A methane anomaly inversion false positive removal method and system suitable for a short-wave infrared spectral imager

Abstract

The application provides a methane anomaly inversion false positive removal method and system suitable for a short-wave infrared spectral imager, which utilizes spectral indexes, i.e., a normalized water index, a solar panel probability index, and a plastic greenhouse and artificial lawn probability index to identify water bodies, solar panels, plastic greenhouses, and artificial lawns; utilizes a constant color space combined with a threshold method to identify shadows; utilizes a threshold method and a band ratio method to identify clouds; and finally obtains an interference ground feature mask to reduce false positives when using a matching filter algorithm and a short-wave infrared satellite remote sensing image to invert methane concentration anomalies. Thus, the application helps to realize accurate, rapid, and low-cost satellite remote sensing of methane concentration anomalies, and lays a key foundation for identifying and quantifying methane emission sources.

Description

TECHNICAL FIELD

[0001] The present application mainly relates to the field of satellite remote sensing technology, and particularly relates to a methane anomaly inversion false positive removal method and system suitable for a short-wave infrared spectral imager. BACKGROUND

[0002] Using short-wave infrared band satellite imaging spectrometer data to invert methane column concentration anomaly, and then identifying and quantifying methane emission sources has become the frontier of international scientific and technological competition. This technology involves a matching filter algorithm, which is essentially a data-driven remote sensing inversion method that can quickly and accurately invert the methane concentration enhancement (also known as methane anomaly) of a region compared to the adjacent region without the aid of complex atmospheric transmission models. At present, the key bottleneck that hinders the technology from moving from academic research to large-scale practical application is that the matching filter algorithm will detect a large number of pseudo-anomalies while inverting the methane anomaly. Generally, the proportion of real methane anomaly in the total anomaly is less than 2%, which leads to the need for high-intensity manual interpretation to identify real anomalies. At the same time, this interpretation process requires a very high professional technical threshold, which results in high cost of implementing the technology process, limiting the wide application of the technology. SUMMARY

[0003] In view of this difficulty, the present application provides a methane pseudo-anomaly ground object automatic identification method suitable for short-wave infrared remote sensing images. The inventors have determined through a large number of experimental studies that the common ground objects that cause methane pseudo-anomaly inversion include five types of ground objects, i.e., solar panels, artificial lawns, plastic greenhouses, water bodies, clouds, and shadows. The present application describes a process and method for automatically extracting the above ground objects. By using the present application, the false positives that occur when using the matching filter algorithm and short-wave infrared satellite remote sensing images to invert methane concentration anomaly can be greatly reduced, thereby helping to realize accurate, rapid, and low-cost satellite remote sensing of methane concentration anomaly, laying a key foundation for identifying and quantifying methane emission sources.

[0004] Based on the above problems, the present application adopts the following technical solutions:

[0005] A methane anomaly inversion false positive removal method suitable for a short-wave infrared spectral imager, comprising the following steps:

[0006] Step 1. Calculate the normalized water body index, solar panel probability index, and plastic greenhouse and artificial lawn probability index based on spectral characteristics, and identify water bodies, solar panels, plastic greenhouses, and artificial lawns in the hyperspectral data;

[0007] Step 2. Based on the normalized water index, solar panel probability index, and plastic greenhouse and artificial turf probability index calculated in step 1, the threshold values of each extracted ground object are obtained using an adaptive algorithm, and further interference ground object masks of water, solar panels, plastic greenhouses, and artificial turf are obtained;

[0008] Step 3. Based on color space transformation, the RGB color space is transformed into an invariant color space, and shadows are identified by thresholding method;

[0009] Step 4. Based on the band ratio method and thresholding method, clouds are identified;

[0010] Step 5. The interference ground objects in the remote sensing image are removed using the identified interference ground object masks, and the false positives in the short-wave infrared satellite remote sensing image inversion of methane concentration anomalies are reduced.

