小波变换和连续投影算法在火龙果总酸无损检测中的应用

发布时间：2017-12-27 04:03

本文关键词：小波变换和连续投影算法在火龙果总酸无损检测中的应用　出处：《光谱学与光谱分析》2016年05期 　论文类型：期刊论文

【摘要】：应用可见/近红外光谱技术、小波变换(WT)和连续投影算法(SPA),对火龙果总酸含量(TA)进行精确、快速的无损检测,为火龙果内部品质无损检测提供科学依据。利用Maya2000光纤光谱仪采集380~1 099nm范围的火龙果漫反射光谱数据,通过WT消噪、SPA优选波长和偏最小二乘回归(PLSR)分析方法,建立了火龙果总酸的定量预测模型。试验结果表明:经过WT消噪联合SPA优选波长压缩光谱变量后建立的WT-SPA-PLSR模型,预测精度都高于全谱PLSR模型。由全部样本的原始光谱变量作为输入变量建立PLSR模型的预测相关系数(R_p)为0.851 394,预测均方根误差(RMSEP)为0.086 848;全部样本的原始光谱数据使用dbN(N=2,3,…,10)小波进行分解消噪,其中消噪效果最优的是db4小波2层分解(db4-2),WT-PLSR模型的Rp为0.915 635,RMSEP为0.066 752,小波变换消噪后的光谱预测模型精度明显提高;原始光谱经过db10-3小波消噪联合SPA算法,从570个光谱变量中优选出530,545,604,626,648,676,685,695,730,897,972,1 016nm共12个变量作为输入变量,建立WT-SPA-PLSR预测模型,模型的RP为0.882 83,RMSEP为0.077 39。SPA算法适合火龙果TA模型的光谱变量选择,能够有效提取与总酸相关度高的波长变量,增加了预测模型的精度和稳定性。研究结果表明小波变换技术联合连续投影算法的漫反射近红外光谱无损检测火龙果总酸含量具有可行性。
[Abstract]:The visible / near infrared spectroscopy, wavelet transform (WT) and continuous projection algorithm (SPA) are applied to detect the total acid content (TA) of Pitaya accurately and quickly, so as to provide a scientific basis for pitaya's internal quality nondestructive testing. Using Maya2000 optical fiber spectrometer to collect pitaya diffuse reflectance data in 380~1 099nm range, we established a quantitative prediction model of Pitaya total acid by WT denoising, SPA preferred wavelength and partial least squares regression (PLSR) analysis method. The experimental results show that the prediction accuracy of the WT-SPA-PLSR model is higher than that of the full spectrum PLSR model after the WT denoising combined with the SPA optimization of the wavelength compression spectrum variable. The predicted correlation coefficient (R_p) of PLSR model was 0.851394 and the root mean square error (RMSEP) of all models was 0.086848. The original spectral data of all samples were dbN (N=2,3,...). 10), wavelet decomposition denoising, the denoising effect is the optimal DB4 wavelet decomposition level 2 (db4-2), WT-PLSR model Rp was 0.915635, RMSEP was 0.066752. The wavelet transform de noised spectral prediction model accuracy is obviously improved; the original spectra by db10-3 wavelet denoising combined SPA algorithm optimized 530545604626648676685695730897972,1 016nm total of 12 variables as input variables from 570 spectral variables, establish WT-SPA-PLSR prediction model, model RP was 0.88283, RMSEP was 0.07739. SPA algorithm is suitable for spectral variable selection of Pitaya TA model. It can effectively extract wavelength variables with high total acid correlation and increase the accuracy and stability of prediction models. The results show the feasibility of nondestructive detection of total acid content of Pitaya by wavelet transform combined with continuous projection method.
【作者单位】：华南农业大学电子工程学院;华南农业大学南方农业机械与装备关键技术教育部重点实验室;国家柑橘产业技术体系机械研究室;华南农业大学工程学院;
【基金】：国家现代农业产业技术体系建设专项项目(CARS-27) 教育部高等学校博士学科点专项科研基金项目(20124404120006)资助
【分类号】：O657.33;TS255.7
【正文快照】： 1.华南农业大学电子工程学院,广东广州5106422.华南农业大学南方农业机械与装备关键技术教育部重点实验室,广东广州5106423.国家柑橘产业技术体系机械研究室,广东广州5106424.华南农业大学工程学院,广东广州510642引言水果品质无损检测一直是农业工程领域重要的研究课题[1]。

【相似文献】