Hyperspectral remote sensing image methane point source detection method based on superpixel matching filter

Abstract

The application provides a hyperspectral remote sensing image methane point source detection method based on superpixel matching filtering, which comprises the following steps: acquiring an original hyperspectral remote sensing image and a standard methane spectral absorption rate; obtaining a true color image according to the original hyperspectral remote sensing image, and obtaining a background spectrum according to the true color image; constructing a methane target spectrum according to the standard methane spectral absorption rate and the background spectrum; filtering the methane target spectrum and the original hyperspectral image by using a matching filtering algorithm to obtain a methane point source filtering result; and obtaining a final methane point source position according to the methane point source filtering result. The application can reduce the complexity of methane point source detection and improve the detection accuracy.

Description

Technical Field

[0001] This invention relates to the field of atmospheric technology, and more specifically, to a method for detecting methane point sources in hyperspectral remote sensing images based on superpixel matched filtering. Background Technology

[0002] As one of the major greenhouse gases, accurate monitoring of methane is of significant environmental and social importance. Current methane detection technologies focus on detection sensors, traditional estimation methods, and deep learning methods.

[0003] Short-wave infrared sensors are more sensitive to near-ground methane concentration changes than thermal infrared sensors and are now widely used in multi-scale methane monitoring. Among these, the large data volume and temporal resolution of multispectral data make it suitable for long-term point source monitoring; however, its accuracy and detection limit for methane point sources are significantly lower than those of hyperspectral data, so it is often used in conjunction with hyperspectral data. Hyperspectral remote sensing images, benefiting from their fine and nearly continuous spectral resolution, have high potential in methane point source detection and concentration inversion. Traditional estimation methods mainly include physical algorithms, CO2 surrogate methods, and matched filtering methods. Physical algorithms are based on the radiative transfer equation and are greatly affected by weather and surface illumination. The CO2 surrogate method, a type of proxy algorithm, utilizes the fact that both CO2 and methane have absorption characteristics at 1.6 μm, allowing concentration inversion using CO2 within a common absorption band. Using CO2 as a reference eliminates the optical path changes caused by scattering, thus correcting the methane concentration. Matched filtering directly obtains the methane enhancement, rather than the overall methane concentration of the entire region, making it more computationally efficient than the methods mentioned above and suitable for point source detection tasks. Deep learning has been widely applied in the environmental field, including solar energy forecasting and wind energy forecasting, but its application in methane emission detection is relatively limited. Current deep learning networks for methane detection mostly use CNNs and other underlying architectures.

[0004] In summary, the development trend of detection sensors in methane monitoring shows a focus on high spatial, high spectral, and high temporal resolution, as well as the integration of multi-source remote sensing, providing a relatively complete data source for multi-scale methane emission monitoring. Traditional methods such as physical algorithms, CO2 surrogate methods, and matched filtering are still widely used, but their effectiveness varies in specific applications, requiring flexible selection based on data and objectives. Deep networks for methane detection show potential in methane emission detection tasks, but their current application scope is relatively limited. Summary of the Invention

[0005] To overcome the shortcomings of current methods for detecting methane point sources, this invention provides a method for detecting methane point sources in hyperspectral remote sensing images based on superpixel matching filtering.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0007] This invention proposes a method for detecting methane point sources in hyperspectral remote sensing images based on superpixel matched filtering, comprising the following steps:

[0008] Acquire raw hyperspectral remote sensing images and standard methane spectral absorbance;

[0009] A true-color image is obtained from the original hyperspectral remote sensing image, and the background spectrum is obtained from the true-color image.

[0010] Construct the target methane spectrum based on the standard methane spectral absorbance and background spectrum;

[0011] A matched filtering algorithm was used to filter the methane target spectrum and the original hyperspectral image to obtain the methane point source filtering result;

[0012] The final methane point source location is obtained based on the methane point source filtering results.

[0013] Preferably, acquiring a hyperspectral remote sensing image and obtaining a true-color image from the hyperspectral remote sensing image includes synthesizing the hyperspectral remote sensing image using RGB bands to obtain a true-color image.

[0014] Preferably, obtaining the background spectrum from the true-color image includes segmenting the background plots of the true-color image using a superpixel segmentation algorithm, and then performing superpixel aggregation on the segmented plots to obtain the background spectrum.