[0011] Further, the normalized water index NDWI calculation expression in step 1 is as follows:

[0012]

[0013] Wherein, G and SWIR represent two bands in the remote sensing image with wavelength ranges of 542-549nm and 1603-1622nm, respectively.

[0014] Further, the solar panel probability index SPI calculation expression in step 1 is as follows:

[0015]

[0016]

[0017] Wherein, SWIR1 and SWIR2 represent two bands in the hyperspectral image with wavelength ranges of 1509-1518nm and 1728-1733nm, respectively, and temp SPI represents the intermediate value for calculating the solar panel probability index.

[0018] Further, the plastic greenhouse and artificial turf probability index PRPI calculation expression in step 1 is as follows:

[0019]

[0020]

[0021] Wherein, SWIR2, SWIR3, and SWIR4 represent three bands in the hyperspectral image with wavelength ranges of 1728-1733nm, 1721-1728nm, and 1745-1750nm, respectively, and temp PRPIThe intermediate value of the probability index of the plastic greenhouse and the artificial turf is calculated.

[0022] Further, in step 2, the NDWI histogram, the SPI histogram and the PRPI histogram are obtained respectively, wherein the NDWI histogram is a bimodal curve, and the trough is the water body recognition threshold; the SPI histogram and the PRPI histogram are approximately normal distribution curves, and the μ+3σ approximate point is the feature object recognition threshold; based on the above rules, the feature object recognition threshold is extracted by an adaptive calculation method, and further, the interference feature object masks of the water body, the solar panel, the plastic greenhouse and the artificial turf are obtained.

[0023] Further, in step 3, the specific process is as follows:

[0024] The spectral sensitivity of the red-green-blue sensor is f R (λ), f G (λ) and f B (λ) respectively, and the ground patch image measurement value under the SPD illumination of the incident light e(λ) is obtained.

[0025] C=m b (n,s)∫ λ f C (λ)e(λ)c b (λ)dλ

[0026] +m s (n,s,v)∫ λ f C (λ)e(λ)c s (λ)dλ

[0027] Wherein, C={R,G,B} is the response of the Cth sensor, c b (λ) and c s (λ) are the ground albedo and the Fresnel reflectance respectively, λ is the wavelength, n is the surface patch normal, s is the light source direction, v is the observer direction, and the geometric terms m b and m s represent the geometric dependence of the object and the surface reflection component respectively.

[0028] When the overall white condition is met, the reflectance calculation formula of the object is as follows:

[0029] C b =em b (n,s)k C

[0030] Wherein, C b ∈{R b ,G b ,B bThe red, green, blue sensor responses under white light source are given, object color and sensor surface reflectivity are k C , light intensity is e and object geometry is m b (n,s) on the brightness of the surface;

[0031] Based on the color space transformation, the RGB color space is transformed into the invariant color space (C1, C2, C3);

[0032] Wherein,

[0033]

[0034]

[0035]

[0036] The reflectivity of the object is calculated by substituting the formula into the above formula, and the C1, C2, C3 calculation formula only depends on the sensor and the surface reflectivity:

[0037]

[0038]

[0039]

[0040] Wherein, k B is the sensor surface reflectivity of the blue band with a wavelength range of 464-469nm, k G is the sensor surface reflectivity of the green band with a wavelength range of 548-552nm, k R is the sensor surface reflectivity of the red band with a wavelength range of 635-641nm;

[0041] Combined with the threshold method of red, green, blue and near-infrared band four bands, the shadow area and the ground building group are distinguished.

[0042] Further, the threshold formula for identifying shadow is as follows:

[0043]

[0044] Wherein, C1, C2, C3 respectively correspond to the values of three bands of the invariant color space, R, G, B are the radiance values of red, green and blue bands with a wavelength range of 635-641nm, 548-552nm and 464-469nm respectively, and NIR is the radiance value of the near-infrared band with a wavelength range of 856-860nm.

