Optimization of Cocoa Pods Maturity Classification Using Stacking and Voting with Ensemble Learning Methods in RGB...

Area of Science:

• Agricultural Science
• Computer Vision
• Artificial Intelligence

Background:

• Early determination of cocoa pod maturity is crucial for harvest quality, yield optimization, and resource management.
• Immature or overripe pods result in lower quality cocoa beans, impacting profitability.
• Current methods for assessing maturity can be subjective and prone to human error.

Purpose of the Study:

• To develop and evaluate an objective, rapid method for assessing cocoa pod maturity using artificial intelligence and computer vision.
• To improve decision-making for optimal harvest timing, maximizing plantation yield and quality.
• To reduce losses associated with premature or late harvesting.

Main Methods:

• Utilized computer vision techniques with the gray level co-occurrence matrix (GLCM) algorithm for feature extraction.
• Analyzed images in both RGB (red, green, blue) and LAB (luminance, axis between red and green, axis between yellow and blue) color spaces.
• Applied and combined various classification algorithms using stacking and voting ensemble techniques for enhanced accuracy.

Main Results:

• Ensemble methods, particularly in the LAB color space, achieved superior performance.
• Voting and stacking techniques in the LAB color space scored 98.49% and 98.71% accuracy, respectively.
• RGB color space analysis yielded slightly lower but still high accuracy, with voting at 96.59% and stacking at 97.06%.

Conclusions:

• The combination of computer vision, AI, and ensemble methods offers a highly effective approach for accurate cocoa pod maturity classification.
• The proposed method significantly surpasses existing literature results, demonstrating its potential to revolutionize cocoa farming practices.
• Further exploration of ensemble techniques is recommended for optimizing performance in complex agricultural classification tasks.