基于图像的血型卡灌装质量在线检测系统  

Online Detection System for Blood Card Filling Quality Based on Image

作　　者：陈曦[1] 柴晨晨 胡亚乔 刘凯龙 张旭[1] 

机构地区：[1]河北工业大学,天津300130

出　　处：《包装工程》2018年第3期157-161,共5页Packaging Engineering

基　　金：天津市科技计划(14ZCDZSY00032)

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。The work aims to measure the solid-liquid height of blood card after filling and check whether bubbles exist 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 calculated. 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 with manual measurement results, the error of solid-liquid mixed phase height was less than ±0.1 mm, the error of liquid phase height was less than ±0.2 mm, and the detection accuracy of bubbles was more than 97%. The need for real-time detection could be met in efficiency. The total time of height measurement and bubble detection was lower than 1.5 s, which met the real-time requirements. The detection system designed can solve online detection problems of blood card filling quality effectively.

关 键 词：血型卡 图像处理 SOBEL边缘检测 SVM HOG特征 交叉验证 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]