Automated Vision-Based Grading and Authenticity Verification of Harumanis Mangoes using Shape and Contour Analysis

Abstract

Harumanis mangoes are a premium agricultural product from Perlis, Malaysia, valued for their unique quality and export potential. Traditional grading methods rely on manual inspection, which is time-consuming, inconsistent, and susceptible to misclassification and fraud involving visually similar mango varieties. To overcome these limitations, an automated vision-based grading system was developed using NI myRIO and LabVIEW. The system captures top-down images in a controlled environment and processes them through a two-level shape analysis, including general shape and tail contour matching, to verify varietal authenticity. Surface defects such as black stains, brown stains, and bruises are detected using custom image processing pipelines, while weight estimation is performed using volume approximation with a correction model based on linear regression. The system achieved 100 percent identification accuracy for authenticity and demonstrated a strong correlation between estimated and actual weights, with an R-squared value of 0.9537 and an 80 percent reduction in mean absolute error. These results highlight the system’s effectiveness in providing fast, consistent, and non-destructive grading for Harumanis mangoes in post-harvest applications.