Loess soil lead pollution inversion method based on Gaofen-5 hyperspectral data
Through the processing of Gaofen No. 5 hyperspectral data, efficient, rapid and accurate identification of the distribution of illite minerals in lead-contaminated yellow cotton soil was achieved, solving the difficulties in identification in the existing technology, reducing environmental protection costs, and improving investigation efficiency.
Patent Information
- Application Number
- CN202410810077.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-21
- Publication Date
- 2024-11-12
AI Technical Summary
The prior art is difficult to effectively identify pollutant information in soil in mining geological environment surveys, especially the remote sensing spectral response of illite minerals in lead-contaminated yellow cotton soil lacks effective treatment methods, and the traditional method has a large workload, time and The investment is large and it is difficult to meet the needs of efficient monitoring.
Using the yellow mulberry soil lead pollution inversion method based on Gaofen 5 hyperspectral data, the hyperspectral data set of Gaofen 5 satellite was obtained, radiation calibration and atmospheric correction were performed, short-wave infrared bands were cut, and the Savitzky-Golay filtering and spectroscopy were used. Feature fitting method, matching and statistical analysis, output illite mineral distribution and lead element content value images to achieve efficient, fast and accurate pollutant identification.
Efficiently, quickly and accurately identify the distribution range of lead element pollution in illite in mining areas polluted by heavy metals, reducing environmental protection and soil restoration costs, solving the difficult problems in identifying heavy metal pollution, and improving the efficiency of mine geological environment surveys .
Smart Images

Figure CN118937280A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of remote sensing identification of geological mineral resources, and in particular to a loess lead pollution inversion method based on GF-5 hyperspectral data. Background Art
[0002] The current mine geological environment survey mainly relies on geochemical methods, including a large number of sample collection, sample preparation, pollutant testing and other links, which requires a lot of manpower, financial resources and material resources, but it is difficult to improve work efficiency. To this end, predecessors have proposed a large number of rapid mine geological environment survey methods based on remote sensing data (Landsat TM / ETM, WorldView-2 and ASTER data). These methods are mainly for open-pit mines, tailings ponds, ore dressing plants, spoil dumps and other mining facilities for human-computer interactive image interpretation and identification. These methods can effectively identify the pollution sources caused by mineral development, but the spatial distribution of pollutants is not well identified. These methods do not specifically propose effective processing methods for the remote sensing spectral response of characteristic minerals related to pollutants.
[0003] At present, the key to using remote sensing technology to detect the pollution status of mining areas lies in the detailed interpretation of the spatial distribution characteristics of pollution sources and the use of geographic information analysis technology and risk management methods to conduct pollution risk assessment. However, it is difficult to play a role in the spatial distribution characteristics of the mining pollutants themselves. At the same time, some characteristic minerals in the soil of mining areas also show changes in near-infrared key spectral bands with different degrees of pollution. The spectral changes of these minerals have characteristic spectral bands in the short-wave infrared band of the domestic high-resolution data of the fifth satellite and can be identified. For a long time, the field of remote sensing technology has been relatively unfamiliar with the detection of the spectral changes of pollutant indicator minerals in the field of mining geological environment, and there have been fewer attempts to remotely sense the illite minerals contaminated by lead in mining soils.
[0004] At present, the demand for minerals is increasing, and heavy metal pollution in the soil of mining areas caused by mining development is widespread and serious. The country pays more and more attention to the work of mine environmental protection and soil remediation. To carry out mine environmental protection and soil remediation, it is necessary to accurately and efficiently identify the main pollutants in the mining area and their distribution patterns. Traditional geochemical survey methods have accurate data and reliable results, but the workload of sample collection and analysis is large, and the time and cost investment are high, which makes it difficult to meet the needs of efficient monitoring. Therefore, there is an urgent need for a method that can carry out efficient, rapid and accurate identification of pollutant distribution patterns in mining areas threatened by heavy metal pollution.
[0005] Through the above analysis, the problems and defects of the prior art are as follows:
[0006] (1) Currently, there are few high-spatial-resolution data bands in multispectral data sources, and the spectral information is scarce, making it difficult to effectively identify pollutant information in the soil.
[0007] (2) The prior art does not propose an effective processing method specifically for the remote sensing spectral response related to illite minerals in lead-contaminated loess soil.
[0008] (3) The current remote sensing survey method is relatively effective for pollution sources such as mine spoil dumps, tailings ponds, and open-pit mines, but it is not very effective for the pollution status of heavy metal-contaminated soil. Summary of the invention
[0009] In order to overcome the problems existing in the above prior art, the purpose of the present invention is to provide a loess soil lead pollution inversion method based on the GF-5 hyperspectral data, which has the characteristics of efficient, rapid and accurate identification of the distribution of lead content in loess soil.
