Coal mine fire area anomaly delineation method based on pulse coupled neural network image fusion

Abstract

The application discloses a coal mine fire area anomaly delineation method based on pulse coupled neural network image fusion, first, the measured natural potential and temperature data of a detection area are preprocessed, including data review, gridding registration and image normalization, to obtain standardized natural potential images and temperature images; then, the two source images are decomposed into low-frequency and high-frequency sub-band images by using non-subsampled contourlet transform; then, the low-frequency and high-frequency sub-band images are fused by using a fusion rule based on dynamic weighting and pulse coupled neural network (PCNN); finally, a fused image is obtained through NSCT inverse transformation, and the coal mine fire area range is delineated by comparing with drilling data. Through the multi-scale image fusion technology, the complementary information of the natural potential and the temperature is effectively integrated, the accuracy and reliability of the coal mine fire area anomaly recognition are significantly improved, and a new technical means is provided for the coal mine fire area detection.

Description

Technical Field

[0001] This invention belongs to the field of geophysical exploration and image processing technology, specifically relating to a method for delineating coal mine fire zone anomalies based on pulse-coupled neural network image fusion. Background Technology

[0002] Geophysical exploration plays a vital role in resource exploration and engineering geology due to its non-destructive nature, wide coverage, and ability to acquire multi-parameter information. Coal mine fire zone detection, a crucial aspect of coal mine safety management, effectively infers the spatial distribution of underground fire zones by measuring differences in the natural potential field of the underground medium and the characteristics of surface thermal radiation.

[0003] Currently, the interpretation of parameters used in coal mine fire zone detection, such as spontaneous potential and temperature, largely relies on the subjective interpretation of single images by technicians. This method of interpretation is easily influenced by personal experience, and the interpretation results for the same dataset may differ. Furthermore, it is difficult to quantitatively integrate the comprehensive information reflected by different parameters, thus limiting the objectivity and accuracy of the interpretation.

[0004] Chinese patent application CN120630334A discloses a method and device for detecting coal mine fire zones by fusing natural potential and temperature images, including the following steps: reviewing relevant data to conduct surface geological surveys and obtain geological characteristics of the mining area; conducting exploration of the surface portion of the target area, including methods such as UAV infrared thermometry, UAV aerial mapping, and on-site investigation and temperature verification; detecting the distribution of natural potential and magnetic properties of the fire zone by combining potential and magnetic methods to infer the depth and extent of coal seam combustion; obtaining data on the vertical combustion state, temperature distribution, and geological structure of the coalfield fire zone through drilling verification and borehole temperature measurement; and fusing and identifying the collected information. At the technical level, this patent application essentially remains at the level of macroscopic spatial superposition of multi-source data and manual comprehensive interpretation, lacking in-depth pixel-level or feature-level intelligent image fusion algorithms. It fails to mathematically and rigorously solve the problem of unifying the dimensions and feature scales of heterogeneous data such as natural potential and temperature. In terms of application effect, its delineation of abnormal areas highly depends on manually setting thresholds and manual extraction, which is highly subjective and easily influenced by human experience, leading to discrepancies and multiple interpretations in the interpretation results. At the same time, the method is not capable of adaptively preserving and depicting the complex high-frequency texture details of the fire zone boundary.

[0005] Chinese patent application CN107780971A discloses a method for detecting coalfield fire zones by a stepwise magnetic-electric-thermal source. After determining the direction of underground coal fire spread based on the magnetic anomaly boundary, the magnetic anomaly maximum zone, and the natural potential anomaly boundary of the fire zone detection area, a borehole is drilled. Based on the highest temperature, C2H4 and C2H2 concentrations in the borehole, the combustion center zone, high-temperature zone, and low-temperature zone of the fire zone are determined. Technically, this method does not perform true mathematical algorithm fusion of multi-source geophysical data (such as pixel-level or feature-level fusion). It merely uses different methods to perform preliminary "range reduction" in space. Furthermore, the acquisition of the final fire zone parameters is extremely dependent on destructive and costly dense physical drilling projects, failing to deeply explore and couple the deep hidden features of the geophysical data itself. In terms of application effectiveness, this method not only has a long exploration cycle and high economic cost, but also its fire zone boundary delineation accuracy is completely limited by the borehole mesh density (40-100 meters). It is difficult to objectively, continuously and with high resolution delineate the precise boundaries and fine textures of the fire zone without damage or with minimal drilling. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing methods for interpreting coal mine fire zone detection data, which are characterized by strong subjectivity and insufficient comprehensive utilization of multi-source information. This invention provides a method for delineating coal mine fire zone anomalies based on pulse-coupled neural network image fusion. This method can objectively and accurately fuse natural potential and temperature information and effectively delineate coal mine fire zones.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A method for delineating coal mine fire zone anomalies based on pulse-coupled neural network image fusion includes: (1). Data preprocessing and parameter acquisition: The measured data collected in the exploration area were preprocessed to obtain the spontaneous potential parameters and temperature parameters respectively; the measured geophysical and thermal infrared data collected in the field were cleaned, gridded, registered and normalized to obtain standardized spontaneous potential images and temperature images.

