摘要
针对煤质近红外分析模型中光谱样本的特点,提出了基于并行最小二乘回归估计(P-LSRE)方法的争议光谱判别。首先,将光谱的全谱区划分为4个子区间,利用并行最小二乘回归估计方法以中心光谱为基准并行拟合建模光谱数据集;然后,以由相似度与相异度构成的相关性测度为衡量标准判别原光谱与拟合光谱的相关关系;最后,对判别信息进行有机融合标定光谱的属性。实验结果表明,该方法可准确地识别争议光谱,且具有较强的泛化性,适用于不同信噪比的光谱数据检测。
Aiming at the spectral characteristic in coal-NIRS model, a dispute spectrum distinguishing method is put for- ward based on parallel least square regression estimation （P-KSRE）. Firstly, the spectrum is divided into 4 subintervals, and P-LERE is employed to concurrently fit modeling data set with the central spectrum as the reference ;then,the relation measure that consists of similarity and dissimilarity is regarded as the measurement criterion to distinguish the correlation between the original spectrum and fitted spectrum;finally, the distinguishing information is fused organically to calibrate the property of the spectrum. The experiment results indicate that the method can accurately distinguish dispute spectrum, has strong generalization and is suitable for the spectrum data detection with different signal to noise ratios.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2013年第4期902-907,共6页
Chinese Journal of Scientific Instrument
基金
江苏省研究生培养创新工程基金项目(CXZZ12-0933)
高等学校博士学科点专项科研基金项目(20110095110011)
国家自然科学基金项目(61072094)资助
