多尺度特征融合的钢坯红外成像测温系统实验设计  

Experimental Design of Steel Billet Infrared Imaging Temperature Measurement System Based on Multi-scale Feature Fusion

作　　者：张利欣[1] 南清荣 曾慧[1] 王粉花[1] ZHANG Lixin;NAN Qingrong;ZENG Hui;WANG Fenhua(School of Intelligent Science and Technology,University of Science and Technology Beijing,Beijing 100083,China)

机构地区：[1]北京科技大学智能科学与技术学院,北京100083

出　　处：《实验室研究与探索》2023年第12期15-19,共5页Research and Exploration In Laboratory

基　　金：国家自然科学基金项目(U2013202);教育部产学合作协同育人项目(201902020003);北京科技大学2020年本科教育教学改革项目(JG2020M24)。

摘　　要：针对冶金加热炉内钢坯温度测量,设计了基于多尺度特征融合的红外成像测温系统。首先,依据测温方式及应用现场环境,设计了测温系统整体结构;然后,根据系统测温原理和温度标定实验,推导系统测温模型,并建立图像灰度与目标温度的映射关系;最后,针对钢坯目标的识别问题,提出了一种基于多尺度特征融合的语义分割模型,实现了对钢坯目标的有效识别,识别准确率达到94.89%。实例验证结果表明,该测温系统可以实现炉内钢坯表面温度的测量,符合冶金领域实际应用和人才培养需求。An infrared imaging temperature measurement experimental device based on multi-scale feature fusion is designed for the temperature measurement of steel billets in metallurgical heating furnaces.Firstly,the overall structure of the temperature measurement system is designed based on the temperature measurement method and the application environment.Secondly,based on the system temperature measurement principle and temperature calibration experiments,the system temperature measurement model is derived and the mapping relationship between image grayscale and target temperature is established.Finally,for the recognition of billet targets,a semantic segmentation model based on multi-scale feature fusion is proposed to realize the effective recognition of billet targets.The results show that the temperature measurement system can achieve the measurement of billet surface temperature in the furnace,which meets the practical application requirements in metallurgical field.

关 键 词：红外成像测温 多尺度特征 图像分割 加热炉 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]