作者：常景景¹, 郑鹏¹, 曹满义¹, 王文秀¹, 张乃勇²

作者单位：1. 郑州大学机械与动力工程学院，河南 郑州 450001;
2. 焦作市卷烟物流配送中心，河南 焦作 454000

关键词：粘连颗粒;尺寸测量;机器视觉;贝叶斯估计

摘要：

在高速流水线上对小视场物体大批量在线检测过程中，针对滚子、钢珠、卷烟爆珠等球形颗粒因体积过小、发生破裂、传输振动而引起的颗粒粘连，致使其尺寸无法精准检测的问题，采用机器视觉的检测手段，提出基于贝叶斯估计的尺寸视觉检测方法。首先，经图像采集系统获取粘连颗粒图像，对采集图像预处理去除干扰因素。之后，利用边缘检测方法进行轮廓坐标点提取，通过数据滤波选取有效轮廓点数据，代入到该文所提出的贝叶斯尺寸测量数学模型中，从而实现对粘连颗粒尺寸的精确测量。以卷烟爆珠为例进行试验验证，该方法测得的尺寸最小均方根误差为0.0496 mm，测量误差均值最小为0.02263 mm，验证所提方法的可靠性与稳定性，满足实际工业检测精度要求。该方法的提出可为高速流水线上小视场粘连颗粒测量研究提供基础。

Detection method of touching beads size based on Bayesian estimation
CHANG Jingjing¹, ZHENG Peng¹, CAO Manyi¹, WANG Wenxiu¹, ZHANG Naiyong²
1. School of Mechanical and Power Engineering, Zhengzhou University, Zhengzhou 450001, China;
2. Jiaozuo Cigarette Logistics Distribution Center, Jiaozuo 454000, China
Abstract: In the process of mass on-line detection of objects with small field of view on high-speed pipeline, aiming at the problem that the size of spherical beads such as roller, steel ball and cigarette capsules cannot be accurately detected due to the bead touching caused by too small volume, rupture and transmission vibration, a size detection method based on Bayesian estimation is proposed by using machine vision. Firstly, the image of adhesive particles is acquired by the image acquisition system, and the interference factors are removed by preprocessing the acquired image. Then, the contour coordinate points are extracted by using the edge detection method, and the effective contour points are selected by data filtering, and the data is substituted into the Bayesian size measurement mathematical model proposed in this paper, so as to realize the accurate measurement of the size of cigarette capsules. The results show that the minimum root mean square error is 0.0496 mm and the minimum mean error is 0.02263 mm, which verifies the reliability and stability of the proposed Bayesian mathematical model measurement method and meets the actual industrial detection accuracy requirements. The proposed method provides a basis for the measurement of small field of view particles on high-speed assembly line.
Keywords: touching beads;size measurement;machine vision;Bayesian estimation
2023, 49(4):26-32  收稿日期: 2021-07-06;收到修改稿日期: 2021-09-01
基金项目: 国家重点研发计划项目（2017YFF0206501）
作者简介: 常景景(1993-),女,河南商丘市人,硕士研究生,专业方向为计算机机器视觉
参考文献
[1] 刘磊, 陈爱军, 程楼, 等. 机器视觉波纹阻火盘表面波高测量方法[J]. 中国测试, 2020, 46(6): 44-50
[2] 刘桂雄, 刘思洋, 吴俊芳, 等. 基于深度学习的机器视觉目标检测算法及在票据检测中应用[J]. 中国测试, 2019, 45(5): 1-9
[3] 刘宰豪. 基于凹点和重心检测的粘连类圆形目标图像分割[D]. 西安: 西安电子科技大学, 2019.
[4] 林盼盼, 张凤月, 王立春. 约束条件下的线性贝叶斯估计[J]. 工程数学学报, 2020, 37(3): 269-280
[5] LIRA I, WOGER W. Bayesian evaluation of the standard uncertainty and coverage probability in a simple measurement model[J]. Measurement ence & Technology, 2001, 12(8): 1172
[6] 陈妮, 覃玉荣, 孙鹏飞. 基于脑电自回归预测的实时相位估计方法[J]. 电子测量与仪器学报, 2020, 34(6): 183-190
[7] ELSTER C. Calculation of uncertainty in the presence of prior knowledge[J]. Metrologia, 2007, 44(2): 111
[8] 彭黔荣. 卷烟的爆珠[J]. 中国烟草学报, 2020, 26(2): 113
[9] 俞喆俊, 洪汉玉, 章秀华, 等. 生产线复杂场景钢坯检测识别的定位方法研究[J]. 光电工程, 2012, 39(1): 54-61
[10] 翟昊, 罗晓琳. 基于改进Canny算子的红外弱小目标检测[J]. 舰船电子工程, 2021, 41(5): 150-153
[11] 秦国义, 夏青. 基于视频图像的影子运动轨迹检测方法[J]. 电脑与电信, 2016(6): 68-70
[12] 唐聪, 黎晓东, 孙洁香, 等. 面向烟草行业的智能边缘计算技术研究与设计[J]. 制造业自动化, 2020, 42(12): 135-137
[13] 秦钟伟, 陈捷, 洪荣晶. 基于机器视觉的滚动轴承滚动体检测[J]. 组合机床与自动化加工技术, 2019(1): 118-121