基于图像的血型卡灌装质量在线检测系统-包装工程.PDFVIP

第39 卷 第3 期 包 装 工 程 2018 年2 月 PACKAGING ENGINEERING ·157 · 基于图像的血型卡灌装质量在线检测系统 陈曦，柴晨晨，胡亚乔，刘凯龙，张旭 （河北工业大学，天津 300130 ） 摘要： 目的 测量灌装后血型卡的固液高度，同时检测所灌液体中是否存在气泡。方法 将摄像机捕获的 血型卡图像经旋转矫正、初步裁剪、定位分割后，得到6 个独立的微柱管，利用Sobel 水平边缘检测方 法分别提取出固液混合相和液相的水平边缘位置，进而计算相应高度，同时提出基于SVM 的气泡检测 方法，提取气泡的HOG 特征，选择RBF 核函数来训练检测模型，并建立分类模型，通过交叉验证的方 法获得模型最佳参数。结果 与手工测量结果进行对比，固液混合相高度误差不超过±0.1 mm ，液相高 度测量误差不超过±0.2 mm ，气泡检测正确率高于97% ，在效率上可满足实时检测的需要，高度测量和 气泡检测的时间总和小于1.5 s，满足实时性要求。结论 所设计的检测系统可以有效解决血型卡灌装质 量的在线检测问题。 关键词：血型卡；图像处理；Sobel 边缘检测；SVM；HOG 特征；交叉验证 中图分类号：TP391 文献标识码：A 文章编号： 1001-3563(2018)03-0157-05 DOI ：10.19554/j.cnki.1001-3563.2018.03.030 Online Detection System for Blood Card Filling Quality Based on Image CHEN Xi, CHAI Chen-chen, HU Ya-qiao, LIU Kai-long, ZHANG Xu (Hebei University of Technology, Tianjin 300130, China) ABSTRACT: The work aims to measure the solid-liquid height of blood card after filling and check whether bubbles ex- ist in the liquid filled. Six independent micro pipes were obtained by rotation correction, initial cutting and positioning segmentation of blood card images captured by cameras. The horizontal edge position between solid-liquid mixed phase and liquid phase was extracted in the method of Sobel horizontal edge detection. Then, the corresponding height was cal- culated. Meanwhile, the way of detecting bubbles based on SVM was put forward. The HOG characteristics of the bubble were extracted, and the RBF kernel function was selected to train the detection model and establish the classification model. The best model parameters were obtained by the way of cross validation. Compared wi