Classification of Groundnut Pod Varieties using Histogram-Based Gradient Booster Classifier (HGBC) with Machine Learning Techniques

• Benjamin BELLO

  Author

• Jibril D. JIYA

  Author

• Aje TOKAN

  Author

• Ahmed MOHAMMED

  Author

Groundnut (Arachis hypogaea) is an important economic and food crop widely cultivated for human consumption, livestock feed and industrial applications. Groundnut pod varieties differ in size, shape, and surface characteristics (texture) which influence the threshing efficiency, seed damage, cleaning performance, and threshing machine parameter for post-harvest systems. However, conventional pod variety identification in most local processing environments is still performed manually, making the process labour-intensive, subjective, slow, and unsuitable for integration into intelligent threshing systems. This study developed a machine vision-based classification framework for automated classification of groundnut pod varieties using a Histogram Based Gradient Boosting Classifier (HGBC) that support intelligent threshing machine design and operation.  A total of 360 groundnut pod images comprising Exdakar (116), Jarma (99), and Samnut26 (145) varieties were acquired under controlled imaging conditions. The methodology involved image acquisition, preprocessing, segmentation, geometric and texture feature extraction, dimensionality assessment using Principal Component Analysis (PCA), and supervised machine learning classification. Extracted engineering features included pod length, width, height, area, perimeter, aspect ratio, circularity, weight, and texture, which are critical parameters for machine interaction analysis and threshing component design. The dataset was divided into a 70:15:15 ratio i.e 70% training, 15% validation, and 15% testing for model development and evaluation.  The HGBC model achieved an overall classification accuracy of 91.9%, with model performance of 95% precision, 95% recall, and 95% F1-score for Exdakar, 88% precision, 95% recall, and 91% F1-score for Jarma, and 93% precision, 88% recall, and 90% F1-score for Samnut26. The results demonstrate the technical feasibility of integrating machine vision techniques and intelligent classification into groundnut threshing systems for adaptive operations, improved threshing efficiency, and reduced seed damage.

• FJET_21_82_82

Copyright (c) 2026 FUDMA Journal of Engineering and Technology

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.