基于机器视觉的冻干粉中的异物检测分类技术研究

摘要

In this paper,digital image processing technology is used to detect and classify the foreign matter of lyophilized powder.In order to more efficiently detect and classify the existing foreign matter in lyophilized powder such as fibers,hair,particles of glass and other visible foreign matter,BP neural network and support vector machine (SVM) algorithms together with principal component analysis (PCA) feature extraction are used to do classification and recognition.Through industrial small sample data simulation,test results show that both methods have good feasibility and practicability.By comparison it is showed that the recognition based on PCA and SVM algorithm is higher than the recognition based on PCA and BP algorithm.%利用数字图像处理技术对冻干粉不良品进行检测分类.为了更高效地自动检测并分类出冻干粉中存在的纤维、毛发、玻璃碎屑等可见异物,研究了基于主成分分析(PCA)特征提取,并用BP神经网路和支持向量机(SVM)算法进行分类识别.通过工业小样本数据仿真实验,测试结果表明,两种方法都具有较好的可行性和实用性,相比之下基于PCA与SVM算法的识别率比基于PCA与BP算法的识别率高.