基于YOLOv8的蕾丝花边缺陷智能检测与质量评级系统OA

To address the low efficiency and insufficient accuracy of lace trimmings by manual inspection in industrial production,as well as the significant cost increase associated with deploying multiple industrial cameras for full-coverage inspection,a deep learning based surface defect detection and grading system of lace trimmings was proposed.The system simulated an actual production line to build an online inspection platform,dynamical images were captured by industrial cameras,real-time defect detection with a YOLOv8 model accelerated by TensorRT was adopted.It accurately recognized four common surface defects in lace trimmings.Based on defect type,frequency and severity,the system automatically assigned a quality grade to each product.A multi-threaded human-machine interface,developed in Python and PyQt5,provided real-time detection visualization,quality grading,performance metric display and inspection log recording,thereby supporting production monitoring,data analysis and process optimization.Experimental results demonstrated that detection accuracy of the system was higher than 91%and the processing speed was 62 frames per second,satisfying industrial real-time requirements.It significantly enhanced the precision and efficiency of lace fabric detection,offering practical value for advancing automation in textile manufacturing.