成像光谱技术在苹果内外部品质检测中的应用研究  

Application of Spectral Scattering Images on Detection of External and Internal Quality Parameters of Apples

作　　者：尚小燕[1] 路绍军[1] 孔英秀[1] 

机构地区：[1]西安工业大学光电工程学院,西安710032

出　　处：《内燃机与配件》2016年第10期144-147,共4页Internal Combustion Engine & Parts

基　　金：西安市科技计划项目(CXY1520(3))资助

摘　　要：采用光谱成像技术,对苹果的果径及糖度检测进行研究。通过一次性采集苹果不同波长的光谱图像,采用最小外接矩形法及多元线性回归法建立标定方程,对苹果果径进行检测,检测的正确率为97.76%。通过选取波长为633nm、649nm、669nm、766nm、850nm和905nm的苹果光谱图像,使用洛伦兹函数拟合感兴趣区域的图像灰度分布,将拟合所得参量与苹果的糖度进行逐步多元线性回归,建立633nm、649nm和669nm三波长最优组合的苹果糖度预测模型,校正集相关系数Rc=0.89,校正标准差SEC=0.491,验证集相关系数Rv=0.89,验证标准差为SEV=0.525。实验表明:利用光谱成像技术可同时进行苹果的内外部品质检测,为实现对苹果的综合品质在线检测提供依据。The spectral imaging technology was used to detect the apple diameter and sugar content in apple. The apple spectral images at different wavelength were captured at once. The apple diameter was measured on the basis of the calibration equation, which was established by the minimum enclosing rectangle method and the multiple linear regression. The detection accuracy was 97.76%. The apple spectral images at wavelength of 633 nm, 649 nm, 669 nm, 766 nm, 850 nm and 905 nm were chosen, and the image gray distribution in region of interest was fit by Lorentzian function with three parameters. The stepwise multiple linear regression model relating Lorentzian parameters to fruit sugar content was developed, and the sugar content prediction model of apple was built with the best combination in wavelength of633 nm, 649 nm and 669 nm. The correlation coefficient of calibration set, Rc =0.89; the standard error of calibration set, SEC =0.491; the correlation coefficient of validation set, Rv=0.89; the standard error of validation set, SEV=0.525. Results show that detection of external and internal quality parameters of apple could be performed simultaneously by use of the spectral imaging technology, and this could provide basis for realizing comprehensive quality on-line detection of apple.

关 键 词：光谱成像技术 苹果 果径 糖度 多元线性回归 

分 类 号：S661.1[农业科学—果树学] TP391.41[农业科学—园艺学]