[0045] Further, in step 4, the apparent reflectance for identifying the cloud layer is calculated based on the spectral characteristics of the cloud layer; and the equivalent apparent reflectance is calculated after the narrow waveband of the high-resolution remote sensing image is synthesized into an equivalent wide waveband,

[0046] The calculation formula of the apparent reflectance for identifying the cloud layer is as follows:

[0047]

[0048] Wherein p is the apparent reflectance of the top layer of the atmosphere; D is the average distance from the earth to the sun; E sun is the average solar spectral irradiance outside the atmospheric layer, L is the radiance value, and θ is the solar zenith angle at the time of image scanning;

[0049] The calculation formula of the equivalent apparent reflectance calculated after the narrow waveband of the high-resolution remote sensing image is synthesized into an equivalent wide waveband is as follows:

[0050]

[0051] Wherein R is the equivalent apparent reflectance value of the synthesized wide waveband; P λ is the apparent reflectance when the wavelength is λ. The threshold method is used to identify the cloud layer by calculating the equivalent apparent reflectance of the corresponding waveband of the hyperspectral data in the wavelength range of 433-471 nm, the wavelength range of 510-640 nm, the wavelength range of 1350-1358 nm, and the wavelength range of 2000-2030 nm.

[0052] Further, the threshold formula for identifying the cloud layer is as follows:

[0053]

[0054] Wherein B1, B2, B3 and B4 are the equivalent apparent reflectances calculated by using the wavelength range of 433-471 nm, the wavelength range of 510-640 nm, the wavelength range of 1350-1358 nm, and the wavelength range of 2000-2030 nm, respectively.

[0055] On the other hand, the present application provides a methane abnormal inversion false positive removal system suitable for a short-wave infrared spectral imager, which specifically comprises the following modules,

[0056] The first module: calculate the normalized water body index, the solar panel probability index, the plastic greenhouse and artificial lawn probability index based on the spectral characteristics, and identify the water body, the solar panel, the plastic greenhouse and the artificial lawn in the hyperspectral data;

[0057] The second module: based on the solar panel probability index, the plastic greenhouse and artificial lawn probability index calculated in step 1, the threshold values of each extracted ground object are obtained by using an adaptive algorithm, and interference ground object masks of four types of ground objects, i.e., water body, solar panel, plastic greenhouse and artificial lawn, are further obtained;

[0058] The third module: based on color space transformation, the RGB color space is transformed into an invariant color space, and the shadow is identified by threshold value method;

[0059] The fourth module: based on the band ratio method and the threshold value method, the cloud layer is identified;

[0060] The fifth module: the interference ground objects existing in the remote sensing image are removed by using the identified interference ground object masks, and the false positive in the short-wave infrared satellite remote sensing image inversion of methane concentration is reduced.

[0061] Compared with the prior art, the present application has the following beneficial effects:

[0062] 1. The calculation formula of the normalized water body index is improved, and the water body, as a disturbance ground object, can be better identified;

[0063] 2. The solar panel probability index and the plastic greenhouse and artificial lawn probability index are proposed, and three types of interference ground objects, i.e., solar panel, plastic greenhouse and artificial lawn, are effectively identified;

[0064] 3. The method for adaptively extracting the identification threshold values of four types of interference ground objects, i.e., water body, solar panel, plastic greenhouse and artificial lawn, is proposed, and the interference ground objects can be flexibly identified. BRIEF DESCRIPTION OF DRAWINGS

[0065] Figure 1 The flowchart of the present application.

[0066] Figure 2 The result map of the original remote sensing image using the matched filter to invert the methane concentration;

[0067] Figure 3 The result map of the processed image using the matched filter to invert the methane concentration in the embodiment of the present application. DETAILED DESCRIPTION

[0068] In order to facilitate those skilled in the art to understand and implement the present application, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application.

