基于高光谱成像技术的药用胶囊铬含量快速检测方法研究  

Rapid Detection for Toxic Capsules with Chromium Based on Hyperspectral Imaging Technology

作　　者：陈丰农 陈浦兰[3] 谢亚平 应南娇 Hamed Hamid Muhammed 杨勇 

机构地区：[1]杭州电子科技大学生命信息与仪器工程学院,浙江杭州310018 [2]School of Technology and Health,KTH Royal Institute of Technology [3]新疆医科大学附属肿瘤医院经管办,新疆乌鲁木齐830011

出　　处：《计量学报》2017年第6期765-769,共5页Acta Metrologica Sinica

基　　金：国家自然科学基金(61305037;31401667)

摘　　要：研究了基于高光谱成像技术快速检测药用胶囊铬含量的方法。首先应用传统的原子吸收分光法对药用明胶空心硬胶囊分析结果作为对照组,再对1048个胶囊(含正常和含有重金属铬)样本的高光谱数据进行降维和定性分析,最后选择偏最小二乘判别分析方法对光谱数据进行处理。当用4个偏最小二乘算子(LV)作为偏最小二乘法模型中的输入特征,分类准确率可达100%,2种样本的交叉验证和样本预测相关系数分别为0.923和0.972;敏感度与特异度均为100%。The rapid detection method of chromium content in medicinal capsules based on hyperspectral imaging was studied. Firstly, analysis of medical empty hard gelatin capsule was used as the control group by traditional atomic absorption spectroscopy. Then, the hyperspectral data of 1048 samples were analyzed by dimensionality reduction and qualitative analysis. Finally, the partial least squares discriminant analysis method was used to process the spectral data. If 4 LV are as input features in the partial lease squares discrimination analysis model, the classification accuracy reached 100%, the correlation coefficient of cross validation and sample prediction are 0.923 and 0.972, respectively. Sensitivity and specificity are both 100%.

关 键 词：计量学 胶囊检测 铬 高光谱图像 光谱分析 PLS-DA模型 

分 类 号：TB99[一般工业技术—计量学]