摘要
浙江移动借助5G实现了某汽车零部件制造商的某品牌Logo外观的缺陷检测并自动标志。用结构光照射零部件,3D摄像头会采集零部件在各个角度的图像,并通过自行研发的神经网络结构程序进行AI最佳算法分析,由此发现并自动标记出有瑕疵的不良品。此外,由于AI具备深度学习的特征,随着垂直领域缺陷数据的不断累积,公司后续还将通过大数据平台进行最新算法升级推荐。由于电镀件的特殊性,上下料环节需要人工参与,以免对零件造成二次损伤。本次针对某品牌Logo电镀件,样机可实现精度0.5mm的外观不良检测,检测周期10s/件,实现95%以上的准确率。
Zhejiang Mobile borrowed 5 G to realize defect detection and automatic marking of the logo appearance of a car parts manufacturer.By illuminating parts with structured light,the 3 D camera will collect the images of parts from different angles,and analyze the best algorithm of AI through the self-developed neural network structure program,so as to find and automatically mark the defective products In addition,because AI has the characteristics of indepth learning,with the accumulation of defective data in the vertical field,the company will follow-up Tongda data platform for the latest algorithm upgrade recommendation.For Volkswagen logo electroplated parts,the prototype can achieve 0.5 mm accuracy of poor appearance detection,detection cycle of 10 s/piece,achieve more than 95% accuracy.
出处
《中国仪器仪表》
2020年第3期34-36,共3页
China Instrumentation
