中文摘要：

以喷洒不同浓度杜邦万灵的香梨作为研究对象,探讨了应用高光谱成像技术检测香梨表面农药残留的方法。运用376-1051nm高光谱成像系统采集200个香梨的高光谱图像,其中120个香梨为建模集,80个香梨为预测集。运用多元散射校正（MSC）对光谱数据进行预处理,然后采用连续投影算法（SPA）提取了11个特征波长。基于处理后的光谱数据,分别运用多元线性回归法（MLR）和主成分回归法（PCR）两种算法分别建立农药残留检测的模型,比较两种模型的结果。通过比较,采用MLR建立的农药残留检测模型效果较优,其校正集相关系数（Rc）为0.973,校正均方根误差（RMSEC）为0.260,预测的正确率可以达到91.7%,对较低浓度残留的预测正确率达到80%。研究表明,应用高光谱成像技术可以成功地检测香梨表面农药残留,并且对低浓度检测也有很好的效果。

英文摘要：

The hyperspectral imaging technology is used to detect pesticide residues on surface of Kuerle fragrant pear. The hyperspectral imaging system （376-1051nm） is applied to extract hyperspectral image of 200 pears, including 120 pears of calibration set and 80 pears of prediction set. In the study, multiplicative scatter correction （MSC）is used for pre-processing, and successive projections algorithm （SPA） is applied for the optimal wavelength selection in the calibra- tion set, and there are 11 optimal wavelengths. Based on spectral data, multiple linear regression （MLR） and principal component regression （PCR） are applied to set mathematical models between the relative reflectance of hyperspectral im- age and the concentration of pesticide residues, respectively. After comparison and contrast, the mathematical model based on MLR is better to detect the pesticide residues on surface of Kuerle fragrant pear, and the correlation coefficient calibration （Rc） and root mean square error of calibration （RMSEC） are 0. 973 and 0. 260, respectively. Moreover, MLR gives 91.7% detection rate, and for low concentration, MLR gives 80% detection rate. The results demonstrate that the hyperspectral imaging technique with the visible-near infrared region is a reliable method for the pesticide residues detec- tion, and has good effect for low concentration detection.