[0015] Preferably, obtaining the standard methane spectral absorbance includes:

[0016] Based on the MODTRAN platform, the standard atmospheric temperature curve was selected as the median atmospheric model for simulation to obtain the absorption characteristics of methane in the short-wave infrared range, which was used as the standard methane spectral absorbance.

[0017] Preferably, the method of using a matched filtering algorithm to filter the methane target spectrum and the original hyperspectral image to obtain the methane point source filtering result includes:

[0018]

[0019] Where P is the methane point source distribution probability, i.e., the methane point source detection filtering result, x p The methane enhancement value of the methane point source is used as the result of the probabilistic processing, which is the methane point source detection filtering result of the test data.

[0020] Preferably, obtaining the location of the most likely methane point source based on the methane point source filtering results includes:

[0021] The methane point source filtering results are sequentially subjected to Gaussian filtering and eight-neighborhood feature estimation to obtain the location of the most likely methane point source.

[0022] Preferably, the step of sequentially performing Gaussian filtering and eight-neighborhood feature estimation on the methane point source filtering results includes:

[0023] For the methane point source filtering result, the average gray value of all pixels within a 3×3 region centered on each pixel is taken as the gray value of the center point. A two-dimensional Gaussian function is used as the weight allocation method, and the image is convolved using a Gaussian kernel generated by the Gaussian function. Specifically:

[0024]

[0025] Where x and y represent the two-dimensional planar coordinates of a pixel, σ is the standard deviation, and optimization is performed using the separating properties of the Gaussian function:

[0026]

[0027] Preferably, the step of sequentially performing Gaussian filtering and eight-neighborhood feature estimation on the methane point source filtering results further includes:

[0028] After Gaussian filtering, an eight-neighborhood feature estimation algorithm based on 8-connected neighborhood labeling is used to perform eight-neighborhood feature estimation on the Gaussian-filtered image. The statistics of pixel values ​​within the eight-connected neighborhood can be expressed as:

[0029] N8(P)={(x+i,y+j)|i,j∈{-1,0,1}}\(x,y)

[0030] For any pixel in an image, its 8-connected neighborhood N8(P) is defined as the set containing P(X,Y) and its eight surrounding pixels:

[0031] The specific algorithm steps are as follows:

[0032] 1) Region labeling: Traverse each pixel in the image and use a labeling function f(P) to record the region to which each pixel belongs;

[0033] 2) Isolated point detection: For each pixel P(X,Y), check its 8-connected neighborhood N8(P). If P belongs to a specific category and there are no other pixels of the same category in its 8-connected neighborhood, then P is considered an isolated point.

[0034] 3) Isolated Point Removal: The pixel value of the detected isolated point P is set to the background value of that point or the average value of its neighboring pixels. Here, the isolated point is set to the background value, and its mathematical expression is as follows:

[0035] Let I(X,Y) represent the pixel value at coordinates (X,Y) in the image, and B(X,Y) represent the background pixel value. For each pixel P(X,Y), the statistics of the pixel values ​​in its 8-connected neighborhood can be expressed as:

[0036]

[0037] If P is an isolated point, then S(P) will not contain pixel values ​​of the same category as P. In this case, the pixel values ​​of P can be updated as follows:

[0038]

[0039] This invention also proposes a hyperspectral remote sensing image methane point source detection system based on superpixel matched filtering, used to implement the above-described method, comprising:

[0040] The data acquisition module is used to acquire raw hyperspectral remote sensing images and standard methane spectral absorbance.

[0041] The true-color image and background spectrum acquisition module is used to obtain a true-color image from the original hyperspectral remote sensing image and to obtain a background spectrum from the true-color image.

[0042] The target spectrum construction module is used to construct the target spectrum of methane based on the standard methane spectral absorbance and background spectrum.

[0043] The filtering module is used to filter the methane target spectrum and the original hyperspectral image using a matched filtering algorithm to obtain the methane point source filtering result;

[0044] The result acquisition module is used to obtain the final methane point source location based on the methane point source filtering results.