[0010] In order to achieve the above object, the technical solution adopted by the present invention is:
[0011] A method for inverting lead pollution in loess soil based on GF-5 hyperspectral data, comprising the following steps:
[0012] Step 1: Obtain the hyperspectral data set of the Gaofen-5 satellite, use the bias and gain data to perform radiometric calibration on the digital quantization value of the surface, and obtain the surface radiometric brightness value observed from the satellite perspective; perform atmospheric correction on the radiometric brightness image according to the atmospheric transmission model to obtain an image reflecting the surface reflectivity value;
[0013] Step 2: Clip the spectral bands of the 273rd to 317th bands of the atmospheric-corrected reflectance image to obtain a short-wave infrared hyperspectral dataset with a wavelength range of 2032 to 2403 nm; perform Savitzky-Golay filtering on the clipped GF-5 short-wave infrared hyperspectral dataset to reduce the noise of the spectral data;
[0014] Step 3: Select the standard spectrum of illite minerals in the USGS spectral library and cut it to the wavelength range of 2032-2403 nm as the reference spectrum; use the spectral feature fitting method to match the filtered and smoothed GF-5 short-wave infrared hyperspectral data set to obtain a matching value image;
[0015] Step 4: Perform statistical analysis on the matching value image, and generate a mask image from the range of the mean to the sum of the mean and the variance; perform mask processing on the short-wave infrared hyperspectral data set in the wavelength range of 2032 to 2403 nm after Savitzky-Golay filtering based on the mask image, and output the distribution section of illite minerals on the soil surface;
[0016] Step 5: Select the 2032-2403nm band as the peak search range, and perform first-order derivative operations on each pixel of the GF-5 short-wave infrared hyperspectral data set; select the wavelength value of the band with a derivative of 0 and store it in the position of the image pixel response, and output the spectral peak position image;
[0017] Step six, based on the relationship between the lead content value and the absorption peak position in the wavelength range of 2032-2403nm, calculate the lead content value of illite in each pixel of the spectral peak position image, and output the lead content value image.
[0018] Furthermore, in step 1, the GF-5 hyperspectral image is subjected to radiation calibration and atmospheric correction operations to obtain images reflecting the surface reflectance values, including:
[0019] (1) Obtain the hyperspectral data set of the Gaofen-5 satellite, use the bias and gain data in it to calibrate the digital quantization value of the surface, and obtain the surface radiation brightness value observed from the satellite perspective;
[0020] Among them, the radiation correction model is:
[0021] L = Gain × DN + Offset;
[0022] Where L is the radiation brightness value, Gain is the gain, Offset is the offset, and DN is the pixel brightness value;
[0023] (2) Perform atmospheric correction calculation on the radiance image according to the atmospheric transmission model;
[0024] Among them, the atmospheric correction model is:
[0025]
[0026] Where L is the total radiation brightness received by the sensor pixel; ρ is the surface reflectivity of the pixel; ρ e is the average surface reflectivity around the pixel; S is the albedo of the large balloon surface; L a is the atmospheric backscattering radiance; A and B are two coefficients that depend on atmospheric conditions and geometric conditions. Further, in step 2, the reflectance image after atmospheric correction is clipped in the spectral bands of bands 273 to 317, and the short-wave infrared hyperspectral data set with a wavelength range of 2032 to 2403 nm is obtained, including:
[0027] For the reflectance image spectral band clipping, among the 330 bands in the wavelength range of 390 to 2500 nm of the GF-5 hyperspectral image dataset, the spectral bands of bands 273 to 317 were selected for spectral clipping to obtain a shortwave infrared hyperspectral dataset in the wavelength range of 2032 to 2403 nm.
[0028] Furthermore, the Savitzky-Golay filtering algorithm formula in step 2 is:
[0029]
[0030] In the formula, h i / H is the smoothing coefficient, which is obtained by fitting the polynomial using the least squares method;
[0031] In the formula, x k is the kth data point in the original data sequence, w is the order of the filter, the least squares method is used to fit the local data segment of the kth data point, and the fitting function is used to estimate the value of each data point to achieve smoothing.
[0032] Further, step three specifically includes:
[0033] (1) Select the standard spectrum of illite mineral in the USGS spectral library and cut it to the wavelength range of 2032-2403 nm, and use it as the image reference spectrum;
[0034] (2) The processed GF-5 shortwave infrared hyperspectral dataset and the reference spectrum screened and processed in the USGS spectral library were subjected to envelope correction to normalize the data of both. The formula is as follows:
[0035]
[0036] In the formula, R Cj is the envelope removal value of band j, R j is the original spectral reflectance of band j, K is the slope between the starting point band and the end point band in the absorption curve, λ j is the jth band, R end and R stand is the original spectral reflectance at the starting point and the end point of the absorption curve, λ end and λ stand is the wavelength of the starting and ending points in the absorption curve;
[0037] (3) The spectral feature fitting method is used to match the GF-5 short-wave infrared hyperspectral data set after filtering, smoothing and envelope correction to obtain a matching value image; the curve shape of the reference spectrum is adjusted by comparing and increasing the constant k, thereby achieving pixel spectrum fitting matching; the larger the gray value of the fitting image, the higher the matching degree, and the greater the possibility of becoming the target mineral. The calculation formula is as follows:
[0038]
[0039] In the formula, is the transformed reference spectrum characteristic curve, p is the original spectrum characteristic curve of the reference spectrum, Both and p are processed by removing the envelope, k is a constant and k is not equal to -1.
[0040] Further, step four specifically includes:
[0041] (1) Performing statistical analysis on the matching value image, taking the range from the mean of pixel statistics to the sum of the mean and variance to generate a mask image;
[0042] (2) Based on the mask image, the GF-5 shortwave infrared hyperspectral dataset in the wavelength range of 2032-2403 nm after Savitzky-Golay filtering was masked to output the sections with purer distribution of illite minerals.
[0043] Further, step five specifically includes:
[0044] (1) Obtain the GF-5 short-wave infrared hyperspectral data set image that reflects the purer distribution of illite minerals after mask processing;
[0045] (2) Select the 2032-2403 nm band as the peak search range, perform first-order derivative operations on each pixel, select the wavelength value of the band with a derivative of 0 and store it in the position of the image pixel response, and output an image reflecting the spectral peak position. The calculation formula is as follows:
[0046]
[0047] Where Δλ=λ j -λ i ,λ j >λ i .
