近红外光谱检测梨果硬度研究  

Study on pear firmness detection by using near infrared reflectance spectroscopy based on CARS

作　　者：王晓明[1] 章海亮[1] 罗微[1] 刘雪梅[1] 

机构地区：[1]华东交通大学,南昌市330013

出　　处：《中国农机化学报》2015年第6期120-123,142,共5页Journal of Chinese Agricultural Mechanization

基　　金：江西省科技支持项目(20142BDH80021)

摘　　要：应用近红外漫反射无损检测梨果硬度。采集240个梨样品的光谱数据,150个梨样品数据用于建模集来建立模型,另外90个样品数据用于预测集,评价模型的性能。采集完整梨果的近红外漫反射光谱（350~1 800nm）,为简化模型,原始光谱经小波算法去除噪声,然后采用CARS算法,波长由最初的1 451个降低为58个,采用偏最小二乘回归建立梨果硬度的定量预测数学模型。研究发现,CARS算法在简化模型同时,模型的预测评价指标优于原始光谱建模的模型,硬度定量预测数学模型的决定系数（Rp2）为0.66,均方根误差（RMSEP）为1.66。近红外漫反射光谱作为一种无损的检测方法,可用于评价梨果的硬度。The objective of the study was to non-destructively detect the firmness of pear fruit by using the near-infrared diffuse reflectance（NIR）spectra.A total of 240 pear fruit samples were used to develop the calibration and prediction models.In order to simply the calibration model,wavelet algorithm was used to remove noise and a total of 58 characteristic wavelengths were selected by using competitive adaptive reweighted sampling（CARS）algorithm and these 58 variables were used as input to partial least squares regression（PLSR）to quantitatively analyze firmness of pear fruit.The results showed that using CARS algorithm can simplify and improve the model greatly and the determination coefficients of the model（R2p）is 0.66,and root mean standard error of the model（RMSEP）is 1.66.The research indicates that NIR spectroscopy combined with CARS algorithm could provide an accurate,reliable and nondestructive method for assessing the internal quality index firmness of pear fruit.

关 键 词：近红外漫反射光谱 硬度 竞争性自适应重加权算法 梨 

分 类 号：O433.5[机械工程—光学工程]