基于太赫兹成像的输送带撕裂检测与分类识别研究  

Research on Conveyor Belt Tear Detection and Classification Based on Terahertz Imaging

作　　者：蒋悦 凌平平 徐伟[1] JIANG Yue;LING Pingping;XU Wei(School of Electronics and Information Engineering,Tiangong University,Tianjin 300387,China;Tianjin Key Laboratory of Optoelectronic Detection Technology and System,Tianjin 300387,China)

机构地区：[1]天津工业大学电子与信息工程学院,天津300387 [2]天津市光电检测技术与系统重点实验室,天津300387

出　　处：《测试技术学报》2024年第2期100-108,共9页Journal of Test and Measurement Technology

摘　　要：针对现有输送带撕裂检测方法存在灵敏度和安全性能低,且无法消除复杂工作环境带来的影响等问题,提出了一种基于太赫兹成像技术的输送带撕裂检测方法。设计并搭建了连续波太赫兹反射式成像系统,采集输送带撕裂的太赫兹图像;对原始图像进行滤波等处理获得低噪声太赫兹图像;最后搭建了基于机器学习太赫兹图像自动分类识别系统,该系统通过提取太赫兹图像的灰度直方图统计特征和几何特征,构建太赫兹图像特征库,并利用特征选择去除特征冗余,最终结合分类器实现对输送带撕裂种类的自动分类与识别。结果表明:通过太赫兹反射式成像系统对输送带进行成像实验,验证了太赫兹波成像技术检测输送带撕裂的可行性;太赫兹图像自动分类识别系统实现了对输送带3种撕裂种类的自动分类与识别,在组合特征下,使用支持向量机(Support Vector Machine,SVM)的分类准确率可达91.6%。Aiming at the problems of existing conveyor belt tear detection methods such as low sensitivity and safety,and inability to eliminate the impact of complex working environments,a conveyor belt tear detection method based on terahertz imaging technology was proposed.This method first designs and builds a continuous wave terahertz reflective imaging system to collect terahertz images of conveyor belt tears;And then performing processing such as filtering on the original image to obtain a low noise terahertz image;Finally,an automatic classification and recognition system for terahertz images based on machine learning is established.The system extracts the statistical and geometric features of the grayscale histogram of terahertz images,con-structs a terahertz image feature library,and uses feature selection to remove feature redundancy.Finally,combined with a classifier,automatic classification and recognition of conveyor belt tear types is achieved.The results show that the feasibility of terahertz wave imaging technology for detecting conveyor belt tears has been verified through imaging experiments on conveyor belts using a terahertz reflective imaging system;The terahertz image automatic classification and recognition system achieves automatic classification and recognition of three types of tear on conveyor belts.Under combined features,the classification accuracy using Support Vector Machine(SVM)can reach 91.6%.This study lays the foundation for the application of terahertz imaging technology in conveyor belt tear detection.

关 键 词：成像系统 太赫兹成像 输送带 撕裂检测 特征提取 机器学习 

分 类 号：TN29[电子电信—物理电子学]