传统外部品质分选目前已经无法满足人们对水果口感品质的需求，对传统外部品质分选线进行糖度分选改造，实现糖度分选，对确保水果的口感品质具有重要意义。分别采集两种不同检测方式下获取的脐橙的近红外漫反射光谱，其中环形发射与接收漫反射方式的光谱能量要比多点发射与接收漫反射方式强，波峰与波谷位置大致相同。近红外漫反射光谱经基线校正、多元散射校正、一阶、二阶导数等方法进行光谱数据预处理，以减少杂散光和噪声的影响。分别建立脐橙在两种不同漫反射检测方式下的糖度偏最小二乘（PLS）模型进行对比分析。实验结果表明，采用基线校正预处理方法获得的结果最优；环形发射与接收漫反射检测方式下的糖度模型预测相关系数为0.81，预测均方根误差为0.46°Brix，多点发射与接收漫反射检测方式下的糖度模型预测相关系数为0.76，预测均方根误差为0.53°Brix。研究表明，应用PLS建模结合近红外漫反射光谱对传统外部品质分选线中的糖度分选线功能进行升级改造是可行的。

Traditional quality sorting methods have been unable to meet people's increasing demands for fruit flavour and quality. Producers must therefore develop their traditional quality sorting methods to achieve sugar content sorting and ensure favourable flavour and quality. To address this, the near-infrared reflection spectra of navel oranges were collected separately through two different detection methods. The spectral energy of their ring transmission and diffuse reflection had to be stronger than that of the multi-point transmission and diffuse reflections. The positions of their peaks and troughs had to be approximately the same. The near-infrared diffuse reflectance spectra were preprocessed using baseline correction, multivariate scattering correction, first and second derivatives to reduce the influence of stray light and noise. A Partial Least Squares (PLS) model for the sugar content information that was collected through the two different reflection detection methods was established for their comparison and analysis. The experimental results show that the baseline correction preprocessing method produced the best results between the two methods. Its predicted correlation coefficient of sugar under ring transmission and diffuse reflection detection was 0.81 and its root mean square error was 0.46° Brix. The estimated correlation coefficient of the sugar content model using the multi-point transmission and diffuse reflection detection method was 0.76 and its root mean square error was 0.53° Brix. This research shows that it is feasible to use PLS modeling and near-infrared diffuse reflectance spectrum to upgrade the sugar content sorting methodology used on production lines.