为实现苹果可溶性固形物的便携式快速检测, 搭建了以STS光谱仪和自制样品杯作为光谱检测装置的苹果可溶性固形物便携式检测平台。 采用自行设计的检测平台采集了苹果的近红外漫反射光谱, 对比分析了不同的光照角度、 光源与探头距离对光谱响应特性的影响, 建立了苹果可溶性固形物偏最小二乘模型(PLS)和最小二乘支持向量机模型(LS-SVM), 采用连续投影算法及主成分分分析法对最小二乘支持向量机模型进行了优化, 并对比分析了两种检测模型的优劣。 其中当光源距探头距离为15 mm光源角度为45°时, 结合偏最小二乘法建立苹果的可溶性固形物定量检测模型精度最高。 模型的预测集相关系数为0.924, 预测均方根误差为0.334%。 实验结果表明, 采用四周照射、 底部接收并结合避光圈的这种结构布置能够有效的克服杂散光现象并且提高了光谱中的有效信息。 研究可为快速、 便携的苹果可溶性固形物检测仪器的设计提供参考依据和理论支撑。

In order to detect soluble solids content(SSC) in fruit conveniently and rapidly, a portable soluble solids content spectrometer for apple was designed based on STS spectrometer. The NIR spectral data were obtained by the portable fruit soluble solids content spectrometer and recorded with the integration time of 100 ms in the wavelength range of 630~1 125 nm. Meanwhile, two structural parameters light source angle α and distance between light source and probe W were analyzed for investigating the influence of the response properties of visible-NIR spectra. The partial least square regression model and least squares support vector machine model were established. By comparison, partial least square regression model performed better. When the distance between probe and light source was 15 mm and the angle of light source was 45, its performance was the best with the root mean square error(RMSEP) of prediction set of 0.334% and correlation coefficient of prediction set of 0.924.