At present, machine vision based defect detection is a hot research channel to realize automatic defect detection technology. Based on the ultra thin Gigabyte single port network transformer chip product as the carrier, the paper puts forward a system research on the automatic defect detection technology based on machine vision to eliminate and sort the defective chip products in the manufacturing process. The research focuses on the development of a set of high efficiency, precision, stable ultrathin single port network transformer chip automatic defect detection platform. The research content mainly focuses on the design of detection platform, vision system setting, image analysis and processing, and control system implementation. The results show that the accuracy of defect detection is 95%, and the detection efficiency is 38% higher than that of manual testing, which meets the research expectation.