Proceedings of the Annual Meeting of the Cognitive Science Society

Human visual recognition is shaped by constraints such as sensitivity to peripheral evidence, integration of complementary visual cues, and tolerance for ambiguous category boundaries. We ask whether these principles can improve automated inspection in Surface Mount Technology (SMT), where defects are often subtle, off-center, and difficult to distinguish categorically. We implement these ideas in a standard classifier using lightweight modules for position-aware normalization, Fourier and color/position cue integration, and ambiguity-aware learning over confusable defect classes. Experiments on industrial and public PCB datasets show that these cognitively inspired modifications improve defect recall under low false-alarm conditions and encourage more consistent focus on defect-relevant regions. The results suggest that human-inspired perceptual constraints offer a useful framework for designing robust machine vision systems in applied settings.