多光谱融合的海洋沉积物碳含量检测

Carbon Content Detection of Marine Sediments Based on Multispectral Fusion

海洋沉积物中碳的变化是衔接海洋生态系统的过去与未来的信息桥梁,揭示了海洋生态过程变化规律。因此开展海洋沉积物碳含量的研究,对掌握海洋生态系统碳循环规律,研究全球碳循环,研究对气候变化的响应和反馈有着重要的作用。光谱技术是一种快速、无损的测量方法,在定量分析中已有很成熟的应用。多光谱融合通过将多个光谱数据结合一起,获得比单一光谱更丰富的信息,有利于物质的分析研究。将多光谱融合应用于海洋沉积物碳含量的研究,以青岛海洋潮间161份沉积物为样品,分别采用海洋光学QE65000光谱仪(光谱仪1)和AVANTES光纤光谱仪AvaSpec-ULS2048(光谱仪2)采集沉积物可见-近红外光谱。将两种光谱仪的光谱进行多光谱融合,分别采用偏最小二乘回归算法(PLSR)和BP神经网络算法(BPNN)建立沉积物碳含量模型。在PLSR沉积物碳含量建模结果中,多融合光谱结果优于光谱仪2,略低于光谱仪1,RPD值为1.968;在BPNN沉积物碳含量建模结果中,多融合光谱结果优于两个单光谱仪,RPD值为2.235。将多光谱融合后的光谱划分多个波段,分别建立沉积物碳含量模型,寻找沉积物碳的特征波段。通过分析多光谱融合各波段模型结果,560~790 nm的建模效果最好,R_(c)^(2)为0.949,RMSEC为0.550,R_(p)^(2)为0.874,RMSEP为0.733,RPD值为2.823。预测效果相较于光谱仪1、光谱仪2、多光谱融合全波段都有了显著的提高。因此采用多融合光谱特征波段建立海洋沉积物碳含量模型,能够提高海洋沉积物碳含量的预测结果,建立准确度更高的沉积物碳模型,为沉积物碳的快速测定打下基础。

The change of carbon in marine sediment is the information bridge between the past and the future of the marine ecosystem,which reveals the law of the marine ecological process.Therefore,the research on the carbon content of marine sediments plays an important role in mastering the carbon cycle law of the marine ecosystems,studying the global carbon cycle,and studying the response and feedback to climate change.Spectrum technology is a fast and non-destructive measurement method,which has been widely used in quantitative analysis.Multispectral fusion based on spectral technology,through the combination of multiple spectral data,we can get more information than a single spectrum,which is conducive to the analysis of substances.In this paper,multispectral fusion was applied to the study of carbon content in marine sediments.161 samples of Qingdao,China intertidal sediments are taken as samples.The visible near infrared spectra of sediments were collected by QE65000 spectrometer(spectrometer 1)and AVANTES optical fiber spectrometer(AvaSpec-ULS2048)(spectrometer 2),respectively.The spectra of the two spectrometers were fused,PLSR and BPNN were used to establish the carbon content model.The results of PLSR modeling showed that the results of multi fusion spectrum were better than that of spectrometer 2,slightly lower than that of spectrometer 1,and the RPD value was 1.968.The results of BPNN modeling showed that the results of multi fusion spectrum were better than those of two single spectrometers,and the RPD value was 2.235.The spectra after multispectral fusion were divided into several bands to find the characteristic bands of carbon in sediments.By analyzing the results of the multispectral fusion model,560~790 nm was the best,the R_(C)^(2) was 0.949,the RMSEC was 0.550,the R_(P)^(2) was 0.874,the RMSEP was 0.733,and the RPD was 2.823.Compared with spectrometer 1,spectrometer 2 and multispectral fusion,the prediction effect was significantly improved.Therefore,using the multi fusion spectral characteristic band to establish the model of the carbon content of marine sediments can improve the prediction results of the carbon content in marine sediments.It can establish a more accurate model of the organic carbon content of sediments,which will lay a foundation for the rapid determination of the organic carbon content of sediments.