基于高光谱技术苹果硬度快速无损检测方法的建立

Establishment of rapid and non-destructive detection method of apple firmness using hyperspectral images

中文关键词:  苹果  高光谱图像  硬度  洛伦兹函数  偏最小二乘法

英文关键词:apple  hyperspectral imaging  firmness  Lorentzian function  partial least squares

中文摘要:

      目的 利用高光谱技术建立苹果内部品质无损检测的方法。方法 将高光谱图像的光谱信息和空间信息结合, 采用洛伦兹函数对苹果高光谱的空间散射曲线进行拟合, 提取拟合曲线的相关参数, 利用拟合参数对苹果硬度进行建模分析。结果 拟合曲线与原散射曲线的相关系数R达到0.99以上。分析比较多种统计建模方法对不同拟合参数的建模效果, 结果表明: 在524~1016 nm波段范围内, 利用偏最小二乘法(partial least squares, PLS)对拟合曲线的峰值建立硬度的预测模型, 校正集预测值与标准值的相关系数Rc=0.89, 校正集标准误差SEC=0.71×105 Pa, 验证集预测值与标准值的相关系数Rv=0.88, 验证集标准误差SEV=0.88×105 Pa。结论 利用高光谱散射成像技术, 采用偏最小二乘的方法对拟合峰值建模, 可以实现苹果硬度的快速无损检测。

英文摘要:

      Objective To establish a new method for apple firmness assessment by using the hyperspectral imaging technology. Methods Based on the spectral information and spatial information of hyperspectral images, Lorentzian distribution function was proposed to fit spectral scattering profiles at individual wavelengths, and fitting parameters was used for modeling analysis of the firmness of the apple. Results The average correlation coefficient R between fitting curve and scattering profile was greater than 0.99. Comparison of various image analysis and modeling methods with different fitting parameters of prediction results, partial least squares (PLS) model with fitting parameter had a good prediction of apple firmness at the wavelength of 400~ 1100 nm with Rc =0.89, SEC=0.71?105 Pa, and Rv =0.88, SEV=0.88?105 Pa, respectively. Conclusion Hyperspectral scattering imaging method is useful for rapid and non-destructive measurement of apple firmness.