[0045] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0046] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0047] This invention proposes a method for detecting methane point sources in hyperspectral remote sensing images based on superpixel matched filtering. By introducing superpixel segmentation technology, the initial hyperspectral remote sensing image is divided into multiple sub-regions, thereby simplifying the image's texture features, reducing computational complexity, and improving analysis efficiency. Secondly, Gaussian filtering is used to smooth the image, and combined with octet region feature statistics, background interference is effectively suppressed and isolated points are removed. This invention can reduce the complexity of methane point source detection and improve detection accuracy. Attached Figure Description

[0048] Figure 1 This is a flowchart illustrating the method described in Example 1;

[0049] Figure 2 This is a system block diagram of the method described in Example 2;

[0050] Figure 3 The original hyperspectral image is the one obtained using the method described in Example 2.

[0051] Figure 4 The matched filtering result is the result of the method described in Example 2;

[0052] Figure 5 The result of Gaussian filtering is shown in Example 2.

[0053] Figure 6 The final methane point source location is the one described in Example 2.

[0054] Figure 7 This is an enlarged view of the final methane point source location in the method described in Example 2;

[0055] Figure 8 This is a schematic diagram of the system described in Example 3. Detailed Implementation

[0056] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent.

[0057] To better illustrate this embodiment, some parts in the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions;

[0058] It will be understood by those skilled in the art that certain well-known structures and their descriptions may be omitted in the accompanying drawings.

[0059] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0060] Example 1

[0061] This embodiment provides a method for detecting methane point sources in hyperspectral remote sensing images based on superpixel matched filtering, such as... Figure 1 As shown, it includes the following steps:

[0062] Acquire raw hyperspectral remote sensing images and standard methane spectral absorbance;

[0063] A true-color image is obtained from the original hyperspectral remote sensing image, and the background spectrum is obtained from the true-color image.

[0064] Construct the target methane spectrum based on the standard methane spectral absorbance and background spectrum;

[0065] A matched filtering algorithm was used to filter the methane target spectrum and the original hyperspectral image to obtain the methane point source filtering result;

[0066] The final methane point source location is obtained based on the methane point source filtering results.

[0067] In the specific implementation process, the original hyperspectral remote sensing image and standard methane spectral absorbance are first acquired. Then, the background spectrum is obtained based on the original hyperspectral remote sensing image. Next, the methane target spectrum is constructed based on the standard methane spectral absorbance and the background spectrum. The methane target spectrum and the original hyperspectral image are filtered using a matched filtering algorithm to obtain the methane point source filtering result. Finally, the final methane point source location is obtained based on the methane point source filtering result.

[0068] Example 2

[0069] This embodiment provides a method for detecting methane point sources in hyperspectral remote sensing images based on superpixel matched filtering, including the following steps:

[0070] Acquire raw hyperspectral remote sensing images and standard methane spectral absorbance;

[0071] A true-color image is obtained from the original hyperspectral remote sensing image, and the background spectrum is obtained from the true-color image.

[0072] Construct the target methane spectrum based on the standard methane spectral absorbance and background spectrum;

[0073] A matched filtering algorithm was used to filter the methane target spectrum and the original hyperspectral image to obtain the methane point source filtering result;

[0074] The final methane point source location is obtained based on the methane point source filtering results.

[0075] like Figure 2 As shown, the method described in this embodiment consists of four steps: First, using GF-5 02AHSI data as the original hyperspectral image, a superpixel segmentation and aggregation strategy is employed to divide the hyperspectral image into superpixels. Second, the standard methane spectral absorbance simulated by MODTRAN is fused with the background spectrum. Then, a matched filtering algorithm is used to search for methane emission points in each superpixel block. Finally, Gaussian filtering and feature statistical analysis are used to obtain the final methane emission source locations.

[0076] The details are as follows:

[0077] (1) MODTRAN methane point source spectrum simulation

[0078] AHSI is one of the main payloads of the GF5-02 satellite. It can acquire spectral information of ground objects with a 60km swath, 30m spatial resolution, and a spectral range of 400-2500nm (spectral resolution: 5-10nm) with a total of 330 spectral channels.