[0048] Furthermore, the relationship between the lead content value in step 6 and the absorption peak position in the wavelength range of 2032 to 2403 nm is as follows:
[0049] y = -217.68x + 1706.6;
[0050] Where y is the adsorption value of lead ion crystal layer; x is the position of hydroxyl (OH) absorption peak, unit: cm -1 .
[0051] Another object of the present invention is to provide a loess soil lead pollution inversion system based on GF-5 hyperspectral data using the loess soil lead pollution inversion method based on GF-5 hyperspectral data, and the illite lead content inversion system in loess soil comprises:
[0052] The hyperspectral data set acquisition module is used to acquire the hyperspectral data set of the Gaofen-5 satellite and perform radiation calibration and atmospheric correction operations to obtain an image reflecting the surface reflectance value;
[0053] The image clipping and filtering module is used to clip the spectral bands of images reflecting the surface reflectance values, obtain the GF-5 shortwave infrared hyperspectral data set and perform Savitzky-Golay filtering and smoothing;
[0054] Image matching module, used to determine the reference spectrum and match it with the filtered and smoothed GF-5 short-wave infrared hyperspectral data set, obtain the matching value image and perform statistical analysis;
[0055] The data mask processing module is used to perform mask processing on the filtered GF-5 short-wave infrared hyperspectral data set and output the section with purer distribution of illite minerals;
[0056] The lead content inversion module is used to perform first-order derivative operations on the short-wave infrared hyperspectral data set, output the spectral peak position image, calculate the lead content value of each pixel illite, and output the lead content value image.
[0057] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the loess lead pollution inversion method based on the GF-5 hyperspectral data.
[0058] Another object of the present invention is to provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the loess lead pollution inversion method based on the GF-5 hyperspectral data.
[0059] Another object of the present invention is to provide an information data processing terminal, which is used to implement the loess lead pollution inversion system based on the GF-5 hyperspectral data.
[0060] The present invention obtains the GF-5 hyperspectral image; the GF-5 remote sensing image is a hyperspectral satellite remote sensing image, which has 330 bands, especially the short-wave infrared part in the wavelength range of 2032 to 2403nm, which reflects the spectral characteristics of surface materials more accurately, and combines the relationship between the lead element content value and the absorption peak position in the wavelength range of 2032 to 2403nm, so as to carry out efficient, rapid and accurate identification and distribution of pollutants in mining areas threatened by heavy metal pollution.
[0061] The present invention is a remote sensing identification method specifically proposed for illite minerals, which can be applied to the analysis and identification of heavy metal pollution in mine geological environment surveys. Heavy metal pollution in the soil of mining areas caused by mining development in these areas is common and serious. The country pays more and more attention to the work of mine environmental protection and soil remediation. To carry out mine environmental protection and soil remediation work, it is necessary to accurately and efficiently identify the main pollutants in the mining area and their distribution patterns. The traditional geochemical survey method has accurate data and reliable results, but the workload of sample collection and analysis is large, and the time and financial investment are high, which is difficult to meet the needs of efficient monitoring. Therefore, the present invention is extremely important for reducing the cost of environmental protection and soil remediation in such heavy metal-contaminated areas. The inversion method provided by the present invention can efficiently, quickly and accurately identify the distribution range of illite contaminated by heavy metal lead in a relatively short time by obtaining the hyperspectral characteristics of clay minerals closely related to illite minerals, thereby solving the difficult problem of heavy metal pollution identification.
[0062] The present invention is based on the shortwave infrared hyperspectral spectrum of the Gaofen-5 satellite, and can efficiently, quickly and accurately identify the distribution of loess soil lead pollution associated with illite in the Loess Plateau area where loess soil is concentrated. BRIEF DESCRIPTION OF THE DRAWINGS
[0063] Figure 1 The present invention provides a flowchart of a method for inverting lead pollution in loess soil based on the GF-5 hyperspectral data.
[0064] Figure 2 It is a schematic diagram of the inversion image of lead pollution in loess soil based on the GF-5 hyperspectral data provided by an embodiment of the present invention.
[0065] Figure 3 This is the transportation location map of Tongguan County. DETAILED DESCRIPTION
[0066] The present invention will be further described in detail below in conjunction with the accompanying drawings.
[0067] like Figure 1 As shown, the loess soil lead pollution inversion method based on the GF-5 hyperspectral data provided by the embodiment of the present invention includes the following steps:
[0068] S101, obtaining a hyperspectral data set from the Gaofen-5 satellite and performing radiation calibration and atmospheric correction operations to obtain an image reflecting the surface reflectance value;
[0069] S102, performing spectral band clipping on the image reflecting the surface reflectance value to obtain the GF-5 shortwave infrared hyperspectral dataset and performing Savitzky-Golay filtering and smoothing processing;
[0070] S103, determining a reference spectrum and matching it with the filtered and smoothed GF-5 shortwave infrared hyperspectral data set to obtain a matching value image and perform statistical analysis;
[0071] S104, performing mask processing on the filtered GF-5 short-wave infrared hyperspectral data set, and outputting the section with relatively pure illite mineral distribution;
[0072] S105, performing a first-order derivative operation on the short-wave infrared hyperspectral data set to output a spectral peak position image; calculating the lead element content value of illite in each pixel of the spectral peak position image, and outputting a lead element content value image.
[0073] S106, select the GF-5 shortwave infrared hyperspectral data set in the 2032nm-2403nm band to perform first-order derivative operation on each pixel, select the wavelength value of the band with a derivative of 0 and store it in the position of the image pixel response, and output the spectral peak position image.
