论文部分内容阅读
基于高光谱技术研究竞争性自适应重加权算法(competitive adaptive reweighted sampling,CARS)结合极限学习机(Extreme learning machine,ELM)对油桃品种判别的可行性。本文利用高光谱成像技术选取油桃420~1000 nm的高光谱图像数据,经卷积平滑法(Savitzky-Golay smoothing,SG)、附加散射校正算法(Multiplicative Scatter correction,MSC)、基线校正(Baseline)、变量标准化算法(Standard Normalized Varite,SNV)等预处理方法处理原始数据,通过PLSR模型确定Baseline为最佳预处理方法。采用主成分分析法(Principal Component Analysis,PCA)、连续投影算法(Successive Projections Algorithm,SPA)与竞争性自适应重加权算法等提取的特征波长,建立偏最小二乘(Partial Least Square,PLS)和极限学习机鉴别模型进行比较研究。结果显示:基于CARS算法提取的特征波长构建的CARS-ELM和CARS-PLS模型性能最优。CARS-PLS预测集相关系数(RP)和均方根误差(RMSEP)分别为0.942和0.205;CARS-ELM的RP和RMSEP分别为0.931和0.119。说明CARS是一种有效的提取特征波长的方法,且ELM与PLS对模型的预测能力相当,可见利用高光谱图像技术结合CARS-ELM对油桃的品种判别是可行的。
Feasibility of discriminating nectarine varieties using competitive adaptive reweighted sampling (CARS) and extreme learning machine (ELM) based on hyperspectral technique. In this paper, hyperspectral imagery was used to select hyperspectral image data of 420 ~ 1000 nm in nectarine, which was analyzed by Savitzky-Golay smoothing (SG), Multiplicative Scatter Correction (MSC), Baseline , Standard Normalized Varite (SNV) and other pretreatment methods to process the raw data, and the Baseline is the best pretreatment method through the PLSR model. Partial Least Squares (PLS) and Partial Least Squares (PLS) were constructed by using principal component analysis (PCA), continuous projection algorithm (SPA) and competitive adaptive weighted algorithm Extreme Learning Machine identification model for comparative study. The results show that the CARS-ELM and CARS-PLS models constructed based on the characteristic wavelengths extracted by the CARS algorithm have the best performance. The correlation coefficient (RP) and root mean square error (RMSEP) of CARS-PLS prediction set were 0.942 and 0.205 respectively. The RP and RMSEP of CARS-ELM were 0.931 and 0.119 respectively. It shows that CARS is an effective method for extracting characteristic wavelength. And ELM and PLS have the same ability of predicting the model. It is feasible to use hyperspectral imagery and CARS-ELM to discriminate nectarines.