多工位图像拼接的超高叠层薄片计数仪器开发  

Apparatus with sequential image stitching for counting ultrahigh stacked substrates

作　　者：曹三 赵宏[1,2] 甄小刚 肖昌炎 

机构地区：[1]湖南大学电气与信息工程学院 [2]昆山湖大机器人技术有限公司

出　　处：《电子测量与仪器学报》2018年第8期117-123,共7页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61571184,U1613209);长沙市科技计划(kq1706016);湖南省重点研发计划(2016GK2056)资助项目

摘　　要：为解决超高叠层薄片产品计数需求与单一相机测量量程有限间的矛盾,开发了一种基于线性模组多工位成像的机器视觉薄片计数仪器。在描述运动装置和成像模块等硬件系统构成的基础上,重点介绍多图像融合和线检测计数相关的图像处理算法:提出一种外加标记结合图像特征的精确配准拼接算法,实现不同工位图像的融合;运用零阈值Canny算子与Radon变换相结合的线检测算法实现薄片边缘检测,并通过统计Radon空间中极大值点的数量实现叠层薄片计数。实验证明,该仪器及方法对多种类型的超高叠层薄片都具有很高的检测精度,检测精度达99.78%。To solve the contradiction between the requirement of countingultrahigh stacked substrates and the limited measurement range of asingle camera,wedevelopan apparatus equipped with a linear motion module to acquire images sequentially and obtain the quantity of the substrates. With a briefdescription of the overall system configuration,such as the linear motion device and the imaging module,we put emphasis on image processing algorithms corresponding to image fusion and line detection. An image stitching algorithm using the image features combined with external markerswas used to stitch the image sequences,and a line detection algorithm combined zerothreshold Canny method with Radon transform was utilized to detect substrates’ fringes. By finding the localmaximumsamong the Radon space,we obtained the quantity of the stacked substrates. The experimental results show that thedeveloped apparatus and the proposed method achieve high detection accuracy for various types of ultrahigh stacked substrates,and the precision ratio reached 99. 78%.

关 键 词：机器视觉 叠层数量测量 图像拼接 线检测 RADON变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]