As the demand for automation and quality assurance rises in the manufacturing industry, traditional quality control methods relying on manual visual inspection of small components have become increasingly inefficient and pose significant health risks. For small parts like zippers, defects that arise during production not only compromise product quality but can also impair functionality and aesthetics, impacting the end-user experience. Conventional inspection methods typically depend on workers manually inspecting products under strong lighting conditions, which, over time, may lead to adverse effects on their vision. This study addresses these limitations by developing an automated defect detection system based on machine vision technology. Leveraging the YOLOv7 object detection algorithm as its core, the system is integrated with edge computing to enable high-speed, on-site analysis of zipper images. The proposed system can identify and classify defects in real-time, achieving a marked improvement in both the speed and accuracy of defect detection processes compared to traditional methods. Furthermore, by reducing the reliance on human inspectors in high-stress visual environments, this system minimizes potential health risks, offering a safer and more sustainable alternative for quality control in zipper manufacturing. The developed solution presents a significant advancement in automated inspection for small-scale components, contributing to improved quality standards and operational efficiency within the manufacturing sector.