[0074] S107, based on the relationship between the lead content value and the Al-OH absorption peak position in the wavelength range of 2032nm to 2403nm, calculate the lead content value of illite in each pixel of the spectral peak position image, and output the lead content value image.
[0075] The schematic diagram of the inversion image of illite lead content provided by the embodiment of the present invention is as follows Figure 2 shown.
[0076] As a preferred embodiment, the loess soil lead pollution inversion method based on the GF-5 hyperspectral data provided by the embodiment of the present invention specifically includes the following steps:
[0077] S1: Obtain the hyperspectral data set of the GF-5 satellite, use the bias and gain data in it to calibrate the digital quantization value of the surface, and obtain the surface radiation brightness value observed from the satellite perspective. Then, perform atmospheric correction calculation on the radiation brightness image according to the atmospheric transmission model to obtain an image reflecting the surface reflectance value.
[0078] S2: The atmospheric-corrected reflectance image is subjected to spectral band clipping of bands 273 to 317 to obtain a shortwave infrared hyperspectral dataset in the wavelength range of 2032 to 2403 nm.
[0079] In the visible-near infrared range of soil, 600-800nm (16667-12500cm -1 ) band mainly reflects soil organic matter information, 400~560nm (25000~17857cm -1 ) band mainly reflects the information of iron and manganese oxides, 1900nm (5263cm -1 ) band mainly reflects water information, 2200nm (4545cm-1 ) are closely related to heavy metal content and clay minerals (illite, etc.) (Xie Xianli et al., 2007; Chen Yiyun, 2011). At present, the remote sensing field mainly focuses on the sub-pixel classification method of the research object (Shen Wenjuan et al., 2014), and the near-infrared spectral analysis field focuses on improving the heavy metal prediction optimization model of spectral reflectance (Meer, 2018). The displacement mechanism of soil near-infrared spectral bands under the influence of heavy metal ions is not closely connected with its basic research on hyperspectral remote sensing applications. Especially at 2200nm (4545cm -1 ) is mainly caused by Al-OH in clay minerals in the soil (Post et al., 2014), and illite in clay minerals is the main carrier for the adsorption of heavy metal ions (Law et al., 1966; Altaner et al., 1997; He Hongping et al., 1999; Liu Jingjing, 2014).
[0080] S3: Perform Savitzky-Golay filtering on the cropped GF-5 shortwave infrared hyperspectral dataset to reduce the noise of the spectral data.
[0081] S4: The standard spectrum of illite mineral is selected from the USGS spectral library and cropped to the wavelength range of 2032-2403 nm. It is used as the reference spectrum, and the spectral feature fitting method is used to match the filtered and smoothed GF-5 shortwave infrared hyperspectral dataset to obtain the matching value image.
[0082] S5: Perform statistical analysis on the matching value image, take the range from the mean to the sum of the mean and the variance to generate a mask image, and perform mask processing on the GF-5 shortwave infrared hyperspectral data set in the wavelength range of 2032 to 2403 nm after Savitzky-Golay filtering based on the mask image, and output the section with purer distribution of illite minerals.
[0083] S6: Select the 2032-2403 nm band as the peak search range, perform first-order derivative operations on each pixel of the GF-5 shortwave infrared hyperspectral data set, select the band wavelength value with a derivative of 0 and store it in the position of the image pixel response, and output the spectral peak position image.
[0084] S7: Based on the relationship between the lead content value and the Fe-OH absorption peak position in the wavelength range of 2032-2403 nm, the lead content value of illite in each pixel of the spectral peak position image is calculated, and the lead content value image is output.
[0085] The illite lead content inversion method based on the shortwave infrared hyperspectral of the Gaofen-5 satellite provided in the embodiment of the present invention can also be implemented by ordinary technicians in the industry using other steps. Figure 1The method for inverting lead pollution in loess soil based on the GF-5 hyperspectral data provided in the embodiment of the present invention is only a specific embodiment. The remote sensing satellite used in the embodiment of the present invention can be GF-5 or other hyperspectral remote sensing satellites.
[0086] The present invention uses remote sensing image processing software ENVI, ERDAS, PCI and the like as a digital image processing platform, and adopts an image processing scheme that is conducive to the identification of illite minerals to process the GF-5 hyperspectral image.
[0087] According to the characteristics of the GF-5 hyperspectral image, the spectral bands of bands 273 to 317 were selected for spectral clipping to obtain a short-wave infrared hyperspectral data set with a wavelength range of 2032 to 2403 nm. The short-wave infrared band contains rich information, which is conducive to the identification of illite minerals. The illite minerals were identified and masked on the image to obtain the distribution range of illite. The characteristic spectral band position was identified for the masked GF-5 short-wave infrared band hyperspectral image, and then the lead content value of illite in each pixel of the spectral peak position image was calculated based on the relationship between the lead content value and the absorption peak position in the wavelength range of 2032 to 2403 nm, and the lead-rich distribution of illite was determined based on the lead content value.
[0088] In step S1 provided in the embodiment of the present invention, a hyperspectral data set of the Gaofen-5 satellite is obtained, and the bias and gain data therein are used to calibrate the digital quantization value of the surface, and the surface radiation brightness value observed from the satellite perspective is obtained, and then the radiation brightness image is subjected to atmospheric correction operation according to the atmospheric transmission model to obtain an image reflecting the surface reflectance value, including:
[0089] (1) Obtain the hyperspectral data set of the Gaofen-5 satellite, use the bias and gain data in it to calibrate the digital quantization value of the surface, and obtain the surface radiation brightness value observed from the satellite perspective;
[0090] Among them, the radiation correction model is:
[0091] L = Gain × DN + Offset;
[0092] Where L is the radiation brightness value, Gain is the gain, Offset is the offset, and DN is the pixel brightness value.