[0069] Embodiment 1

[0070] As shown in the drawings, Figure 1 the embodiment of the present application provides a methane anomaly inversion false positive removal method suitable for a short-wave infrared spectral imager, which comprises the following steps:

[0071] Step 1. Calculate the normalized water index, solar panel probability index, plastic greenhouse and artificial turf probability index based on spectral characteristics, identify water, solar panels, plastic greenhouses and artificial turf in hyperspectral data;

[0072] This step is based on the spectral characteristics of water, solar panels, plastic greenhouses and artificial turf, and obtains the normalized water index suitable for hyperspectral resolution data, and proposes the solar panel probability index and the plastic greenhouse and artificial turf probability index, and extracts four types of interference ground objects of water, solar panels, plastic greenhouses and artificial turf in the image.

[0073] This step is improved on the basis of the traditional normalized water index. The traditional normalized water index uses near-infrared and short-wave infrared bands for calculation, but through comparative observation, it is found that the difference between water and other ground objects in NDWI calculated by green band and short-wave infrared band in hyperspectral imager is larger, which is more suitable for extracting water. Formula 1 gives the calculation method of the normalized water index NDWI of this embodiment, where G and SWIR represent the green band and the short-wave infrared band of the remote sensing image respectively. is the use of G and SWIR to extract water, because water has a higher reflectivity in the G band, and a lower reflectivity in the SWIR band. In the Gao Fen 5 data, the 222nd band (wavelength: 1606nm) represents the SWIR band, and the 38th band (wavelength: 545nm) represents the green band.

[0074]

[0075] Formula 2-3 gives the calculation method of the solar panel probability index (SPI), where SWIR1 and SWIR2 represent two short-wave infrared bands of the hyperspectral image, and these two bands represent the 211th band (wavelength: 1514nm) and the 237th band (wavelength: 1732nm) of Gao Fen 5 respectively.

[0076]

[0077]

[0078] Formula 4-5 gives the calculation method of the plastic greenhouse and artificial turf probability index (PRPI), where SWIR2, SWIR3 and SWIR4 represent three short-wave infrared bands of the hyperspectral image, and these three bands represent the 237th band (wavelength: 1732nm), the 236th band (wavelength: 1724nm) and the 239th band (wavelength: 1749nm) of Gao Fen 5 respectively.

[0079]

[0080]

[0081] The result histogram calculated by the above three indexes and the feature recognition threshold has the following rules:

[0082] (1) The NDWI histogram is a bimodal curve, and the trough is the water body recognition threshold

[0083] (2) The SPI and PRPI histograms are approximately normal distribution curves, and μ+3σ is approximately the feature recognition threshold.

[0084] According to the above rules, the adaptive calculation method is written to extract the feature recognition threshold, and further to obtain the interference feature mask of the four types of features of water body, solar panel, plastic greenhouse and artificial lawn.

[0085] Step 2. Based on the normalized water body index, solar panel probability index, plastic greenhouse and artificial lawn probability index calculated in step 1, the adaptive algorithm is used to obtain the threshold of each extracted feature, and further to obtain the interference feature mask of the four types of features of water body, solar panel, plastic greenhouse and artificial lawn;

[0086] Step 3. Based on the color space transformation, the RGB color space is transformed into the invariant color space, and the shadow is recognized by threshold method;

[0087] Formula 6 gives the calculation method of the ground patch image measurement value under the SPD illumination of incident light e(λ) using the red, green and blue sensors with spectral sensitivity f R (λ), f G (λ) and f B (λ) respectively:

[0088] C=m b (n,s)∫ λ f C (λ)e(λ)c b (λ)dλ+m s (n,s,v)∫ λ f C (λ)e(λ)c s (λ)dλ (6)

[0089] C={R,G,B} gives the response of the Cth sensor. c b (λ) and c s (λ) are the ground albedo and Fresnel reflectance respectively. λ is the wavelength, n is the surface patch normal, s is the light source direction, and v is the observer direction. The geometric terms m b and m s represent the geometric dependence of the object and surface reflection components respectively.

