基于智能视觉的透明液体可见异物检测软件设计  

Automatic Impurity Detecting Software Based on Intelligent Visual Technology

作　　者：张朴[1] 程明[1] 

机构地区：[1]南京工业大学自动化与电气工程学院,南京211816

出　　处：《计算机测量与控制》2015年第3期1034-1037,共4页Computer Measurement &Control

摘　　要：目前在国内的透明液体生产线上,液体通常采用人工检测,效率低、准确率,所以基于图形处理算法以及QT,设计了一套基于智能视觉的透明液体杂质颗粒检测软件,重点实现了软件与算法的改进优化;获取的连续图像经自适应中值滤波预处理去除图像噪声后,通过改进的序列图像二次差分算法提取出序列图中的细小的异物,然后根据通过SIFT特征度量的MeanShift目标跟踪算法,跟踪液体图像中的细小异物;并根据异物的运动轨迹及大小,检测到有无异物以及异物的类型,进而进一步判断透明药品质量是否合格;结果测试表明,软件的准确率达到95%以上,检测分辨率达到了44μm,基本可以替代透明液体生产线上的人工检测。the domestic pharmaceutical manufacturers are almost transparent using artificial method to detect the product, hut this method is of low efficiency, and low accuracy rate. So we designed a automatic impurity detecting system based on QT and image processing module. After the noise of Continuous images was removed by adaptive median filtering. We proposed the improved second--difference algorithm to detect moving visible impurities, and used the SIFT features and MeanShift algorithm to track impurities. Finally we can judge whether the product is qualified. Experimental results showed that the accuracy rate of the system can reach 90% and the software can detect the impurity that is above 40 microns. Basically it can replace the original manual testing in medical production line.

关 键 词：智能视觉 杂质颗粒检测 序列图像二次差分 异物跟踪 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]