Development of a fuzzy model for differentiating peanut plant from broadleaf weeds using image features

Area of Science:

• Agricultural Engineering
• Computer Vision
• Machine Learning

Background:

• Weed identification is crucial for efficient crop management and reducing herbicide use.
• Automated systems require accurate plant discrimination for precision agriculture.

Purpose of the Study:

• To develop and evaluate a fuzzy logic model integrated with decision trees for distinguishing peanut plants from common weeds.
• To assess the effectiveness of different feature selection and decision tree algorithms for this classification task.

Main Methods:

• Extracted color and wavelet-based texture features from plant images.
• Applied Principal Component Analysis (PCA) and Correlation-based Feature Selection (CFS) for feature selection.
• Utilized J48, Random Tree (RT), and Reduced Error Pruning (REP) decision tree algorithms.

Main Results:

• CFS-selected features yielded the highest classification accuracies, reaching up to 80.83% for four plant categories.
• When differentiating peanut from weeds, accuracies exceeded 90% on both training and testing datasets.
• J48-CFS and REP-CFS models were identified as most suitable for fuzzy system development.

Conclusions:

• The developed decision tree-based fuzzy logic model effectively discriminates weeds from peanut plants.
• This approach is suitable for implementation in machine vision-based precision cultivating systems.
• Further refinement could improve accuracy in complex multi-class scenarios.