大豆品质近红外光谱分析仪测量数据处理

利用自行设计的CCD近红外光谱仪器、采用光纤探头测得的大豆样品漫反射近红外光谱数据中含有各种噪声。经过分析,该光谱数据的噪声包括随波长变化的随机变量,引起光谱基线平移的随机常量,以及由于装样不同、样品散射系数变化而引起的光谱基线旋转。对于随机变量噪声,通过用背景、白板、样品各组重复采集、递归平均再计算光谱数据的方法减小,同时可节省存储空间,减小三组数据的重迭区域,增加光谱数据的有效范围。分别采用了局部平均阈值法、Savitzky-Golay卷积平滑、小波消噪等三种方法进一步消除随机变量噪声,结果表明:卷积平滑、小波消噪消除原始光谱数据噪声的效果较好。利用高阶小波分解变换获得光谱基线数据,将其从光谱数据中扣除,从而实现了光谱的基线校正。

Processing methods of data measured by NIR spectrum soya quality analyzer

There are some kinds of noises in the soya diffuse reflection NIR spectrum measured by optic-fibre probe and CCD prototype spectrometer.In order to the spectrum show the character information of soya samples,the proper methods should be used to de-noise based on the specialties of the noises.The noises included a random variable changing with the wavelength,a random constant bringing the spectrum baseline drift,and the spectrum baseline circumgyration brought by the difference of sample loading and the chan...