铜污染土壤光谱特征及弱差信息的LH-PSD甄测模型

土壤受重金属污染后,会影响农作物及食品安全,危及人体健康,因此寻找快速、高效甄测土壤重金属污染信息的方法尤为关键。传统化学分析方法存在过程繁杂、费时耗力等缺点,而高光谱遥感因光谱分辨率高、信息量大、快速无损等特点在环境监测等应用方面优势明显。由于电磁遥感信号反射、辐射过程复杂,通过仪器获取的土壤高光谱数据难以直接解析出重金属污染信息,因而,研究并寻求一种能够有效挖掘土壤重金属污染信息的方法对高光谱遥感监测污染意义重大。不同浓度铜(Cu)污染会使土壤理化性质改变,引发土壤光谱产生微弱变化,该研究目的是对Cu污染土壤光谱的特征及弱差信息进行识别、提取与分析,进而挖掘光谱中的重金属污染信息。采用包络线去除(CR)对光谱进行预处理,通过定义局部极大值均值(LMM)与半波高(HWH),结合时频分析的短时傅里叶变换(STFT)及能量谱密度(PSD),构建LH-PSD甄测模型。通过模型对极相似土壤光谱进行处理,所获PSD分布图使光谱间的微弱差异可视化显现,并显著区分了相似光谱,验证了模型对光谱特征及弱差信息的甄别能力和有效性。同时应用该模型,对不同Cu污染梯度的土壤实验光谱进行重金属污染信息的提取与分析,研究结果表明,LH-PSD甄测模型中,LMM与HWH可有效提取光谱间差异特征并以阶梯状显现。经模型处理后得到的可视化PSD分布图能直观定性判别土壤是否受重金属Cu污染,即当土壤受重金属Cu污染后,相同采样频数下,在频率为100与600 Hz附近PSD分布会出现明显空缺分离,随着Cu污染浓度的增加,在100~600 Hz之间PSD的分布呈逐渐稀疏态势。能量值E可定量化监测土壤Cu污染程度,即随着土壤中Cu污染浓度的增加,E值呈下降趋势,且与Cu含量的相关系数达到-0.910 5,显著相关。为检验模型的可靠性,研究结合栽种玉米作物的土壤光谱,经LH-PSD甄测模型对其进行分析,所得可视化的PSD分布图结果与实验分析中基本一致,且能量值E的监测结果与土壤中Cu含量相关系数达到-0.973 9,相关性显著,验证了模型的可靠性。因此,LH-PSD甄测模型实现了对土壤光谱从光谱域到时频域的甄析,为深度挖掘重金属污染的光谱特征及弱差信息提供一种新思路。

The LH-PSD Analysis Model of Cu Contaminated Soil Spectral Characteristics and Weak Characteristic Information

Soil environmental safety is of great significance. When soil is contaminated by heavy metals, it will affect the safety of crops and foods and endanger human health. Therefore, it is particularly critical to look for ways to rapidly and efficiently measure heavy metal pollution in soil. Traditional chemical analysis methods have some disadvantages such as complicated process, time-consuming and labor-consuming. Hyperspectral remote sensing has obvious advantages in environmental monitoring and other applications because of its high spectral resolution, large amount of information, and rapid losslessness. Due to the complex reflection and radiation process of electromagnetic remote sensing signals, the soil hyperspectral data acquired by the instrument is difficult to directly interpret the information of heavy metal pollution. Therefore, it is very important to find out a method that can effectively excavate heavy metal pollution information in soils. The soil physicochemical properties will change because of different concentrations pollution of Cu, causing slight changes in the soil spectrum, the purpose of this study is to identify, extract and analyze the characteristics and weak difference information in the spectrum of Cu contaminated soil, and then tap the heavy metal pollution information in the spectra. In this paper, the continuum removal(CR) was used to preprocess the spectrum, the LH-PSD analysis model for analyzing soil spectra was constructed bydefining local maximum mean (LMM) and half wave height(HWH), combined with Short-time Fourier transform (STFT) of the time-frequency analysis method and power spectral density (PSD). The extremely similar soil spectrum was processed by the LH-PSD model, and the PSD distribution map visualized the faint differences between the spectra, and significantly distinguished the similar spectra, which verified the ability of the model to discriminate spectral features and weak differential information. At the same time, this model was used to extract and analyze heavy metal pollution information from experimental soil spectra with different Cu pollution gradients. The results of the study show that CR can plan the spectrum to the same background and highlight the differences between spectra, LMM and HWH of LH-PSD detection model can effectively extract the characteristics of the difference between the spectra and appear in a ladder. The visualized PSD map obtained after the model processing can directly and qualitatively discriminate whether the soil is contaminated by heavy metal Cu. Specifically, when the soil is contaminated by heavy metal Cu, at the same sampling frequency, the PSD distribution at frequencies of 100 and 600 Hz will be obviousvacant separation, with the increase of Cu pollution concentration, the distribution of PSD between 100~600 Hz is gradually sparse. The energy value E can be used to quantitatively monitor the degree of soil Cu pollution. That is, as the concentration of Cu in the soil increases, the E value decreases, and the correlation coefficient with the Cu content reaches -0.910 5, which is significantly correlated. In order to test the reliability of the model, the soil spectra of the planted corn crop was combined and analyzed by the LH-PSD detection model. The result of the visualized PSD map was basically similar to that in the experimental analysis, The correlation coefficient of the energy value E with Cu content in soil reaches -0.973 9, which has a significant correlation. The monitoring effect is ideal and the reliability of the model is verified. Therefore, through the LH-PSD analysis model, the dissection of soil spectrum from the spectral domain to the time-frequency domain provides a new idea for deepening the spectral features and weak information of heavy metal pollution spectra.