基于机器视觉的织带瑕疵检测设备设计OA

To overcome the problems of low efficiency and limited accuracy of manual inspection,a multi-lane parallel online webbing defect inspection equipment was developed.The mechanical system integrates synchronized conveying and a stop-at-position belt-holding repair unit.On the algorithm side,a YOLO detection algorithm based on Edge-Adaptive Multi-scale(EAM)feature enhanced target detection network was proposed.The Edge-Guided Directional Adaptive Gating Module(EGDAGM)module was embedded in the shallow and middle layer of the backbone to enhance thin strip-like structures and high-frequency textures.The Mixed Local Channel Attention Module(MLCAM)module was introduced at the top stage to improve the discriminability power of high-level semantics.In addition,the Adaptive Weighted Pyramid Fusion Networ(AWPFN)replaced the conventional Path Aggregation Network(PAN)and Feature Pyramid Network(FPN)to achieve branch-weighted cross-scale fusion.The simulation results show that compared with the original YOLOv11 model,the mAP50 of the improved model on the Web5D dataset is increased to 93.2％,and the mAP50-95 is increased to 52.9％.Field tests demonstrated that the equipment reached a single-lane speed of up to 30 m/min,supported six-lane parallel inspection with coordinated stop control,significantly reduced manual workload and improved detection stability.The proposed equipment provides an efficient automated solution for online quality inspection of webbing materials.