基于高光谱成像技术的新疆冰糖心红富士苹果分级和糖度预测研究

对新疆冰糖心红富士苹果采用高光谱成像技术进行分级和糖度预测研究.在糖度预测分析中,使用正交试验设计方法确定影响预测效果的主要因素是预测回归方法、光谱预处理方法和波长合并,次要因素是光谱校正处理方法、数据类型和实测值归一化处理.提取平均光谱,经过白板校正,采用一阶微分光谱预处理,10个波长的光谱合并,基于多元线性回归方法建立苹果糖度的预测模型,其验证集苹果糖度的预测模型相关系数为0.911,预测均方根误差为0.76％ Brix,相对分析误差为2.44.在分级研究中,选择712 nm波长图像,Gamma灰度变换增强图像,大津算法阈值确定后分割图像,基于形态学处理剔除果梗区域,提取苹果分割后区域的面积、充实度、周长、平均灰度等特征,采用二次判别分析分级苹果,验证集苹果分级准确率达到89.5％.结果表明,高光谱图像技术既能够准确预测新疆冰糖心红富士苹果糖度品质,也可以用于基于外部品质特征的分级研究.

Prediction of the Sugar Degree and Grading of Xinjiang Fuji Apple by Hyper-Spectral Imaging Techniques

The study was conducted to investigats the prediction of the sugar degree and the grading of Xinjiang Fuji apple using hyper-spectral imaging technology.First,in the research of sugar degree prediction,the main factors impacting the prediction performance of the sugar degree were prediction regression method,pretreatment of spectra,wavelength combination.The secondary factors were correction treatment of spectra,datatypes and actual determination method.The mean spectra was obtained when correcting spectra,preprocessing of spectra using first order,the spectra of 10 wavelengths were combined.The prediction model of sugar degree of Xinjiang Fuji Apple were built based on multiple linear regression method.The related coefficients of prediction model of apple sugar degree collected by the test,were 0.911,the predicted mean square error was 0.76% Brix,the opposite analysis error was 2.44,in the research of apple grading,712 nm wave image were collected,which were strengthened by the changeof Gamma grey,the images were separated when the algorithm threshold value were determined.The fruit stalk area were rejected based on morphological techniques.The characteristics,substantical conditions,circumference,average grey of the region in separated apple,were obtained.The apples were secondarily graded,the accuracy rate of grading apples were 89.5%.The results showed that the hyper-spectral image techniques both can accu-rately predict sugar degree quality of Xinjiang Fuji Apple and be used to study apple based on its external quality characteristics.