基于LIBS技术和卷积神经网络的土壤铅含量等级快速分类

On fast classification of lead(Pb) levels in soil based on LIBS technology and convolutional neural network

将激光诱导击穿光谱(LIBS)技术与卷积神经网络(CNN)相结合用于土壤中铅(Pb)含量的分类研究,对5类不同污染程度的样本进行了分类试验。结果表明LIBS-CNN方法可以实现土壤中Pb质量浓度等级的快速准确分类,准确度达到99%以上。相较于常用的化学方法,LIBS技术可以原位快速地对待测样品进行检测,样品预处理简单。因此LIBS-CNN方法可以为后期土壤修复技术提供更加准确有效的数据,且节约大量的时间及成本,提高检测分辨率。

Laser-induced breakdown spectroscopy (LIBS) and convolutional neural network (CNN) were combined to classify soil samples contaminated with different levels of lead (Pb). Overall,five categories of soil samples were used for this experiment. The result indicated that the method of LIBS - CNN could be implemented to quickly classify the level of Pb in the soil samples,with accuracy of measurement over 99%. Compared to commonly used chemical methods, LIBS and CNN have achieved better yields and could be an alternative for in-situ fast classification of contaminated soils without complicated pretreatment. This alternative could facilitate later stage of soil remediation technology by helping to save time,reduce costs and improve the rate of detection resolution.