针对传统工业存在对微特电机电枢表面缺陷检测任务人工成本高、工作量大的问题,提出了一种改进YOLOv6 模型的微特电机表面缺陷检测算法.首先,在主干网络加入SimAM注意力模块,加强网络信息传递,提高模型对特征的敏感程度;其次,Neck端使用GSConv新型卷积方式,以减少模型计算量;最后,使用CIoU损失函数解决GIoU损失函数的局限性,以提升模型检测精度.将所提改进算法在微特电机表面缺陷检测公开数据集上MASS-DET上进行训练并测试,实验结果表明,改进后的算法检测精度优于原算法,其中缺陷检测结果的 mAP 值和 mAR 值分别提升了4.7％和2.5％.同时相比于一些其他目前先进的目标检测算法在精度和速度上均有提升,证明了改进算法的有效性.
Surface Defect Detection of Micro Motor Armature Based on Improved YOLOv6 Model
Aiming at the problems of high labor cost and heavy workload in traditional industry,an im-proved YOLOv6 model for surface defect detection of micro motor was proposed.Firstly,the SimAM atten-tion module is added to the backbone network to enhance the network information transmission and improve the sensitivity of the model to features.Secondly,the Neck end uses GSConv new convolution mode to re-duce model calculation.Finally,CIoU loss function is used to solve the limitations of GIoU loss function to improve the accuracy of model detection.The proposed algorithm was trained and tested on MASS DET,a public data set of micro motor surface defect detection.The experimental results show that the detection ac-curacy of the improved algorithm is better than the original algorithm,and the mAP and mAR values of the defect detection results are increased by 4.7% and 2.5%,respectively.At the same time,compared with some other advanced target detection algorithms,the accuracy and speed are improved,which proves the ef-fectiveness of the improved algorithm.
defect detectionYOLOv6attention mechanismGSConvloss function
万方数据
维普
