基于激光诱导击穿光谱的柑橘叶片黄龙病检测

Detecting Huanglongbing in citrus leaves based on laser induced breakdown spectroscopy

采用激光诱导击穿光谱(LIBS)结合化学计量学的方法对柑橘叶片黄龙病进行定性检测。试验结果显示:柑橘叶片中营养元素P(Ⅱ)、Mn(Ⅰ)、Si(Ⅰ)和Fe(Ⅰ)的LIBS信号强度与柑橘叶片的健康程度有直接关系,其中健康、中度感染黄龙病和重度感染黄龙病的柑橘叶片中P(Ⅱ)、Mn(Ⅰ)、Si(Ⅰ)和Fe(Ⅰ)的特征光谱强度呈依次减少的关系;然后再分别建立5个特征光谱以及采用光谱融合方法将5个特征光谱融合的偏最小二乘判别分析(partial least square discriminant analysis,PLS-DA)模型,并对其判别模型进行分析,其中Fe(Ⅰ)的RMSEC为0.394,Rc为0.871,总误判率为23.1%;RMSEP为0.454,Rp为0.841,总误判率为26.6%。光谱融合的RMSEC为0.341,Rc为0.905,总误判率为15.5%,RMSEP为0.395,Rp为0.867,总误判率为22.7%;利用归一化、多元散射校正(MSC)、标准正态变换(SNV)和正交信号校正(OSC)4种预处理方法对原始光谱进行预处理,并建立PLS-DA模型。研究结果表明,利用LIBS技术结合OSC光谱预处理和PLS-DA建模方法,模型的RMSEC为0.027,Rc为0.994,总误判率为0;RMSEP为0.023,Rp为0.995,总误判率为0,对3种类别的柑橘叶片能进行较好地分类。

The laser induced breakdown spectroscopy(LIBS)combined with chemometrics was used to qualitatively detect Huanglongbing in citrus leaves.The results showed that the LIBS signal intensities of elements P(Ⅱ),Mn(Ⅰ),Si(Ⅰ)and Fe(Ⅰ)in citrus leaves were directly related to the healthiness of citrus leaves,where the intensity of the characteristic peaks of P(Ⅱ),Mn(Ⅰ),Si(Ⅰ)and Fe(Ⅰ)in healthy,moderately and severely infected citrus leaves had a decreasing trend.Five characteristic spectrum were analyzed and fused together by spectral fusion to establish a partial least squares discriminant analysis(partial least square,PLS)model,in which the root mean square error(RMSEC)of the modeling set for Fe(Ⅰ)was 0.394,the correlation coefficient(Rc)of the modeling set was 0.871,and the total false positive rate was 23.1%.The prediction set mean square error(RMSEP)was 0.454,and the prediction set correlation coefficient(Rp)was 0.841,with an overall misspecification rate of 26.6%.The RMSEC for spectral fusion was 0.341 and Rc was 0.905,with an overall false positive rate of 15.5%,and the RMSEP was 0.395 and Rp was 0.867,with an overall false positive rate of 22.7%.Meanwhile,four pre-processing methods including normalization,multiplicative scatter correction(MSC),standard normal variate(SNV)and orthogonal signal correction(OSC)were used to reduce the effects of noise and errors on the spectra,and to establish the PLS model.Results of the LIBS technique combined with orthogonal signal correction(OSC)spectral pre-processing and partial least squares(PLS)modeling methods showed that the RMSEC was 0.027 and Rc was 0.994 with a total false positive rate of 0.The RMSEP was 0.023 and Rp was 0.995 with a total false positive rate of 0.The three categories of citrus leaves were better classified.It is indicated that the feasibility of using LIBS technology to detect nutrients in citrus leaves.It will provide a method for rapidly detecting Huanglongbing in citrus leaf.