[0093] (2) Perform atmospheric correction calculation on the radiance image according to the atmospheric transmission model;
[0094] Among them, the atmospheric correction model is:
[0095]
[0096] Where L is the total radiation brightness received by the sensor pixel; ρ is the surface reflectivity of the pixel; ρ e is the average surface reflectivity around the pixel; S is the albedo of the large balloon surface; L a is the atmospheric backscattered radiance; A and B are two coefficients that depend on atmospheric and geometric conditions.
[0097] In step S2 provided by the embodiment of the present invention, the reflectance image after atmospheric correction is subjected to spectral band clipping of the 273rd to 317th bands, and the short-wave infrared hyperspectral data set in the wavelength range of 2032 to 2403 nm is obtained, including:
[0098] For the reflectance image spectral band clipping, among the 330 bands in the wavelength range of 390 to 2500 nm of the GF-5 hyperspectral image dataset, the spectral bands of bands 273 to 317 were selected for spectral clipping to obtain a shortwave infrared hyperspectral dataset in the wavelength range of 2032 to 2403 nm.
[0099] In step S3 provided in the embodiment of the present invention, the Savitzky-Golay filtering algorithm formula is:
[0100]
[0101] In the formula, h i / H is the smoothing coefficient, which is obtained by fitting the polynomial using the least squares method.
[0102] In the formula, x k is the kth data point in the original data sequence, and w is the order of the filter. The algorithm uses the least squares method to fit the local data segment of the kth data point and uses the fitting function to estimate the value of each data point to achieve smoothing.
[0103] In step S4 provided in the embodiment of the present invention, the standard spectrum of illite mineral selected in the USGS spectral library is clipped to the wavelength range of 2032-2403 nm, and used as a reference spectrum. The spectral feature fitting method is used to match the filtered and smoothed GF-5 short-wave infrared hyperspectral data set to obtain a matching value image, which specifically includes:
[0104] (1) Select the standard spectrum of illite minerals from the USGS spectral library and cut it to the wavelength range of 2032nm to 2403nm, and use it as the image reference spectrum;
[0105] (2) The envelope correction process is performed on the GF-5 shortwave infrared hyperspectral dataset after the above processing and the reference spectrum screened and processed in the USGS spectral library to normalize the data of the two. The formula is as follows:
[0106]
[0107] In the formula, R Cj is the envelope removal value of band j, R j is the original spectral reflectance of band j, K is the slope between the starting point band and the end point band in the absorption curve, λ j is the jth band, R end and R stand is the original spectral reflectance at the starting point and the end point of the absorption curve, λ end and λ stand are the wavelengths of the starting and ending points in the absorption curve.
[0108] (3) The spectral feature fitting method is used to match the filtered, smoothed and envelope-corrected GF-5 short-wave infrared hyperspectral data set to obtain a matching value image. Generally, there is a certain difference in the characteristic absorption depth between the pixel spectrum after envelope correction and the reference spectrum. It is necessary to adjust the curve shape of the reference spectrum by adding a constant k through comparison to better fit the pixel spectrum. The larger the gray value of the fitting image, the higher the matching degree and the greater the possibility of becoming the target mineral. The calculation formula is as follows:
[0109]
[0110] In the formula, is the transformed reference spectrum characteristic curve, and p is the original spectrum characteristic curve of the reference spectrum (here and p are processed by envelope removal), k is a constant and k is not equal to -1.
[0111] In step S5 provided by the embodiment of the present invention, the above matching value image is statistically analyzed, and a mask image is generated from the range of the pixel statistical value mean to the sum of the mean and the variance. The GF-5 shortwave infrared hyperspectral data set in the wavelength range of 2032 to 2403 nm after Savitzky-Golay filtering is masked according to the mask image to output a section with a relatively pure distribution of illite minerals, specifically including:
[0112] (1) performing statistical analysis on the matching value image, and generating a mask image from the range of the pixel statistical value mean to the sum of the mean and the variance;
[0113] (2) Based on the mask image, the GF-5 shortwave infrared hyperspectral dataset in the wavelength range of 2032-2403 nm after Savitzky-Golay filtering was masked to output the sections with purer distribution of illite minerals.
[0114] In step S6 provided by the embodiment of the present invention, for the GF-5 short-wave infrared hyperspectral data set image reflecting the relatively pure section of illite mineral distribution obtained by the above mask processing, the 2032-2403nm band is selected as the peak search range, the first-order derivative operation is performed on each pixel, the wavelength value of the band with a derivative of 0 is selected and stored in the position of the image pixel response, and the image reflecting the spectral peak position is output, which specifically includes:
[0115] (1) The GF-5 short-wave infrared hyperspectral data set image reflecting the relatively pure distribution section of illite minerals obtained by the above mask processing;
[0116] (2) Select the 2032-2403 nm band as the peak search range, perform first-order derivative operations on each pixel, select the wavelength value of the band with a derivative of 0 and store it in the position of the image pixel response, and output an image reflecting the spectral peak position. The calculation formula is as follows:
[0117]
[0118] Where Δλ=λ j -λ i ,λ j >λ i .