[0090] Equation 7 gives the reflectance of an object when the overall white condition holds:

[0091] C b = em b (n,s)k C (7)

[0092] where for C b ∈ {R b ,G b ,B b} gives the red, green, blue sensor response under a white light source, the object color and sensor and surface reflectance k C , the illumination intensity e and the object geometry m b (n,s). Thus a uniformly colored surface will produce large RGB differences. The RGB is not sensitive to the surface orientation, the light orientation and the illumination intensity.

[0093] Therefore, the invariant color model, i.e. the invariant color space (C1, C2, C3) can be used as the best nonlinear transformation for shadow detection.

[0094] Equations 8-10 give the definitions of C1, C2, C3:

[0095]

[0096]

[0097]

[0098] represents the angle of the object reflectance vector, and thus is invariant for matte, dark objects.

[0099] Equations 11-13 show the derivation of the C1, C2, C3 computation equations from equation 7 into equations 9-10, which depend only on the sensor and surface reflectance:

[0100]

[0101]

[0102]

[0103] However, only the invariant color space transformation cannot distinguish the shadowed areas from the ground building clusters with blue roofs. Usually, the radiance values of the shadowed areas and the ground building clusters with blue roofs are quite different, so the threshold method combining the red, green, blue and near infrared bands can be used to distinguish the shadowed areas from the ground building clusters with blue roofs.

[0104] Equation 14 gives the threshold equation for shadow detection.

[0105]

[0106] In the data of GF-5 No. 02 satellite, the 60th band (wavelength: 639 nm) is selected for the red band (R), the 39th band (wavelength: 549 nm) is selected for the green band (G), the 20th band (wavelength: 468 nm) is selected for the blue band (B), and the 111th band (wavelength: 857 nm) is selected for the near-infrared band (NIR).

[0107] Step 4. Identify the cloud layer based on the band ratio method and the threshold method;

[0108] This step uses the band ratio method combined with the threshold method to identify the cloud layer based on the spectral characteristics of the cloud layer.

[0109] Formula 15 gives the calculation formula of the apparent reflectance for identifying the cloud layer.

[0110]

[0111] where p is the apparent reflectance of the top layer of the atmosphere; D is the average distance from the earth to the sun; E sun is the average solar spectral irradiance outside the atmospheric layer, L is the radiance value, and θ is the solar zenith angle at the time of image scanning.

[0112] Formula 16 gives the calculation method for calculating the equivalent apparent reflectance after synthesizing the narrow bands of the high-resolution remote sensing image into equivalent wide bands.

[0113]

[0114] where R is the equivalent apparent reflectance value of the synthesized wide band; P λ is the apparent reflectance at wavelength λ. The threshold method is used to identify the cloud layer by calculating the equivalent apparent reflectance of the corresponding bands of the hyperspectral data in the four wavelength range intervals: wavelength range 1 (433-471 nm), wavelength range 2 (510-640 nm), wavelength range 3 (1350-1358 nm), and wavelength range 4 (2000-2030 nm).

[0115] Formula 17 gives the calculation formula of the threshold method and the band ratio method for identifying the cloud layer.

[0116]

[0117] where B1, B2, B3, and B4 are the equivalent apparent reflectances calculated using wavelength ranges 1, 2, 3, and 4, respectively. In the data of GF-5, wavelength ranges 1, 2, 3, and 4 correspond to bands 11-20, 30-60, 192, and 270-272, respectively.

[0118] Step 5. Remove the interference ground objects in the remote sensing image using the identified interference ground object mask to reduce false positives when the short-wave infrared satellite remote sensing image is used to retrieve methane concentration anomalies.

[0119] Figure 2 and Figure 3 The results of using the original remote sensing image and the image processed using the present patent to retrieve methane concentration are shown in the graphs.