[0079] This embodiment is based on the MODTRAN platform (http: / / modtran.spectral.com / ) and selects the 1976 US standard atmospheric temperature curve as the effective median atmospheric model for simulation. The transmission and absorption characteristics of methane in the shortwave infrared range of atmospheric greenhouse gases are obtained by atmospheric radiative transfer simulation in MODTRAN.

[0080] The absorbance and transmittance of standard methane can be converted using the following formula:

[0081] n(β A )=-lg[n(β T (1)

[0082] In the formula, β A and β T This represents the absorption and transmission coefficients.

[0083] Methane exhibits strong absorption characteristics around 2300 nm, and the 2100-2450 nm window of GF-5 02AHSI is quite sensitive to changes in XCH4. Therefore, this embodiment uses the spectral characteristics of this window band to detect the location of methane point sources.

[0084] (2) Superpixel segmentation and aggregation

[0085] Superpixel segmentation refers to dividing an initial image into multiple sub-regions (sets of pixels). To simplify image texture and facilitate analysis, superpixel segmentation algorithms are widely used in computer vision. In this embodiment, the Entropy Rate Segmentation (ERS) algorithm is used to aggregate pixels in a hyperspectral true-color image to generate two-dimensional superpixel patches. Specifically, the image is mapped onto an undirected graph, as shown below:

[0086] G=(V,E) (2)

[0087] In the formula, the vertex set V represents the pixels in the initial image, and the edge set E consists of the similarity between pixel pairs defined by the weight function. Then, a subset of the edges is selected... The graph is divided into several connected subgraphs such that the resulting graph G′=(V,O) consists of smaller connected subgraphs. Therefore, a graph-theory-based clustering objective function is proposed, and an entropy rate term H(·) and a balance term B(·) are introduced to optimize superpixel segmentation:

[0088]

[0089] In the formula, H(O) tends to form uniform and compact clusters, while B(O) is used to promote clusters of similar size. The parameter α is introduced as a contribution to the equilibrium term and the entropy rate term. Furthermore, the entropy rate H(·) of a random walk on G′=(V,O) can be expressed as a set of functions:

[0090] H(O)=-∑ i u i ∑ j p i,j (O)logp i,j (O) (4)

[0091] In the formula, p i,j (O) represents the transition probability when a particle randomly walks from vertex i to vertex j. i,j This represents the weight of the edge connecting vertices i and j. It is the sum of the weights of the edges passing through vertex i. The steady-state distribution is represented by u. i =w i / w T The calculation shows that, among which V is the number of vertices.

[0092] This embodiment utilizes a superpixel segmentation and aggregation algorithm to synthesize Gaofen-5 observation data using RGB bands to obtain a true-color image. For the true-color image, the aforementioned superpixel segmentation algorithm is used to segment background plots, and the segmented plots are then superpixel aggregated to obtain the background spectrum. Finally, the methane target spectrum is constructed using the methane spectral absorbance established by the MODTRAN simulation platform.

[0093] (3) Matched filtering

[0094] Matched filtering algorithms have been successfully applied in the inversion of methane point source column concentrations using airborne and aerospace hyperspectral sensors. Based on this, this embodiment uses GF5-02-AHSI data and an improved matched filtering algorithm to perform methane point source column concentration inversion. The core idea of ​​the matched filtering algorithm is to treat each input spectral radiance value as a superposition of the background radiance without methane enhancement and the radiative perturbation caused by changes in methane concentration. This perturbation is modeled as the target spectral radiance, which represents the radiative increment transfer signal of methane absorption per unit area. The target parameter for methane point source inversion is the methane point source column concentration. X relative to the surrounding background b Pixel-by-pixel enhancement, i.e. The specific calculation method is as follows:

[0095]

[0096] In the formula, L iThe input is the observed radiance value, μ and C are the mean and covariance of the background radiation, t is the target spectral radiance perturbation, representing the perturbation relative to the background radiation signal caused by the increased methane concentration, and T represents the transpose. The target spectral radiance perturbation t is obtained by multiplying the unit methane absorption spectrum s by the background radiation mean μ. The unit methane absorption spectrum s is obtained using atmospheric radiative transfer simulation on the MODTRAN platform. Since the real Earth's surface is not a Lambertian surface, its biaxial reflectance characteristics are significant. An empirical Minnaert coefficient k is introduced to describe the biaxial reflectance characteristics of the Earth's surface. When k = 1, it indicates that the Earth's surface is a Lambertian surface; a decrease in the value of k indicates an increase in surface anisotropy.