[0008] (2). Apply non-subsampled contour wave transform to decompose the two source images in multiple scales and directions; for the low-frequency sub-band representing the background field, use dynamic weighting method for fusion; for the high-frequency sub-band representing the detailed features, use fusion rules based on pulse coupled neural network (PCNN) for fusion.

[0009] (3). Image reconstruction and fire zone delineation: The fused image is obtained by inverse transformation of NSCT (non-subsampled contour wave transform), and the anomalies in the fused image are interpreted in combination with geological data to delineate the range of coal mine fire zone.

[0010] Further, in step (1), the data preprocessing specifically includes: First, a comprehensive review of the original measurement data is conducted to identify and remove outliers and distorted data caused by instrument malfunction, environmental interference, or human error; then, through interpolation, sorting, and coordinate transformation, data points of different types or different geometric distributions are converted into grid values ​​with the same resolution, thereby ensuring that all data types in the same area can be precisely aligned in a unified coordinate system and strictly registered according to the grid position; subsequently, the standardized grid values ​​are configured with different color band mapping relationships according to the preset physical property combination characteristics and converted into color images to enhance specific fire zone anomaly combinations before fusion; then, the color images are grayscaled to prepare for subsequent fusion operations; finally, the generated grayscale images are normalized to digitize the image information, and the normalized source image is used as the input for image fusion.

[0011] Furthermore, in step (2), the image decomposition part specifically includes: using NSCT (non-subsampled contour wave transform) to decompose the natural potential image I... u and temperature image I t The image is decomposed into low-frequency sub-band images and high-frequency sub-band images. The NSCT transform consists of two stages: the non-subsampled pyramidal filter bank (NSPFB) and the non-subsampled directional filter bank (NSDFB). It can achieve sparse representation of the image in multiple scales and directions. This decomposition process eliminates the downsampling operation, thus possessing translation invariance and accurately capturing the contour and texture information in the image.

[0012] For the natural potential image I u The NSCT decomposition process is defined as D NSCT (I u ) The expression is as follows: , Temperature image I t The NSCT decomposition process is defined as D NSCT (I t The expression is as follows: , in, D NSCT (*) indicates a multi-scale, multi-directional decomposition operation of non-subsampled contour wave transform; k This indicates the number of multiscale decomposition layers in the non-subsampled contour wave transform. H u and L u These represent the natural potential images. I uThe high-frequency directional sub-band image set and the low-frequency sub-band image set are obtained after non-subsampled contour wave transform decomposition; and The images of natural potentials at the 1st and 2nd are respectively represented. k High-frequency subband images and low-frequency subband images corresponding to the scale decomposition layer; H t and L t Temperature images are shown respectively. I t The high-frequency directional sub-band image set and the low-frequency sub-band image set are obtained after non-subsampled contour wave transform decomposition; and The temperature images are respectively represented in the first... k High-frequency subband images and low-frequency subband images corresponding to the scale decomposition layer.

[0013] Furthermore, in step (2), for the low-frequency subband image of the natural potential... L u Low-frequency subband image of temperature L t The fusion rule takes into account the contribution of physical properties in the coal mine fire zone, and therefore adopts a dynamically weighted fusion rule, which is defined as follows: , Where w1 + w2 = 1, the weighting coefficients w1 and w2 can be differentiated according to the physical properties of the target to highlight key information. For example, for coal mine fire zones with high potential anomalies and high temperatures, temperature is given a higher weight (i.e., w2 > w1); thus, a low-frequency sub-band image of the natural potential is obtained. L u Low-frequency subband image of temperature L t The fusion result is FL.

