| 摘要: |
| 目的 对套袋和不套袋苹果分别建立反射光谱与糖度预测模型，并对模型的精度进行比较分析，为构建苹果品质分级系统提供理论支撑。方法 采用美国ASD公司的便携式光谱仪和数显折光计分别测量套袋和不套袋烟富3号红富士苹果，以苹果赤道位置4个取样点的反射率光谱和对应位置的糖度为数据源，原始光谱经多元散射校正后，与糖度数据一同用偏最小二乘回归算法，分别建立套袋和不套袋苹果的反射率光谱糖度模型，进行糖度预测。结果 （1）套袋苹果校正集相关系数Rc=0.76，均方根误差RMSEP=0.837 5 Brix；预测集相关系数Rv=0.72，均方根误差RMSEP=0.870 2 Brix；（2）不套袋苹果校正集相关系数Rc=0.69，均方根误差RMSEP=0.904 0 Brix；预测集相关系数Rv=0.63，均方根误差RMSEP=0.913 4 Brix。结论 不套袋苹果的模型精度低于套袋苹果模型精度。相对复杂的表面情况导致不套袋苹果模型精度较差，不套袋苹果的无损检测误差会高于套袋苹果。 |
| 关键词: 苹果 套袋 不套袋 偏最小二乘回归 多元散射校正 糖度 无损检测 |
| DOI：10.12105/j.issn.1672-0423.20180409 |
| 分类号: |
| 基金项目:山东省农业科学院农业科技创新工程CXGC2017B04;中国农业科学院与山东省农业科学院科技创新工程协同创新任务CAAS-XTCX2018023山东省农业科学院农业科技创新工程（CXGC2017B04）；中国农业科学院与山东省农业科学院科技创新工程协同创新任务（CAAS-XTCX2018023） |
|
| Sugar content nondestructive testing study for apple |
|
Wang Fengyun1,Shen Yu1,2,Zhang Chen1,2,Liu Bingfu1,Zheng Jiye1
|
|
1.S&T Information Institute of Shandong Academy of Agricultural Sciences，Jinan 250100，China;2.College of Computer Science and Engineering，Shandong University of Science and Technology，Qingdao 266000，China
|
| Abstract: |
| Purposes Regarding the different sugar content in bagged and unbagged apples，reflectance spectra and sugar prediction models were established respectively. Moreover，the accuracy of these models were compared and analyzed，in order to provide theoretical support for the construction of apple quality classification system.Methods This research applied the ASD’s portable spectrometer and digital refractometer to measure the reflectance spectra of the same batch of Yanfu 3 Hao Red Fuji apples and the sugar content of the corresponding locations.Results After the original spectrum was corrected by multiple scatter，the partial least-squares regression algorithm was used together with the sugar content data to establish the reflectance spectra of the bagged and non-bagged apples-the sugar content model. The bagged apple model correction set correlation coefficient Rc is 0. 76，root mean square error RMSEP is 0. 837 5 Brix；prediction set correlation coefficient Rv is 0. 72，root mean square error RMSEP is 0. 870 2 Brix. Uncorked apple model correction set correlation coefficient Rc is 0. 69，root mean square error RMSEP=0. 904 0 Brix；prediction set correlation coefficient Rv is 0. 63，root mean square error RMSEP is 0. 913 4 Brix.Conclusion The relatively complex surface conditions lead to poorer accuracy of the non-bagged apple model. Non-bagged apples have a higher NDT error than bagged apples. |
| Key words: apple bagged non-bagged partial least square regression multiple scatter correction sugar content nondestructive testing |
