基于光谱技术的Bipls算法结合CARS算法的苹果可溶性固形物含量检测

可溶性固形物含量是判断苹果内部品质的重要参考属性之一。利用高光谱技术获取苹果感兴趣区域的反射光谱,以S-G平滑(Savitzky-Golay smoothing)和直接正交信号校正(Direct orthogonal signal correction, DOSC)算法对光谱数据进行梯度预处理后,用后向区间偏最小二乘法(Bipls)优选出3,5,6,7,8,9,13,14,15,16,17,18,19,20,21,23等16个子区间,共计177个波长。结合竞争自适应重加权采样算法(CARS)再作进一步筛选,提取出449.6,512.9,544.8,547.2,594.3,596.8,928.2 nm等7个特征波长,利用偏最小二乘算法(PLS)建立基于特征波长的可溶性固形物含量检测模型,所得模型评价为R_c=0.906 2,RMSEC为0.482 2,R_p=0.871 6,RMSEP为0.614 0。该算法模型预测性能同Bipls和Bipls-SPA模型相比更为优异,证明了Bipls结合CARS算法在提高苹果可溶性固体物含量检测精度方面的有效性。

Determination of Apple Soluble Solids Content Using Bipls Coupled with CARS Algorithm Based on Spectral Technology

Soluble solid content(SSC) is one of the important reference attributes for judging the internal quality of apples. The hyperspectral technique was used to obtain the reflectance spectrum of the region of interest of the apple, and using the Savitzky-Golay smoothing and direct orthogonal signal correction algorithms(DOSC) to perform gradient preprocessing on the spectral data.Backward interval partial least squares method(Bipls) prefers selecting 16 sub-intervals of 3, 5, 6, 7, 8, 9, 13, 14, 15, 16, 17, 18, 19, 20, 21, 23 for a total of 177 wavelengths. Combined with the competitive adaptive reweighted sampling algorithm(CARS) for further screening, 7 characteristic wavelengths such as 449.6, 512.9, 544.8, 547.2, 594.3, 596.8, 928.2 nm were extracted. Partial least squares algorithm was used to develop SSC determination model based on characteristic wavelengths. The model was evaluated as R_c=0.906 2, RMSEC of 0.482 2, R_p=0.871 6 and RMSEP of 0.614 0. The performance of the model with Bipls and Bipls-SPA is more excellent, the effectiveness of Bipls combined with CARS algorithm in improving the detection accuracy of apple soluble solids content was proved.