基于视频图像处理的液体粘度识别方法研究  

Research on liquid viscosity recognition method based on video image processing

作　　者：许瀚文 马小晶[1] 王宏伟[2] XU Hanwen;MA Xiaojing;WANG Hongwei(College of Electrical Engineering,Xinjiang University,Urumqi 830047,China;School of Control Science and Control Engineering,Dalian University of Technology,Dalian 116024,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830047 [2]大连理工大学控制科学与工程学院,辽宁大连116024

出　　处：《实验技术与管理》2022年第1期30-35,共6页Experimental Technology and Management

基　　金：国家自然科学基金项目(12002296);新疆维吾尔自治区自然科学基金项目(2017D01C085)。

摘　　要：为了在生产过程中实现液体粘度的在线监测,提出一种结合数字图像处理和液滴分析技术的液体粘度识别方法。通过实验获得不同粘度液滴的形成及生长过程图像,利用图像处理技术得到液滴生成周期的轮廓特征数据,并采用随机森林特征选择(RFFS)算法选择高质量特征子集作为输入,对麻雀搜索算法(SSA)优化的SVR模型进行训练,获得能够识别目标液体动力粘度的软测量模型。研究结果表明,基于SSA-SVR软测量模型的液体粘度识别方法能够快速有效识别液体的动力粘度,在所研究的粘度范围内检测精度优于传统测量方法,为液体粘度非接触检测提供了一种新的测量思路。In order to realize the online monitoring of liquid viscosity in the production process,a liquid viscosity recognition method combining digital image processing and droplet analysis technology is proposed.First,the formation and growth process of droplets of different viscosities are obtained through experiments,then the contour feature data of the droplet generation cycle is obtained by image processing technology,and finally,the random forest feature selection(RFFS)algorithm is used to select high-quality feature subsets as input,train the SVR model optimized by the sparrow search algorithm(SSA),and obtain a soft-sensing model that can identify the target liquid dynamic viscosity.The experimental results show that the liquid viscosity recognition method based on SSA-SVR soft sensing model can quickly and effectively identify the dynamic viscosity of the target liquid,and the detection accuracy is better than the traditional viscosity measurement method in the studied viscosity range,which provides a new measurement idea for the non-contact detection of liquid viscosity.

关 键 词：粘度识别 液滴图像分析技术 特征选择 支持向量回归 麻雀搜索算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]