[0120] Embodiment 2

[0121] The present embodiment provides a methane anomaly retrieval false positive removal system suitable for a short-wave infrared spectral imager, which specifically includes the following modules,

[0122] The first module: based on spectral features, calculate the normalized water index, solar panel probability index, plastic greenhouse and artificial turf probability index, identify water, solar panels, plastic greenhouses and artificial turf in hyperspectral data;

[0123] The second module: based on the normalized water index, solar panel probability index, plastic greenhouse and artificial turf probability index calculated in step 1, use an adaptive algorithm to obtain the threshold value of each extracted ground object, and further obtain the interference ground object mask of the four types of ground objects, i.e. water, solar panels, plastic greenhouses and artificial turf;

[0124] The third module: based on color space transformation, transform the RGB color space into an invariant color space, and identify shadows by thresholding;

[0125] The fourth module: based on the band ratio method and thresholding, identify clouds;

[0126] The fifth module: use the identified interference ground object mask to remove the interference ground objects in the remote sensing image, and reduce false positives when the short-wave infrared satellite remote sensing image is used to retrieve methane concentration anomalies.

[0127] The above is only the preferred embodiment of the present application, and does not limit the implementation and protection scope of the present application. For those skilled in the art, it should be realized that any equivalent replacement and obvious changes made according to the content of the present application should be included in the protection scope of the present application.

Claims

1. A methane anomaly inversion false positive removal method suitable for a shortwave infrared spectral imager, characterized in that, Comprising the following steps: Step 1. Calculate the normalized water index, solar panel probability index, plastic greenhouse and artificial turf probability index based on spectral characteristics, identify water, solar panels, plastic greenhouses and artificial turf in hyperspectral data; Step 2. Based on the normalized water index, solar panel probability index, plastic greenhouse and artificial turf probability index calculated in step 1, use the adaptive algorithm to obtain the threshold value of each extracted ground object, and further obtain the interference ground object mask of the four types of ground objects of water, solar panels, plastic greenhouses and artificial turf; Step 3. Based on color space transformation, transform the RGB color space into an invariant color space, and identify the shadow by threshold method; Step 4. Based on the band ratio method and threshold method, identify the cloud layer; based on the spectral characteristics of the cloud layer, calculate the apparent reflectance for identifying the cloud layer; and after the narrow waveband of the high-resolution remote sensing image is combined into an equivalent wide waveband, calculate the equivalent apparent reflectance, The calculation formula of the apparent reflectance for identifying the cloud layer is as follows: wherein is the apparent reflectivity of the top layer of the atmosphere; is the average distance from the Earth to the Sun; is the average solar spectral irradiance outside the atmosphere in the waveband, is the radiance value, is the solar zenith angle at the time of the image scan; After the narrow waveband of the high-resolution remote sensing image is combined into an equivalent wide waveband, the calculation formula of the equivalent apparent reflectance is as follows: wherein is the equivalent apparent reflectance value of the synthesized wide band; is the apparent reflectance when the wavelength is The threshold method is used to identify the cloud layer by calculating the equivalent apparent reflectance of the corresponding wave band of the hyperspectral data in the wavelength range of 433-471 nm, the wavelength range of 510-640 nm, the wavelength range of 1350-1358 nm, and the wavelength range of 2000-2030 nm. Step 5. Remove the interference ground objects existing in the remote sensing image by using the identified interference ground object mask, and reduce the false positives when the short-wave infrared satellite remote sensing image is used to retrieve the abnormal methane concentration.

2. The method for removing false positive of methane anomaly inversion according to claim 1, wherein: The calculation expression of the normalized water index NDWI in step 1 is as follows: Wherein, G and SWIR respectively represent two bands in the wavelength range of 542-549 nm and 1603-1622 nm in the remote sensing image.

3. The method for removing false positive of methane anomaly inversion according to claim 1, wherein: The calculation expression of the solar panel probability index SPI in step 1 is as follows: wherein SWIR1 and SWIR2 represent two bands in the hyperspectral image with wavelength ranges of 1509-1518 nm and 1728-1733 nm, respectively, represents the median value of the computed solar panel probability index.