[0014] Furthermore, in step (2), the PCNN fusion rule is used to fuse the high-frequency subband image of natural potential and the high-frequency subband image of temperature; Natural potential high-frequency subband image H u High-frequency subband image H with temperature t It describes the detailed information corresponding to the source image and processes it using a simplified Pulse Coupled Neural Network (PCNN) model; the PCNN model is defined as follows: , in, The coefficients are calculated using a 3×3 sliding window. grayscale value,kl This represents the extent of connection between the pulse-coupled neural network and surrounding neurons. This represents the feedback input of the neuron, equal to ; It is a connection matrix; For the input connection, it equals the total firing time within the neuron's connection range. α L and V L These are its attenuation coefficient and amplification coefficient, respectively; Indicates the overall internal activity items; β It refers to connection strength; It is a threshold. α θ Its attenuation coefficient, V θ It is its amplification factor; n represents the number of calculation iterations; when At this time, the neuron will generate a pulse. , Ignition time; This represents the pulse output state of the neuron during the nth iteration.

[0015] During the iteration process, important information in the image is extracted by analyzing the ignition time of the coefficients; the high-frequency subband fusion decision rule is as follows: , Where FH represents the high-frequency fused image, and The images of spontaneous potential and temperature are respectively in ( x , y ) ignition time of the pulse-coupled neural network neuron at point; , These represent the firing times of PCNN neurons corresponding to the high-frequency subbands of natural potential and temperature, respectively.

[0016] Furthermore, in step (4), the fused low-frequency subband image and high-frequency subband image are reconstructed using NSCT inverse transform to obtain the natural potential image. I u and temperature image I t The IF (Image Fusion Process) formula is as follows: IF=INNSCT(FL,FH) Furthermore, INNSCT represents the inverse NSCT transform. The decomposition filter and directional filter used in this process are consistent with those used in the decomposition process. In step (3), the reconstructed fused image contains the spatial distribution characteristics of natural potential and temperature. By comparing it with known geological information, the credibility and positioning accuracy of coal mine fire zone identification can be significantly improved.

[0017] Image fusion technology provides a new approach for the comprehensive analysis and interpretation of multi-source coal mine fire zone detection data. Unsampled contour wave transform, as an advanced multi-scale geometric analysis tool, possesses translation invariance and excellent orientation and detail capture capabilities, making it highly suitable for fusing geophysical images with complex structural features.

[0018] The beneficial effects of this invention are: 1. By systematically preprocessing data and fusing NSCT images, effective complementarity and integration of natural potential and temperature data were achieved, reducing the ambiguity of interpretation by a single method.

[0019] 2. The fusion rules adopted have clear physical meanings. The weights of low-frequency fusion are adjustable, and high-frequency fusion can adaptively capture edge texture details through PCNN and adaptively select significant features, making the fusion results more in line with the interpretation needs of coal mine fire zone detection.

[0020] 3. This method has a clear process and is highly operable, providing a new, more objective and reliable technical means for delineating the fire zone range in coal mine fire zone detection. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall structure of the present invention; Figure 2 This is a plane map of the normalized natural potential contour lines; Figure 3 This is a normalized temperature contour map. Figure 4 This is a contour map of the fused data obtained based on NSCT. Figure 5 To compare and verify the area map of the ignition point and the area where the borehole is located. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] The structures, proportions, and sizes illustrated in the accompanying drawings are merely for illustrative purposes and to aid those skilled in the art in understanding and reading the invention. They are not intended to limit the scope of the invention and therefore have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to size, provided they do not affect the effectiveness or purpose of the invention, should still fall within the scope of the technical content disclosed herein. Furthermore, the terms "upper," "lower," "left," "right," "middle," and "one" used in this specification are merely for clarity and not intended to limit the scope of the invention. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the invention's implementation.

[0024] The implementation process of this embodiment strictly follows... Figure 1 The diagram shows the overall structure.

[0025] A method for delineating coal mine fire zone anomalies based on pulse-coupled neural network image fusion includes the following steps: Step 1. Data Acquisition and Preprocessing In this embodiment, measured data from a coal mine fire zone exploration area are used. This area is a typical region for spontaneous combustion of coal seams, and the underground fire zone exhibits physical characteristics of low spontaneous potential and high temperature, making it a key target area for coal mine safety management.

[0026] Spontaneous potential detection and ground temperature measurement were carried out in this area, with a grid spacing of 80 meters (line spacing) × 20 meters (point spacing) and 42 survey lines. The raw data obtained were spontaneous potential and surface temperature data.

[0027] Data preprocessing is fundamental to ensuring effective data fusion. The specific steps are as follows: (1) Data review: First, a comprehensive review of the original measured data is conducted to identify and remove outliers and distorted data points caused by poor electrode contact, environmental thermal interference or human error, so as to ensure the reliability of the data source.

