Related Experiment Video
Updated: Oct 14, 2025
11:49
Cereal Crop Ear Counting in Field Conditions Using Zenithal RGB Images
Published on: February 2, 2019
9.5K
Cotton stubble detection based on wavelet decomposition and texture features
Yukun Yang1,2, Jing Nie1,2, Za Kan1,2
1College of Mechanical and Electrical Engineering, Shihezi University, Shihezi, 832000, Xinjiang, China.
Plant Methods
|November 3, 2021
Summary
This study developed a visual navigation system for cotton residue film recovery, improving efficiency. The system uses texture features and wavelet decomposition for accurate cotton stubble detection, enhancing agricultural practices.
Area of Science:
- Agricultural Engineering
- Computer Vision
- Image Processing
Background:
- Residual film pollution in cotton fields poses a significant environmental challenge.
- Current manual recycling methods are inefficient and labor-intensive.
- Developing automated visual navigation systems is crucial for improving work efficiency.
Purpose of the Study:
- To develop a robust cotton stubble detection algorithm for a visual navigation system.
- To enhance the efficiency and reliability of residual film recovery in cotton fields.
Main Methods:
- Extraction of texture features including GLCM, GLRLM, and LBP from stubble, film, and leaf images.
- Classification using Random Forest, Back Propagation Neural Network, and Support Vector Machine models.
- Investigation of texture features from wavelet decomposition coefficients for improved classification.
Main Results:
- The Back Propagation Neural Network classifier with GLCM texture features from original images showed optimal performance.
- Combining original image texture features with vertical coefficient texture features from coif3 wavelet decomposition yielded the best classification results.
- This combined approach increased classification accuracy by 3.8%, sensitivity by 4.8%, and specificity by 1.2%.
Conclusions:
- The developed algorithm effectively detects cotton stubble under various conditions (location, time, driving anomalies).
- Fusion of wavelet coefficient texture features and original image texture features provides a valuable method for stubble detection.
- This approach offers a reference for stubble detection in diverse crop cultivation systems.