4. The method for removing false positive of methane anomaly inversion according to claim 1, wherein: The calculation expression of the plastic greenhouse and artificial turf probability index PRPI in step 1 is as follows: wherein SWIR2, SWIR3, SWIR4 represent three bands in the hyperspectral image with wavelength ranges of 1728-1733 nm, 1721-1728 nm, and 1745-1750 nm, respectively, The intermediate value represents the probability index of the plastic greenhouse and artificial turf.

5. The method for removing false positive of methane anomaly inversion according to claim 1, wherein: In step 2, the NDWI histogram, the SPI histogram and the PRPI histogram are obtained respectively, wherein the NDWI histogram is a bimodal curve, and the trough is the water body recognition threshold; the SPI histogram and the PRPI histogram are normal distribution curves, and the approximate point is the feature recognition threshold The approximate point is the feature recognition threshold, the feature recognition threshold is extracted by an adaptive calculation method, and the interference feature masks of four types of features, i.e., water body, solar panel, plastic greenhouse and artificial lawn, are further obtained.

6. The method for removing false positive of methane anomaly inversion according to claim 1, wherein: In step 3, the specific process is as follows: Using red-green-blue sensors with spectral sensitivities of , , , the surface patch image measurements under SPD illumination of incoming light are obtained; wherein, is the Cth sensor response, and are the ground albedo and Fresnel reflectance, respectively, is the wavelength, n is the surface patch normal, s is the light source direction, v is the observer direction, the geometric term and denote the geometric dependence of the object and surface reflection components, respectively; When the overall white condition is met, the reflectivity calculation formula of the object is as follows: wherein, The red, green, blue sensor responses under a white light source are given by , the object color and sensor and surface reflectance are , the luminance on Based on color space transformation, transform the RGB color space into an invariant color space (C1, C2, C3); Wherein, Substitute the reflectivity calculation formula of the object into the above formula to obtain the C1, C2, C3 calculation formula which only depends on the sensor and the surface albedo: wherein, is the sensor surface reflectance using the blue band with a wavelength range of 464-469 nm, is the sensor surface reflectance using the green band with a wavelength range of 548-552 nm, is the sensor surface reflectance using the red band with a wavelength range of 635-641 nm, 7. The method for removing false positive of methane anomaly inversion according to claim 6, wherein: The threshold formula for identifying the shadow is as follows: wherein, , , respectively correspond to the values of the three wavebands of the invariant color space, , , respectively are the radiance values of the red, green and blue wavebands with wavelength ranges of 635-641 nm, 548-552 nm and 464-469 nm, is the radiance value using the near-infrared waveband with wavelength range of 856-860 nm.

8. The method for removing false positive of methane anomaly inversion according to claim 1, wherein: The threshold formula for identifying the cloud layer is as follows: wherein , , , are the equivalent apparent reflectances calculated with the wavelength range 433-471 nm, the wavelength range 510-640 nm, the wavelength range 1350-1358 nm, the wavelength range 2000-2030 nm, respectively.

9. A methane anomaly inversion false positive removal system suitable for use in a shortwave infrared spectral imager, the system comprising: Specifically includes the following modules, The first module: calculate the normalized water index, solar panel probability index, plastic greenhouse and artificial turf probability index based on spectral characteristics, identify water, solar panels, plastic greenhouses and artificial turf in hyperspectral data; The second module: based on the normalized water index, solar panel probability index, plastic greenhouse and artificial turf probability index calculated in step 1, use the adaptive algorithm to obtain the threshold value of each extracted ground object, and further obtain the interference ground object mask of the four types of ground objects of water, solar panels, plastic greenhouses and artificial turf; The third module: based on color space transformation, transform the RGB color space into an invariant color space, and identify the shadow by threshold method; The fourth module: based on the band ratio method and threshold method, identify the cloud layer; The fifth module: remove the interference ground objects existing in the remote sensing image by using the identified interference ground object mask, and reduce the false positives when the short-wave infrared satellite remote sensing image is used to retrieve the abnormal methane concentration; The methane anomaly inversion false positive removal system for a shortwave infrared spectral imager is configured to perform the steps of the method of any of claims 1-8.

Images