Teaching mechanical systems integration through authentic project-based learning: An agricultural machine-vision grading project

Project-based learning (PBL) connects theory with engineering practice, but its applications in agricultural machinery education often remain focused on isolated technical tasks. This study investigated an authentic PBL course design implemented in an eight-week machine vision course. The PBL group (n = 12) completed an Agaricus bisporus (button mushroom) grading project involving platform construction, dataset development, model training, lightweight optimization, and edge deployment, while the comparison group (n = 18) received traditional instruction. Although the pre-test difference was not statistically significant ( p = .138), its moderate effect size suggested that baseline equivalence could not be fully assumed. After adjusting for pre-test scores using ANCOVA, the PBL group retained higher post-test performance than the traditional instruction group (adjusted means: 76.28 vs. 73.98; F (1,27) = 4.88, p = .036, partial η 2 = .153). The PBL group also achieved higher normalized learning gains (0.54 ± 0.13 vs. 0.43 ± 0.12, p = .029). Project evidence further documented students’ engineering practices and project performance during the authentic engineering workflow.