Weed and Corn Seedling Detection in Field Based on Multi Feature Fusion and Support Vector Machine

Area of Science:

• Agricultural Engineering
• Computer Vision
• Machine Learning

Background:

• Precision agriculture relies on accurate weed and crop detection for efficient herbicide application and fertilization.
• Existing methods may lack the accuracy needed for real-time robotic applications in diverse field conditions.

Purpose of the Study:

• To develop and validate a robust method for identifying and locating corn seedlings and weeds.
• To enhance the capabilities of agricultural robots for precise weeding and fertilization.

Main Methods:

• Established an image dataset of corn seedlings and weeds.
• Extracted and reduced dimensionality of multiple features (HOG, LBP, Hu moments, Gabor, GLCM, GLCM-G).
• Utilized Support Vector Machine (SVM) classification with optimal feature fusion.
• Employed K-means clustering for image segmentation and connected component analysis.

Main Results:

• The fusion of rotation invariant Local Binary Pattern (LBP) and gray level-gradient co-occurrence matrix features with SVM achieved the highest classification accuracy.
• The method accurately detected and located various weed species and corn seedlings in field images.
• Successfully provided positional data for herbicide spraying robots and fertilization machines.

Conclusions:

• The proposed multi-feature fusion approach significantly improves weed and crop detection accuracy in precision agriculture.
• This technique enables more effective automated weed management and targeted fertilization, reducing crop damage and resource waste.