【摘要】 X射线荧光(XRF)光谱法作为重金属现场快速检测的重要技术，当用于土壤重金属检测时，由于土壤所含元素较多，XRF光谱中存在部分元素特征谱峰交叉重叠问题，影响重金属特征谱峰提取及定量分析准确性。针对该问题，提出一种基于小波峰锐化-高斯混合模型(WPS-GMM)的土壤XRF光谱重叠峰解析方法。该方法通过离散小波变换对重叠峰进行峰锐化，以增强信号中的重要局部特征，明确子峰峰位，并以其作为先验约束条件，对重叠峰构建高斯混合模型，进一步通过最大化似然估计模型参数，获得重叠峰中各子峰中心峰位、积分面积等相关信息，实现重叠峰的解析。将所建WPS-GMM方法用于土壤XRF光谱中Ni Kα-Co Kβ、 Cu Kα-Ni Kβ、 Zn Kα-Cu Kβ三组典型重叠峰解析及重叠峰中主峰对应重金属Ni、 Cu、 Zn的定量分析，并将结果与传统仅基于高斯混合模型(GMM)的重叠峰解析方法对比，验证所建方法的准确性。结果表明：与GMM方法相比，所建WPS-GMM方法对土壤XRF光谱三组重叠峰中Ni Kα、 Ni Kβ、 Cu Kα、 Cu Kβ和Zn Kα子峰峰位解析的准确率平均分别提高了77.55%、 47.03%、 52.65%、 22.07%和8.43%,积分面积解析准确率平均分别提高了74.05%、 80.17%、 61.62%、 28.29%和43.59%;用于Ni、 Cu、 Zn定量分析的准确率平均分别提高了73.23%、 68.47%、 47.62%;并且用于工业用土、农用土、建筑用土三种不同用途土壤中Ni、 Cu、 Zn准确定量分析均具有更好的普适性。因此所建WPS-GMM方法能够更准确获取不同土壤XRF光谱中重叠峰的子峰信息，更有利于提高土壤重金属XRF定量分析准确性。该研究为土壤重金属XRF现场快速准确检测提供了重要的方法基础。更多还原

【Abstract】 X-ray fluorescence(XRF) spectroscopy is an important technique for rapid and on-site detection of heavy metals. However, when it is used for detecting heavy metals in soil, due to the variety of elements contained in soil, there are some overlapping peaks in the soil XRF spectrum, which affects the accuracy of characteristic peak extraction and quantitative analysis of heavy metals. For this problem, this paper proposed a method for resolving the overlapping peaks in soil XRF spectra based on wavelet peak sharpening and Gaussian mixture model(WPS-GMM). This method first sharpened the overlapping peaks by a discrete wavelet transform to enhance the important local features of the spectral signal and clarify the peak positions of subpeaks. Then, using it as a prior constraint, a Gaussian mixture model of the overlapping peak was constructed. Finally, the relevant information for each subpeak in the overlapping peak was obtained by maximizing the likelihood to estimate the model parameters, thereby achieving the resolution of the overlapping peak. The established WPS-GMM method was applied to the resolution of typical overlapping peaks of Ni Kα-Co Kβ, Cu Kα-Ni Kβ, and Zn Kα-Cu Kβ in soil XRF spectra, as well as the quantitative analysis of Ni, Cu, and Zn corresponding to the main peaks in the overlapping peaks, to verify it saccuracy by comparing with the traditional resolution method for overlapping peaks based solely on the Gaussian mixture model(GMM). The results showed that compared with the GMM method, when the established WPS-GMM method was used for resolving the three overlapping peaks in soil XRF spectra, for the sub peaks of Ni Kα, Ni Kβ, Cu Kα, Cu Kβ and Zn Kα, the accuracy of resolved peak position increased by an average of 77.55%, 47.03%, 52.65%, 22.07%, and 8.43%, respectively, and the accuracy of resolved integral area increased by an average of 74.05%, 80.17%, 61.62%, 28.29%, and 43.59%, respectively; Moreover, the accuracy of quantitative analysis of Ni, Cu and Zn increased by an average of 73.23%, 68.47% and 47.62%, respectively. The established method demonstrated better universality for the accurate quantitative analysis of Ni, Cu, and Zn in soils with three different uses, including industrial, agricultural, and construction, by resolving the overlapping peaks in XRF spectra. Therefore, the established WPS-GMM method can more accurately obtain sub-peak information of overlapping peaks in XRF spectra of different soils, which is more conducive to improving the accuracy of quantitative analysis of heavy metals in soils by XRF. This study will provide an important methodological foundation for the rapid and accurate on-site detection of heavy metals in soil using XRF spectroscopy.更多还原