基于光谱微分滤波及多元校正的马铃薯干物质含量快速检测

DRY MATTER CONTENT DETECTING OF POTATOES BASED ON SPECTRA DERIVATIVE,NORRIS FILTERING AND MULTIVARIATE APPROACH

针对生产实际中缺乏快速的品质检测手段影响马铃薯产业发展的问题,开展基于漫反射光谱的马铃薯干物质含量检测研究。采用一、二阶微分及Norris微分滤波对光谱数据进行预处理,以消除干扰信息的影响。分析了主成分回归(PCR)和偏最小二乘(PLS)两种多元校正法在建立校正模型中的特点,分别建立了干物质含量校正模型,通过外部验证确定适合的建模方法。131个样品的检测研究结果表明,一阶微分光谱Norris滤波(分段长度为17点,分段间距为4点)处理后,采用PLS法的建模与预测效果最好,模型相关系数r为0.898,均方根校正误差(RMSEC)为1.72%,均方根预测误差(RMSEP)为2.34%,明显优于采用原始光谱、一阶微分光谱、二阶微分光谱及二阶微分Norris滤波光谱的建模与预测结果。

To solve the problem that there is no fast quality examination method,which influence potato industry development, nonde- structive and quick determination of dry matter content of potato was studied using visible - NIR diffuse reflectance spectroscopy. The spectra pre -processing methods including fist derivative, second derivative and Norris derivative filtering were applied to eliminate the effect of interference signal on spectra data. Principal component regression（PCR） and partial least square （PLS）methods were used to develop calibration models. The prediction set was used to evaluate the predictive ability of the models. 131 samples experimental re- suhs showed that the best model was obtained by PLS with the first derivative spectra processed by the Norris filtering method,that the segment length was 17 points and the gap between segment was 4 points. The correlation coefficientr, root mean square errors of calibra- tion （ RMSEC）, and root mean square errors of prediction （RMSEP） were 0. 898,1.72% and 2.34%, respectively. The prediction per- formance is better than the models established by the original spectra, the first derivative spectra,the second derivative spectra, and the second derivative spectra processed by the Norris filtering method. It is possible that measuring drymatter content of potato use visible - NIR spectrometric technique nondestructively.