[0119] In step S7 provided by the embodiment of the present invention, the lead content value of illite in each pixel of the spectral peak position image is calculated based on the relationship between the lead content value and the Fe-OH absorption peak position in the wavelength range of 2032 to 2403 nm, and the lead content value image is output, which specifically includes:
[0120] (1) The relationship between the lead content and the position of each OH absorption peak in the wavelength range of 2032-2403 nm is as follows:
[0121] y = -217.68x + 1706.6;
[0122] Where y is the lead content, x is the hydroxyl (OH) absorption peak position (cm -1 ).
[0123] (2) Calculate the lead content value of illite in each pixel of the spectral peak position image and output the lead content value image.
[0124] The Gaofen-5 satellite was officially put into use in March 2019. It is one of the most important remote sensing satellites in my country's high-spectral resolution earth observation system. The main sensors carried on the Gaofen-5 satellite include a visible short-wave infrared hyperspectral camera, a full-spectrum spectral imager, an atmospheric trace gas differential absorption spectrometer, an atmospheric main greenhouse gas monitor, an atmospheric aerosol multi-angle polarization detector, and an atmospheric environment infrared very high spectral resolution detector. Among them, the hyperspectral camera and the full-spectrum camera fill the gap in the domestic surface hyperspectral-multispectral comprehensive observation, and will play an important role in the fields of ecological environment, disaster monitoring, and resource investigation.
[0125] The loess soil lead pollution inversion system based on the GF-5 hyperspectral data provided by the embodiment of the present invention includes:
[0126] The hyperspectral data set acquisition module is used to acquire the hyperspectral data set of the Gaofen-5 satellite and perform radiation calibration and atmospheric correction operations to obtain an image reflecting the surface reflectance value;
[0127] The image clipping and filtering module is used to clip the spectral bands of images reflecting the surface reflectance values, obtain the GF-5 shortwave infrared hyperspectral data set and perform Savitzky-Golay filtering and smoothing;
[0128] Image matching module, used to determine the reference spectrum and match it with the filtered and smoothed GF-5 short-wave infrared hyperspectral data set, obtain the matching value image and perform statistical analysis;
[0129] The data mask processing module is used to perform mask processing on the filtered GF-5 short-wave infrared hyperspectral data set and output the section with purer distribution of illite minerals;
[0130] The lead content inversion module is used to perform first-order derivative operations on the short-wave infrared hyperspectral data set, output the spectral peak position image, calculate the lead content value of each pixel illite, and output the lead content value image.
[0131] Example:
[0132] like Figure 3 As shown in the figure, this experiment uses the gold mining area of Tongguan County as the experimental area. Tongguan County is located at the eastern end of the Guanzhong Plain, adjacent to Lingbao City, Henan Province in the east, Huayin City, Shaanxi Province in the west, Qinling Mountains and Luonan County, Shaanxi Province in the south, and the Yellow River and Weihe River in the north, facing Dali County, Shaanxi Province and Ruicheng County, Shanxi Province across the water. East longitude 110°09′30″~110°25′32″, north latitude 34°23′30″~34°39′00″, total area 526km 2 The Tongguan County gold mining area is located in the southern part of the county. The mining area is 18km from east to west and 8-10km from north to south, with a total area of 162km 2, accounting for 42% of the county's land area. Among them, the industrial reserves of gold mines exceed 100t, accompanied by silver and lead. In addition, there are non-metallic minerals such as iron ore, quartz, graphite, smelting crystal, marble, and diabase. Judging from the relative density of gold-quartz veins in the mining area, the main gold mining areas are distributed in the three mining areas of Dongtongyu Gold Mining Area, Matai (Yu) Gold Mining Area, and Xitongyu Gold Mining Area. The polymetallic ores in the mining area are mainly aluminum, followed by copper and zinc, and are also accompanied by silver. Lead, zinc, copper, silver and gold are the same mineralization series in Xiaozouling, Tongguan County, and have an intrinsic essential connection in the mineralization mechanism. The waste slag produced by the development of Tongguan gold mine has obvious regional distribution patterns. The southern mountainous area is mainly mining waste slag and a small amount of ore dressing tailings slag. While occupying land and farmland, mineral processing tailings contain high concentrations of Pb elements, which cause farmland pollution due to irrigation with mining wastewater and heavy rains. In winter and spring, strong winds and dust blown by the wind into villagers' courtyards and surrounding farmland, causing an increase in the heavy metal content in farmland soil.
[0133] The bias and gain data of the collected GF-5 hyperspectral image dataset were used to calibrate the digital quantization value of the surface, and the surface radiance value observed from the satellite perspective was obtained. Then, the radiance image was atmospherically corrected according to the atmospheric transmission model to obtain an image reflecting the surface reflectance value. The reflectance image after atmospheric correction was cropped in the spectral bands of the 273rd to 317th bands to obtain a short-wave infrared hyperspectral dataset with a wavelength range of 2032 to 2403nm. The cropped GF-5 short-wave infrared hyperspectral dataset was subjected to Savitzky-Golay filtering to reduce the noise of the spectral data. The standard spectrum of illite minerals was selected from the USGS spectral library and cropped to the wavelength range of 2032 to 2403nm. It was used as the reference spectrum, and the spectral feature fitting method was used to fit the filtered and smoothed GF-5 short-wave infrared hyperspectral dataset. The spectral data set is matched to obtain a matching value image; the matching value image is statistically analyzed, and a mask image is generated from the range of the mean to the sum of the mean and the variance, and the GF-5 short-wave infrared hyperspectral data set in the wavelength range of 2032 to 2403 nm after Savitzky-Golay filtering is masked according to the mask image, and the section with a relatively pure distribution of illite minerals is output; the 2032 to 2403 nm band is selected as the peak search range, and the first-order derivative operation is performed on each pixel of the GF-5 short-wave infrared hyperspectral data set, and the wavelength value of the band with a derivative of 0 is selected to store in the position of the image pixel response, and the spectral peak position image is output; according to the relationship between the lead content value and the Al-OH absorption peak position in the wavelength range of 2032 to 2403 nm, the heavy metal lead content value of each pixel of the spectral peak position image is calculated, and the lead content value distribution image is output. The present invention can efficiently, quickly and accurately identify the distribution range of illite contaminated by heavy metal lead.
