摘要
基于分段光谱特征值提取法和小波变换算法等多个数据预处理方法,分别针对分段基线差异及光谱噪声等严重影响激光诱导击穿光谱(LIBS)信号质量的主要影响因素,开展光谱信号预处理研究。基于实验室LIBS实验装置,通过实验验证,基于多通道光谱仪不同波段光谱特征值提取,提出了一种简单易行的多组数据中特征值点连接的方法,有效地提高了LIBS光谱信号的基线平直度,并得出以小波变换算法进行LIBS谱线信号去噪的最佳算法参数。在上述工作的基础上,使用基于误差反向传播的人工神经网络方法,实现了纯铜和不锈钢等物质种类的有效识别,研究结果表明,综合利用多数据处理方法进行LIBS技术中光谱信号处理可以有效提高谱线分析和识别的质量。
Based on multiple signal process methods, such as segmented spectral feature extraction and wavelet transform algorithm, the pre﹣spectrum signal treatment technique was investigated to decrease the difference of segmented spectral baseline and lower the spectral noise, and thus the signal quality in laser﹣induced breakdown spectroscopy （LIBS） was improved. Based on extracting the characteristic value in different spectral bands of multi﹣channel spectrometers, a simple method was presented to connect the characteristic value in different segmented data and effectively flats the signal baseline. Through analyzing the experimental data, the wavelet transform was used to lower the noise and obtain the optimum parameters. On the basis of the above work, artificial neural network based error back propagation was adopted to identify spectral line of the copper and stainless steel sample successfully. All the results illustrate that the utilization of multiple data processing method for spectral signal processing in LIBS technique can improve the quality of line＆#39;s analysis and recognition.
出处
《红外与激光工程》
EI
CSCD
北大核心
2014年第11期3807-3812,共6页
Infrared and Laser Engineering
基金
国家自然科学基金(61177082
61205074)
北京市自然科学基金(4122063)
关键词
激光诱导击穿光谱
基线校正
谱线识别
小波算法
人工神经网络
laser﹣induced breakdown spectroscopy
baseline correction
line＆#39
s recognition
wavelet transform
artificial neural network