[0097] L H =L T (cosθ s / cosγ s ) k (6)

[0098] In the formula, L H To observe radiance below a flat surface, L T The radiance observed below the surface in mountainous areas. cosθ s It is the cosine of the zenith angle of the sun's incidence, cosγ s It is the cosine of the zenith angle relative to the sun's incidence.

[0099] The target spectrum is obtained by fusing the spectral absorbance adjustment factor obtained from MODTRAN simulation with the background spectrum obtained from superpixel aggregation. Target characteristics are then extracted from the target spectrum, and subsequently, spectral matching filtering is performed with the original GF5 observation data to obtain the probability of methane point source distribution. The probabilistic formula is as follows:

[0100]

[0101] In the formula, x p The methane enhancement value of the methane point source is used as the result of the probabilistic processing, which is the methane point source detection filtering result of the test data.

[0102] (4) Background suppression and isolated point removal

[0103] Based on prior knowledge, locations with methane point source emissions typically exhibit high concentrations at the center, gradually decreasing with increasing distance from the center. Therefore, some isolated points in the obtained methane enhancement image may be due to inherent data issues and are not the point sources to be extracted. Thus, this embodiment employs Gaussian filtering and octet region feature statistics to achieve background suppression and isolated point removal.

[0104] 1) Gaussian filtering algorithm: Taking each pixel as the center, the average of the gray values ​​of all pixels within its 3×3 region is taken as the gray value of the center point. A weighted average is used, with points closer to each other having a higher weight and points farther away having a lower weight. Based on this, this embodiment uses a two-dimensional Gaussian function as the weight allocation method and uses a Gaussian kernel (Gaussian template) generated by the Gaussian function to perform a convolution operation on the image, as detailed below:

[0105]

[0106] In the formula, x and y represent the two-dimensional planar coordinates of the pixel, and σ is the standard deviation. Optimization is then performed using the separating properties of the Gaussian function:

[0107]

[0108] 2) Eight-connected isolated point removal: After Gaussian filtering, noise in the image is effectively suppressed, but some isolated pixels may be generated, which are visually inconsistent with the surrounding environment. To remove these isolated points, this embodiment uses a region labeling algorithm based on eight connectivity. An eight-connected region refers to a connected region in a two-dimensional image formed by a pixel and its eight neighboring pixels (including horizontal, vertical, and diagonal directions). This connectivity definition is used to identify and analyze continuous objects in an image. Eight connectivity means that a pixel can be connected to its neighboring pixels in the horizontal, vertical, or diagonal directions. The mathematical definition of eight connectivity is expressed as follows:

[0109] N8(P)={(x+i,y+j)|i,j∈{-1,0,1}}\(x,y) (10)

[0110] For any pixel in an image, its 8-connected neighborhood N8(P) is defined as the set containing P(X,Y) and its eight surrounding pixels:

[0111] The specific algorithm steps are as follows:

[0112] 1. Region labeling: Traverse each pixel in the image and use a labeling function f(P) to record the region to which each pixel belongs.

[0113] 2. Isolated Point Detection: For each pixel P(X,Y), examine its 8-connected neighborhood N8(P). If P belongs to a specific category (e.g., boundary pixels after edge detection) and there are no other pixels of the same category within its 8-connected neighborhood, then P is considered an isolated point.

[0114] 3. Isolated Point Removal: The pixel value of the detected isolated point P is set to either the background value or the average value of its neighboring pixels to remove the isolated point. Here, the isolated point is set to the background value. Its mathematical expression is as follows:

[0115] Let I(X,Y) represent the pixel value of the image at coordinates (X,Y), and B(X,Y) represent the background pixel value. For each pixel P(X,Y), the statistics of the pixel values ​​in its 8-connected neighborhood can be expressed as:

[0116]

[0117] If P is an isolated point, then S(P) will not contain pixel values ​​of the same category as P. In this case, the pixel values ​​of P can be updated as follows:

[0118]

[0119] This embodiment utilizes a Gaussian filtering algorithm to process the image. Gaussian filtering effectively reduces random noise in the image while preserving edge features and avoiding excessive blurring. Considering the actual characteristics of methane distribution, which tends to appear in clusters rather than independent distributions, a feature statistical analysis of 8-connected regions is further performed on the image based on Gaussian filtering. Through these steps, this embodiment effectively removes 8-connected isolated points from the image, thereby improving the overall image quality and consistency. Furthermore, connectivity is utilized to remove scattered methane points, retaining connected methane sample points to improve detection accuracy.

