中文摘要：

[目的]建立赣南脐橙颜色指标定量数学模型,探索用颜色进行水果分级的新方法。[方法]采用色差计来测量50个赣南脐橙样本的表面颜色,用近红外漫反射光谱并结合多元校正算法偏最小二乘法（PLS）,建立了赣南脐橙颜色指标L、a、b的定量模型。[结果]在全波段范围内,原始光谱所建模型最佳,其颜色指标L所建校正模型相关系数（r）为0.933,预测均方根偏差（RMSEP）为1.330,完全交互验证相关系数（rcross）达0.926;颜色指标a所建校正模型相关系数为0.970,预测均方根偏差为1.524,完全交互验证的相关系数达0.967;颜色指标b所建校正模型相关系数为0.893,预测均方根偏差为2.676,完全交互验证的相关系数达0.875。[结论]原始光谱所建模型最好,但其模型的校正均方根偏差和完全交互验证均方根偏差都偏高。

英文摘要：

[Objective] The quantitative mathematical model of color index in Navel Orange in the South of Jiangxi was set up and a new methods of classifying fruits by color was discussed.[Method] The surface colors in 50 of Navel Oranges samples were measured by using color reader.The quantitative models of the color indexs（L,a,b）was develop by the Near-infrared diffuse reflectance spectroscopy with partial least squares method.[Result] In whole wave band range,the optimum model was established by using original spectrum,the correlation coefficient（r）of the calibration model established by color index L was 0.933,the root mean square error of the prediction（RMSEP） was 1.333 and the correlation coefficient of full cross validation（rcross） was 0.926.The correlation coefficient of the calibration model established by color index a was 0.970,RMSEP was 1.524 and the correlation coefficient of full cross validation（rcross） was 0.967.The correlation coefficient of the calibration model established by color index b was 0.89 3,RMSEP was 2.676 and the correlation coefficient of full cross validation（rcross）was 0.875.[Conclusion] The model established by original spectrum was best,the root mean square error of full cross validation were both slightly higher.