuracy (≥0.999) and F1 scores (≥0.998).
- ShuffleNet-v2 demonstrated superior efficiency and reliability with fast inference times.
- The DCNNs successfully discriminated weeds based on herbicide susceptibility.
Conclusions:
- DCNNs trained on herbicide spectra can precisely identify weeds for selective herbicide application.
- This technology supports reduced herbicide usage and can be integrated into smart sprayer systems.
- Machine vision-based autonomous spot-spraying systems can benefit from this DCNN approach.