[0120] This embodiment uses GF5-02-AHSI data as the original hyperspectral image, such as Figure 3 As shown, this is the original hyperspectral remote sensing image, with six detection regions marked A1-A6. Detection was performed on these regions, as follows... Figure 4 As shown, this represents the methane point source detection probability after matched filtering. Figure 5 The diagram shown is the structure after Gaussian filtering, compared to... Figure 4 In other words, Figure 5 Reduce random noise in the image and preserve the edge features of the image, such as Figure 6 The image shows the final methane point source location. Figure 7 for Figure 6 The magnified image of the corresponding detection area shows the location of the methane point source in the bright area.

[0121] This embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for detecting methane point sources in hyperspectral remote sensing images based on superpixel matched filtering.

[0122] Example 3

[0123] This embodiment provides a hyperspectral remote sensing image methane point source detection system based on superpixel matched filtering, used to implement the method described in Embodiment 1 or 2, such as... Figure 8 As shown, it includes:

[0124] The data acquisition module is used to acquire raw hyperspectral remote sensing images and standard methane spectral absorbance.

[0125] The true-color image and background spectrum acquisition module is used to obtain a true-color image from the original hyperspectral remote sensing image and to obtain a background spectrum from the true-color image.

[0126] The target spectrum construction module is used to construct the target spectrum of methane based on the standard methane spectral absorbance and background spectrum.

[0127] The filtering module is used to filter the methane target spectrum and the original hyperspectral image using a matched filtering algorithm to obtain the methane point source filtering result;

[0128] The result acquisition module is used to obtain the final methane point source location based on the methane point source filtering results.

[0129] The same or similar labels correspond to the same or similar parts;

[0130] The terms used to describe positional relationships in the accompanying drawings are for illustrative purposes only and should not be construed as limiting this patent.

[0131] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A hyperspectral remote sensing image methane point source detection method based on superpixel matching filter, characterized in that, The method comprises the following steps: Obtaining an original hyperspectral remote sensing image and a standard methane spectral absorption rate; Obtaining a true color image according to the original hyperspectral remote sensing image, and obtaining a background spectrum according to the true color image; Constructing a methane target spectrum according to the standard methane spectral absorption rate and the background spectrum; Filtering the methane target spectrum and the original hyperspectral image by using a matching filter algorithm to obtain a methane point source filtering result; Obtaining a maximum possible methane point source position according to the methane point source filtering result; Obtaining a hyperspectral remote sensing image, and obtaining a true color image according to the hyperspectral remote sensing image, which comprises synthesizing the hyperspectral remote sensing image through RGB bands to obtain the true color image; Obtaining a background spectrum according to the true color image, which comprises segmenting a background block of the true color image by using a superpixel segmentation algorithm, and aggregating the segmented blocks to obtain the background spectrum; Obtaining a standard methane spectral absorption rate, which comprises: Based on a MODTRAN platform, selecting a 1976 US standard atmospheric temperature curve as a median atmospheric model for simulation to obtain absorption characteristics of methane in a short-wave infrared range as the standard methane spectral absorption rate; Filtering the methane target spectrum and the original hyperspectral image by using a matching filter algorithm to obtain a methane point source filtering result, which comprises: Wherein, P is the methane point source distribution probability, that is, the methane point source detection filtering result, x p is the methane enhancement value of the methane point source, and the probability processing result is the methane point source detection filtering result of the to-be-detected data Obtaining a maximum possible methane point source position according to the methane point source filtering result, which comprises: Sequentially performing Gaussian filtering and eight-neighbor feature estimation on the methane point source filtering result to obtain the maximum possible methane point source position.

