基于机器视觉的棒线材废料尺寸综合测量

A Comprehensive Method for Measuring the Dimensions of Rolling Wire Rod Cut-Off Materials Based on Machine Vision

从控制棒线材剪切废料长度进而提高棒线材轧制过程自动化水平的角度出发,提出了一种基于机器视觉的棒线材剪切废料外形尺寸测量方法,并设计了对应图像采集装置。基于图像合并和UNet神经网络,该方法提出了剪切废料中变形段和正常段的判定标准并给出了变形段长度测量结果。所提出的测量方法对检测环境和目标外形的变化具有鲁棒性,对于样本量要求远低于其他传统神经网络因而更适用于实际工业数据集,而长度方向的量程也可随图像数量的增减而灵活调整。采用真实剪切废料样本进行测量实验,结果表明该测量方法简单有效、精确度高,且由于模型具有轻量化、可预训练的特点,该方法具备在线监测部署潜力。

This paper proposes a machine vision-based method for measuring the shape and size of scrap material resulting from the shearing of control rod wires.The aim is to improve the level of automation in the rod wire rolling process.Additionally,a corresponding image acquisition device has been designed.Based on image merging and the UNet neural network,criteria for determining the deformed and normal sections in the sheared waste material have been proposed,and measurement results for the length of the deformed sections were presented.The measurement method proposed in this paper exhibits robustness to variations in the detection environment and target shapes.In addition,it requires a significantly lower sample size compared to other traditional neural networks,making it more suitable for practical industrial datasets.The measurement range in the length direction can also be flexibly adjusted with the increase or decrease of the number of images.Experimental results demonstrate that the proposed measurement method is simple,effective,and highly accurate.Furthermore,due to the lightweight and pre-trainable nature of the model,this method has the potential for online monitoring deployment.