基于YOLOv5的打磨表面质量快速评价方法

Rapid evaluation method of polished surface quality based on YOLOv5

针对手机壳表面质量缺陷检测目前存在自动化程度低、检测精度低、评估效率低等问题,本文提出了一种基于YOLOv5的打磨表面质量的快速评价方法。首先为提高特征的提取能力对YOLOv5模型进行了改进,并引入注意力机制对特征图的不同通道进行权衡,进一步提高了对较小特征的识别精度。利用改进后的YOLOv5模型对特征进行识别,计算加工件表面单位面积内的特征数量作为表面质量评价方式,对不同打磨参数下加工工件表面质量进行评价。结果表明改进后的YOLOv5提高了对特征的识别精度,振纹的精度提高了19.2%、达79.5%,斑块的精度提高了14.2%、达85.4%,并实验证明该评价方法仅需700 ms就可判定加工件表面质量是否符合标准。

Aiming at the problems of low automation,low detection accuracy,and low evaluation efficiency in the detection of surface quality defects of mobile phone cases,this paper proposes a rapid evaluation method for polished surface quality based on YOLOv5.First,the YOLOv5 model is improved to enhance the feature extraction ability,and an attention mechanism is introduced to weigh the different channels of the feature map,which further improves the recognition accuracy of smaller features.The improved YOLOv5 model is used to identify the features,and the number of features per unit area of the workpiece surface is calculated as the surface quality evaluation method,and the surface quality of the machined workpiece under different grinding parameters is evaluated.The results show that the improved YOLOv5 improves the recognition accuracy of features,the accuracy of vibration pattern is increased by 19.2%、to 79.5%,and the accuracy of the plaque is increased by 14.2%、to 85.4%.Experiments show that the evaluation method only needs 700 ms to determine whether the surface quality of the workpiece meets the standard.