2. The hyperspectral remote sensing image methane point source detection method based on superpixel matching filter according to claim 1, characterized in that, The sequential Gaussian filtering and eight-neighbor feature estimation on the methane point source filtering result comprises: Taking all pixel gray values in a 3*3 region of the methane point source filtering result as a center as an average gray value of the center point, using a two-dimensional Gaussian function as a weight distribution mode, and performing convolution operation on the image by using a Gaussian kernel generated by the Gaussian function, specifically: Wherein, x and y represent two-dimensional plane coordinate positions of the pixel, and sigma is a standard deviation, and optimization is performed by using a separation characteristic of the Gaussian function:

3. The hyperspectral remote sensing image methane point source detection method based on superpixel matching filter according to claim 2, characterized in that, The sequential Gaussian filtering and eight-neighbor feature estimation on the methane point source filtering result further comprises: After the Gaussian filtering, eight-neighbor feature estimation is performed on the Gaussian-filtered image by using an eight-connected neighborhood labeling algorithm, and the statistics of pixel values in the eight-connected neighborhood can be expressed as: N8(P)={(x+i,y+j)|i,j∈{-1,0,1}}\(x,y) For any pixel point in the image, the eight-connected neighborhood N8(P) is defined as a set containing P(X,Y) and eight pixel points around P(X,Y): The specific algorithm steps are as follows: 1) Region labeling: traverse each pixel point in the image, and use a labeling function f(P) to record the region to which each pixel point belongs; 2) Isolated point detection: for each pixel point P(X,Y), check the eight-connected neighborhood N8(P) of P(X,Y), if P belongs to a specific category and there is no other pixel point belonging to the same category in the eight-connected neighborhood of P, then P is considered as an isolated point; 3) Isolated point removal: set the pixel value of the detected isolated point P to a background value or an average value of adjacent pixels, and here the isolated point is set to the background value, which can be mathematically expressed as: Let I(X,Y) represent the pixel value at coordinates (X,Y) in the image, and B(X,Y) represent the background pixel value; for each pixel P(X,Y), the statistics of the pixel values ​​in its 8-connected neighborhood can be expressed as: If P is an isolated point, then S(P) will not contain pixel values ​​of the same category as P; in this case, the pixel values ​​of P can be updated as follows:

4. A hyperspectral remote sensing image methane point source detection system based on superpixel matching filter, for implementing the method of any one of claims 1-3, characterized in that, include: The data acquisition module is used to acquire raw hyperspectral remote sensing images and standard methane spectral absorbance. The true-color image and background spectrum acquisition module is used to obtain a true-color image from the original hyperspectral remote sensing image and to obtain a background spectrum from the true-color image. The target spectrum construction module is used to construct the target spectrum of methane based on the standard methane spectral absorbance and background spectrum. The filtering module is used to filter the methane target spectrum and the original hyperspectral image using a matched filtering algorithm to obtain the methane point source filtering result; The result acquisition module is used to obtain the location of the most likely methane point source based on the methane point source filtering results. Acquiring hyperspectral remote sensing images and obtaining true-color images from them includes synthesizing hyperspectral remote sensing images using RGB bands to obtain true-color images; Obtaining the background spectrum from a true-color image involves segmenting the background plots of the true-color image using a superpixel segmentation algorithm, and then performing superpixel aggregation on the segmented plots to obtain the background spectrum. Obtaining the standard methane spectral absorbance includes: Based on the MODTRAN platform, the 1976 US standard atmospheric temperature curve was selected as the median atmospheric model for simulation to obtain the absorption characteristics of methane in the shortwave infrared range, which was used as the standard methane spectral absorbance. The matched filtering algorithm is used to filter the methane target spectrum and the original hyperspectral image to obtain the methane point source filtering results, including: Wherein, P is the methane point source distribution probability, that is, the methane point source detection filtering result, x p is the methane enhancement value of the methane point source, and the probability processing result is the methane point source detection filtering result of the to-be-detected data Based on the methane point source filtering results, the most likely methane point source locations include: The methane point source filtering results are sequentially subjected to Gaussian filtering and eight-neighbor feature estimation to obtain the location of the most likely methane point source.

5. A computer-readable storage medium having stored thereon a computer program, characterized in that When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-3.

Images