[0134] Since the spectral resolution of the GF-5 hyperspectral image is high, it has up to 330 visible-near infrared bands, and illite minerals have characteristic absorption bands in the near-infrared range of 1900-2500nm. Studies have shown that the absorption peak at 2200nm (4545cm-1) is caused by Al-OH, and the nearby absorption bands should be attributed to the OH functional group. After processing, the GF-5 hyperspectral image can clearly reflect the distribution of illite minerals. The crystal structure changes of illite adsorbing heavy metal ions can be analyzed by mid-infrared spectroscopy (2500-25000nm) and X-ray diffraction. The position of the mid-infrared characteristic band of illite in the wavelength range of 2703-2778nm shifts from 2751nm to 2745nm as the amount of heavy metal ions adsorbed increases. At the same time, the interlayer charge distribution changes, causing the OH octahedral layer to deform and the interplanar spacing to increase. In general, the shift of the mid-infrared band of a mineral is also accompanied by a change in the half-height width parameter. Illite minerals have a characteristic absorption band vertex near 2250nm in the short-wave infrared range of 2032-2403nm, which helps to identify the iron and magnesium content information of illite minerals. Therefore, the technical solution provided by the present invention is to identify illite information based on the GF-5 hyperspectral image according to the USGS spectral library standard spectrum, detect the illite distribution area, and combine the lead content value with the Al-OH absorption peak position in the wavelength range of 2032-2403nm, and then calculate the lead element content value of illite in each pixel of the spectral peak position image, and output the lead element content value image, and then identify the lead-rich illite distribution area according to the lead element content value, and finally find the lead-rich illite through field investigation and verification, which not only avoids the problem of low spectral resolution, inaccurate anomaly identification, and mixed noise that seriously affects the extraction accuracy when using only multi-spectral remote sensing data such as Landsat and ASTER, but also can quickly, accurately and efficiently analyze the degree of heavy metal contamination of illite in the mine geological environment, serve the mine geological environment investigation work, improve work efficiency, and save funds invested by the state and society. Therefore, the method or system provided by the present invention can efficiently, quickly and accurately identify the distribution range of illite contaminated by heavy metal lead in a relatively short period of time, and has not been proposed by predecessors. The illite lead content inversion method provided by the present invention can fill the gap in the relevant field and has certain significance for guiding mine environmental protection work.
Claims
1. A method for inversion of loess lead pollution based on GF-5 hyperspectral data, characterized in that: The steps include: Step 1: Obtain the hyperspectral data set of the Gaofen-5 satellite, use the bias and gain data to perform radiometric calibration on the digital quantization value of the surface, and obtain the surface radiation brightness value observed from the satellite perspective; Perform atmospheric correction on the radiance image according to the atmospheric transmission model to obtain an image reflecting the surface reflectance value; Step 2: Clip the spectral bands of the 273rd to 317th bands of the atmospheric-corrected reflectance image to obtain a short-wave infrared hyperspectral dataset with a wavelength range of 2032 to 2403 nm; perform Savitzky-Golay filtering on the clipped GF-5 short-wave infrared hyperspectral dataset to reduce the noise of the spectral data; Step 3, select the standard spectrum of illite mineral in the USGS spectral library and cut it to the wavelength range of 2032-2403nm and use it as the reference spectrum; The spectral feature fitting method is used to match the filtered and smoothed GF-5 shortwave infrared hyperspectral data set to obtain the matching value image. Step 4: Perform statistical analysis on the matching value image, and generate a mask image from the range of the mean to the sum of the mean and the variance; perform mask processing on the short-wave infrared hyperspectral data set in the wavelength range of 2032 to 2403 nm after Savitzky-Golay filtering based on the mask image, and output the section with relatively pure illite mineral distribution; Step 5: Select the 2032-2403nm band as the peak search range, and perform first-order derivative operations on each pixel of the GF-5 short-wave infrared hyperspectral data set; select the wavelength value of the band with a derivative of 0 and store it in the position of the image pixel response, and output the spectral peak position image; Step six, based on the relationship between the lead content value and the absorption peak position in the wavelength range of 2032-2403nm, calculate the lead content value of illite in each pixel of the spectral peak position image, and output the lead content value image.
2. The method for inversion of loess lead pollution based on GF-5 hyperspectral data according to claim 1 is characterized in that: In step 1, the GF-5 hyperspectral image is subjected to radiometric calibration and atmospheric correction operations to obtain images reflecting the surface reflectance values, including: (1) Obtain the hyperspectral data set of the Gaofen-5 satellite, use the bias and gain data in it to calibrate the digital quantization value of the surface, and obtain the surface radiation brightness value observed from the satellite perspective; Among them, the radiation correction model is: L = Gain × DN + Offset; Where L is the radiation brightness value, Gain is the gain, Offset is the offset, and DN is the pixel brightness value; (2) Perform atmospheric correction calculation on the radiance image according to the atmospheric transmission model; Among them, the atmospheric correction model is: Where L is the total radiation brightness received by the sensor pixel; ρ is the surface reflectivity of the pixel; ρ e is the average surface reflectivity around the pixel; S is the albedo of the large balloon surface; L a is the atmospheric backscattered radiance; A and B are two coefficients that depend on atmospheric and geometric conditions.
