摘要
文章针对目前汽车生产线中焊接检测自动化程度较低、检测鲁棒性较差等问题,利用深度学习算法的特征提取能力,提出了一种基于改进的YOLOv3模型的焊缝缺陷检测方法。该方法基于连通域提取工件上的焊缝图像,并将提取到的焊缝图像输入到改进的深度学习模型中进行训练,通过融合大、中、小3个尺度的感受野,实现不同尺度焊缝缺陷的高精度识别和定位。试验结果表明,本方法可准确识别焊缝缺陷,且具有抗干扰能力强、识别速度快等优点,可有效提升焊装生产线的自动化水平。
Aiming at the problems of reduced welding inspection automation and poor inspection robustness in the existing automobile production lines, using the feature extraction ability of deep learning algorithms, a weld defect detection method based on the improved YOLOv3 model is proposed. This method extracts the weld image on the workpiece based on the connected domain, and inputs the extracted weld image into an improved deep learning model for training. By fusing the three-scale receptive fields of large, medium and small scales, high-precision identification and positioning of weld defects of different scales are realized. Experimental results show that this method can accurately identify weld defects, and has the advantages of strong anti-interference ability and fast identification speed, and can improve the automation level of the welding production line.
作者
马强
李龙涛
耿志卿
李心平
MA Qiang;LI Longtao;GENG Zhiqing;LI Xinping
出处
《上海汽车》
2021年第6期56-62,共7页
Shanghai Auto
