Mass prediction models for Buchanania lanzan seed and kernel based on physical parameters

Fruit/seed are typically graded based on size; however, grading them by mass could be more efficient. Therefore, the present study aims to forecast the mass of Buchanania lanzan seed and kernel using their physical parameters. Linear and nonlinear mass models were developed and analyzed using univariate and multivariate regression techniques. The models were formulated based on dimensions, projected areas, volumes and densities. The results indicated that mass modeling of seed was most accurate when based on width, geometric mean diameter, third projected area, and ellipsoid volume. In contrast, mass prediction for the kernel was best achieved using length, arithmetric mean diameter, first projected area, and ellipsoid volume. The quadratic models, based on ellipsoid volume for seed and artimatic mean diameter for the kernel, demonstrated the highest R2 values and lowest RMSE indicating their effectiveness in mass prediction. These findings are significant for the development of advanced post-harvest machinery, offering a more efficient approach to seed and kernel grading.