3. The method for inversion of loess lead pollution based on GF-5 hyperspectral data according to claim 1 is characterized in that: In step 2, the reflectance image after atmospheric correction is clipped in the spectral bands of bands 273 to 317, and the short-wave infrared hyperspectral dataset with a wavelength range of 2032 to 2403 nm is obtained, including: For the reflectance image spectral band clipping, among the 330 bands in the wavelength range of 390 to 2500 nm of the GF-5 hyperspectral image dataset, the spectral bands of bands 273 to 317 were selected for spectral clipping to obtain a shortwave infrared hyperspectral dataset in the wavelength range of 2032 to 2403 nm.
4. The method for inversion of loess lead pollution based on GF-5 hyperspectral data according to claim 3 is characterized in that: The Savitzky-Golay filtering algorithm formula in step 2 is: In the formula, h i / H is the smoothing coefficient, which is obtained by fitting the polynomial using the least squares method; In the formula, x k is the kth data point in the original data sequence, and w is the order of the filter. The algorithm uses the least squares method to fit the local data segment of the kth data point and uses the fitting function to estimate the value of each data point to achieve smoothing.
5. According to claim 1, a method for inversion of loess lead pollution based on GF-5 hyperspectral data is characterized in that: Step three specifically includes: (1) Select the standard spectrum of illite mineral in the USGS spectral library and cut it to the wavelength range of 2032-2403 nm, and use it as the image reference spectrum; (2) The processed GF-5 shortwave infrared hyperspectral dataset and the reference spectrum screened and processed in the USGS spectral library were subjected to envelope correction to normalize the data of both. The formula is as follows: In the formula, R Cj is the envelope removal value of band j, R j is the original spectral reflectance of band j, K is the slope between the starting point band and the end point band in the absorption curve, λ j is the jth band, R end and R stand is the original spectral reflectance at the starting point and the end point of the absorption curve, λ end and λ stand is the wavelength of the starting and ending points in the absorption curve; (3) The spectral feature fitting method is used to match the GF-5 short-wave infrared hyperspectral data set after filtering, smoothing and envelope correction to obtain a matching value image; the curve shape of the reference spectrum is adjusted by comparing and increasing the constant k, thereby achieving pixel spectrum fitting matching; the larger the gray value of the fitting image, the higher the matching degree, and the greater the possibility of becoming the target mineral. The calculation formula is as follows: T = (p + k) / (1.0 + k); Wherein, T is the transformed reference spectrum characteristic curve, p is the original spectrum characteristic curve of the reference spectrum, both T and p are processed by envelope removal, k is a constant and k is not equal to -1.
6. According to claim 1, a method for inversion of loess lead pollution based on GF-5 hyperspectral data is characterized in that: Step 4 specifically includes: (1) Performing statistical analysis on the matching value image, taking the range from the mean of pixel statistics to the sum of the mean and variance to generate a mask image; (2) Based on the mask image, the GF-5 shortwave infrared hyperspectral dataset in the wavelength range of 2032-2403 nm after Savitzky-Golay filtering was masked to output the sections with purer distribution of illite minerals.
7. According to claim 1, a method for inversion of loess lead pollution based on GF-5 hyperspectral data is characterized in that: Step 5 specifically includes: (1) Obtain the GF-5 short-wave infrared hyperspectral data set image that reflects the purer distribution of illite minerals after mask processing; (2) Select the 2032-2403 nm band as the peak search range, perform first-order derivative operations on each pixel, select the wavelength value of the band with a derivative of 0 and store it in the position of the image pixel response, and output an image reflecting the spectral peak position. The calculation formula is as follows: In the formula, Δλ=λ j -l i ,l j >l i 。 8. According to claim 1, a method for inversion of loess lead pollution based on GF-5 hyperspectral data is characterized in that: The relationship between the lead content in step 6 and the absorption peak position in the wavelength range of 2032-2403nm is as follows: y = -217.68x + 1706.6; Where y is the adsorption value of lead ion crystal layer; x is the position of hydroxyl (OH) absorption peak, unit: cm -1 .
9. A loess soil lead pollution inversion system based on the GF-5 hyperspectral data based on the method according to any one of claims 1 to 8, characterized in that: Illite lead content inversion system includes: The hyperspectral data set acquisition module is used to acquire the hyperspectral data set of the Gaofen-5 satellite and perform radiation calibration and atmospheric correction operations to obtain an image reflecting the surface reflectance value; The image clipping and filtering module is used to clip the spectral bands of images reflecting the surface reflectance values, obtain the GF-5 shortwave infrared hyperspectral data set and perform Savitzky-Golay filtering and smoothing; Image matching module, used to determine the reference spectrum and match it with the filtered and smoothed GF-5 short-wave infrared hyperspectral data set, obtain the matching value image and perform statistical analysis; The data mask processing module is used to perform mask processing on the filtered GF-5 short-wave infrared hyperspectral data set and output the section with purer distribution of illite minerals; The lead content inversion module is used to perform first-order derivative operations on the short-wave infrared hyperspectral data set, output the spectral peak position image, calculate the lead content value of each pixel illite, and output the lead content value image.
Citation Information
Cited By
Sewage heavy metal pollution rapid detection method based on multispectral imaging
CN120741361A
Rapid detection method for heavy metal pollution in sewage based on multispectral imaging
CN120741361B
