基于机器视觉的紧密排列目标检测研究

Research on detection of closely arranged objects based on machine vision

机器视觉技术为工业机器人视觉导引系统提供重要的支持。随着生产效率的不断提高,流水线中常呈现多个货物无规律紧密排列运输的情况。在这种情况下,传统的基于模板匹配的目标检测算法的定位效果和识别精度准确率较低。为了解决这一问题,文中采用了基于深度学习的YOLOv8算法对紧密排列的物体进行检测,并针对紧密排列物体的特点建立相关数据集,通过实验验证了其在结构化环境下的性能。实验结果表明,YOLOv8算法训练后得到模型在拥有较快收敛速度的同时均值平均精度稳定在0.95左右,并且在多个真实环境中,模型对于无序队列目标的识别平均准确度能达到91.64%,为流水线中存在紧密排列现象的物品的定位和识别提供了一种有效的解决方法。

Machine vision technology is an important support for industrial robot vision guidance system.With the continuous rise of production efficiency,Many irregular and closely arranged goods are often transported together on the assembly line.Under such circumstances,the traditional target detection algorithm based on template matching has poor positioning performance and recognition accuracy.To solve this problem,YOLOv8 algorithm based on deep learning was used to detect closely arranged objects,and related data sets were established based on the characteristics of these objects,and the performance in structured environment was verified by experiments.The experimental results show that the model trained by YOLOv8 algorithm has a fast convergence speed and the average accuracy is stable at about 0.95,and in many real environments,the average accuracy of the model for identifying disorderly queue targets can reach 91.64%,which provides an effective solution for positioning and identifying closely arranged items on the assembly line.