采用ORB-SVM模型的铝合金熔滴复合电弧堆积层形貌缺陷快速识别

Rapid Identification of Morphology Defects of Arc Welding Assisted Aluminum Alloy Droplet Deposition Based on ORB-SVM

针对铝合金熔滴复合电弧增材制造堆积层潜在的形貌缺陷,以及图像处理时长导致的整个监控系统滞后的问题,建立了一个快速原位被动视觉系统来识别缺陷类型。首先,基于ORB算法自动提取堆积层形貌关键点的二值字符串描述子;然后,采用BOW模型得到维度相同的图像特征描述向量;最后,利用提取的特征向量对SVM分类器进行训练,并在分类准确性和处理时间方面验证了性能。结果表明,ORB-SVM模型对堆积层形貌识别的准确率为0.96,特征提取、编码与识别总时间约为9ms/帧,为确保熔滴复合电弧增材制造在复杂制造条件下的堆积层精度提供了一种可行的解决方案,并且在实时监控应用中具有较大的潜力。

A fast in-situ passive vision system is established to identify the defect types in the context of the potential morphology defects of the arc welding assisted aluminum alloy droplet deposition additive manufacturing and the lag of the entire monitoring system caused by the image processing time.Firstly,the binary string descriptors of the key points of the deposition layer morphology are automatically extracted based on ORB algorithm;then the BOW model is used to obtain the image feature description vectors with the same dimension;finally,the SVM classifier is trained with the extracted feature vectors,and the performance is verified in terms of classification accuracy and processing time.The results show that the ORB-SVM model has an accuracy of 0.96 for recognizing the deposition layer morphology,and the total time for feature extraction,encoding and recognition of a single image is about 9 milliseconds.In addition,it provides a feasible solution to ensure the deposition layer accuracy of arc welding assisted droplet deposition additive manufacturing under complex manufacturing conditions and has a great potential in real-time monitoring.