在部分生产线上，钢坯从炉内出炉的过程中，需要依靠人眼判断钢坯是否到达指定位置，到位后再利用推钢机推动炉内的钢坯完成出炉过程。在这个过程中，人眼长时间观察摄像头屏幕容易疲劳，人工劳动强度大成本高，生产工作效率较低。针对以上问题，文中提出利用机器视觉系统替代人类视觉系统进行钢坯位置的实时定位，首先将钢坯定位问题转换为关键点检测问题，然后提出了基于ResNet网络和基于关键点分割网络(Key Point Segmentation Network，KPSN)的两种模型来进行关键点检测，最后，通过测试和分析所提出的两种方法，提出了多方法融合的关键点检测方案，降低了极端情况下误检的风险，实际应用表明，文中所提方法具有较高的鲁棒性，达到了实际应用的要求。
Key Point Detection Method of Billet in Furnace Based on Deep Learning
In some production lines,during the process of the billet being released from the furnace,it is necessary to rely on human eyes to judge whether the billet has reached the specified position,and then the steel pusher is used to push the billet in the furnace to complete the firing process.In this process,human eyes are prone to fatigue when observing the camera screen for a long time,labor intensity is high,the cost is high,and production efficiency is low.To solve the above problems,this paper uses a ma-chine vision system instead of the human vision system for the real-time location of billet.Firstly,the billet positioning problem is transformed into the key point detection problem,and then two models based on the ResNet network and key point segmentation net-work(KPSN)are proposed to carry out the key point detection.Finally,through testing and analyzing the two methods proposed,a key point detection scheme of multi-method fusion is proposed,which reduces the risk of false detection in extreme cases.Practical application shows that the proposed method has high robustness and meets the requirements of practical application.
万方数据
维普
