基于高光谱图像技术的罂粟种子活性无损检测

Research on Non-destructive Detection of Papaver somniferum Seed Activity Based on Hyperspectral Images Technology

罂粟种子是否具有活性是打击非法持有毒品原植物种子等相关犯罪的主要法庭证据。为实现罂粟种子活性的快速无损检测,本研究首次采用高光谱图像技术结合种子发芽率建立罂粟种子活性的预测模型,分别采用偏最小二乘回归法(PLSR)、支持向量回归法(SVR)和极限学习机法(ELM)建立预测模型。结果表明,全波段预测时,采用ELM建立的预测模型性能最佳,R_(p)为0.9204;在使用特征波段光谱数据建立预测模型时,相较于全波段模型,在模型性能上均有所提升,其中采用ELM在连续投影算法(SPA)选取的特征波段建立的预测模型的性能最佳,R_(p)为0.9285。表明采用高光谱图像技术检测罂粟种子活性的方法可行,该方法可以无损、快速地检测罂粟种子活性,为打击非法种植毒品原植物的相关犯罪活动提供技术支持。

The viability of Papaver somniferum seeds is the main forensic evidence against crimes related to the illegal possession of drug seed.In order to achieve rapid non-destructive testing of Papaver somniferum seed activity,hyperspectral image technology combined with seed germination rate was used for the first time in this study to establish the prediction model of Papaver somniferum seed activity,and partial least square regression(PLSR),support vector regression(SVR)and extreme learning machine(ELM)were used to establish the prediction model,respectively.The results showed that the prediction model based on ELM had the best performance and R_(p)was 0.9204.Compared with the full-band model,the performance of the prediction model was improved by using the spectral data of the feature band selected by ELM in the continuous projection algorithm(SPA).The performance of the prediction model established by ELM in the feature band selected by SPA was the best,and the R_(p)was 0.9285.The results showed that the method of detecting poppy seed activity by hyperspectral image was feasible.The method could detect poppy seed activity nondestructively and quickly,and provide technical support for combating the related criminal activities of illegal cultivation of drug plants.