基于高光谱成像技术的板栗产地溯源

Origin tracing of Castanea mollissima based on hyperspectral imaging technology

中文关键词:  高光谱成像  支持向量机  遗传算法  板栗  产地溯源

英文关键词:hyperspectral imaging  support vector machine  genetic algorithm  Castanea mollissima  origin tracing

中文摘要:

      目的 建立了一种基于高光谱成像(hyperspectral imaging, HSI)技术的板栗产地溯源模型。方法 采集怀柔、迁西和沂蒙短枝3种不同产地板栗的高光谱图像, 提取感兴趣区域, 建立支持向量机(support vector machine, SVM)板栗产地溯源模型, 通过比较分析不同预处理方法对建模结果的影响, 选出最佳的预处理组合方法, 并使用遗传算法(genetic algorithm, GA)对模型进一步优化。结果 实验结果表明, 经多元散射校正(multiplicative scatter correction, MSC)和移动窗口平滑法(moving window smoothing, MWS)组合预处理后的数据所建立的溯源模型预测性能最好, 分类的预测精确率达到了95%以上, 模型整体的预测准确率为96.61%。经GA对SVM的参数C进行优化, 优化后的模型对怀柔板栗和沂蒙短枝板栗的预测精确率达到了100%, 模型整体的准确率提高到了98.31%。结论 本研究基于高光谱成像技术建立了一种板栗产地溯源模型, 经预处理和参数优化后, 所建立的模型具有较好的预测性能, 为板栗的产地溯源提供了一种新方法。

英文摘要:

      Objective To establish a model of Castanea mollissima origin tracing based on hyperspectral imaging (HSI) technology. Methods Hyperspectral images of 3 kinds of different origin Castanea mollissima from Huairou, Qianxi and Yimeng were collected, region of interest was extracted, and support vector machine (SVM) tracing model of Castanea mollissima origin was established. By comparing and analyzing the influence of different pretreatment methods on the modeling results, the best pretreatment combination method was selected, and the genetic algorithm (GA) was used to further optimize the model. Results The experimental results showed that the traceability model established by the combination of multiplicative scatter correction (MSC) and moving window smoothing (MWS) had the best prediction performance. The prediction accuracy of classification reached more than 95%, and the overall accuracy of the model was 96.61%. The GA was used to optimize the parameter C of SVM, and the prediction accuracy of the optimized model reached 100%, and the overall accuracy of the model increased to 98.31%. Conclusion In this study, based on hyperspectral imaging technology, a Castanea mollissima origin tracing model is established. After preprocessing and parameter optimization, the established model has better prediction performance, which provides a new method for the origin tracing of Castanea mollissima.