[0028] (2) Gridding and registration: Using professional software, the irregularly distributed measurement point data is transformed into regular grid data through Kriging interpolation, and coordinate transformation and unification are performed to ensure that the natural potential and temperature data are strictly spatially registered in the same coordinate system.

[0029] (3) Imaging and Normalization: Standardized grid values ​​were plotted as color contour maps. To facilitate the enhanced display of different property combinations in subsequent fusion analysis, a unified color band scale was used for visualization of the spontaneous potential and temperature images. In particular, to highlight the typical fire zone response pattern of "negative spontaneous potential anomaly and high temperature," the mapping relationship of the color bands was inverted when generating the spontaneous potential contour maps. This artificial setting of the color band mapping made the abnormal areas (negative spontaneous potential anomaly area and high temperature area) related to the target fire zone in the two source images visually consistent before subsequent fusion, both being highlighted features. This method, by flexibly configuring the mapping relationship of the color bands, can also be applied to the pre-fusion enhancement of other property combination patterns, demonstrating its flexibility and specificity in multi-source anomaly collaborative identification. Then, it was converted into a grayscale image. Finally, the grayscale image was normalized to unify the pixel value range to the [0,1] interval, resulting in a standard spontaneous potential source image I with a size of 480×600 pixels. u and temperature source image I t This serves as the input for NSCT image fusion. The normalized image is as follows: Figure 2 and Figure 3 As shown.

[0030] NSCT image fusion NSCT decomposition Preprocessed natural potential image I u and temperature image I t Perform NSCT decomposition.

[0031] Decomposition parameter settings: Based on the image size and detail richness, the number of decomposition layers is set to (4,4,4,4), i.e., a 4-level scale decomposition is performed. The pyramid filter (NSPFB) is set to "maxflat", and the directional filter (NSDFB) is set to "dmaxflat7". This combination achieves excellent multi-scale directional features while maintaining computational efficiency.

[0032] Decomposition process: First, NSPFB performs multi-scale decomposition on the source image to obtain a low-frequency sub-band image (L). u L t The image contains images of high-frequency subbands at four different scales. NSDFB then performs directional decomposition on the high-frequency subbands at each scale. For example, with a directional series vector [2,3,3,4], when the directional series is 2, the high-frequency subband can be decomposed into 2... 2 =4 directional sub-bands; when the directional series is 4, it can be decomposed into 2 4 =16 directional sub-bands. Ultimately, each source image is decomposed into one low-frequency sub-band and a series of high-frequency sub-bands distributed at different scales and in different directions.

[0033] Step 2.2 Subband Image Fusion Low-frequency subband fusion: The low-frequency subband carries the main structural and background field information of the image. A weighted average fusion rule is used: .

[0034] In this embodiment, the target fire zone exhibits characteristics of "negative spontaneous potential anomaly and high temperature," with high temperature being a more direct and crucial indicator of the fire zone. Therefore, the temperature image is given a higher weight in the fusion process, with w1=0.25 (spontaneous potential weight) and w2=0.75 (temperature weight) to better highlight the fire zone information and clearly identify the boundary between the fire zone and the normal surrounding rock.

[0035] High-frequency subband fusion: The high-frequency subband contains detailed information such as image edges and textures. A fusion rule based on a pulse-coupled neural network (PCNN) is adopted.

[0036] PCNN Calculation: A simplified PCNN model is used, with a 3×3 sliding window to process the high-frequency subband coefficients. The gray values ​​of the coefficients are used as external excitations for neurons, and the firing time of the neuron corresponding to each pixel is obtained through iterative calculation.

[0037] , in, The coefficients are calculated using a 3×3 sliding window. grayscale value, kl This represents the extent of connections between the pulse-coupled neural network and surrounding neurons. This represents the feedback input of the neuron, equal to , It is a connection matrix. For the input connection, it equals the total firing time within the neuron's connection range. α L and V L These are its attenuation coefficient and amplification coefficient, respectively. This represents the total activity term, where β is the connection strength. It is a threshold. α θ Its attenuation coefficient, V θ It is its amplification factor, where n represents the number of calculation iterations; when At this time, the neuron will generate a pulse. , Ignition time, This represents the pulse output state of the neuron during the nth iteration.

[0038] In this embodiment, the PCNN parameters are set as follows:α L =0.2, α 𝜃 =0.4, V L =0.1, V θ =20, β =0.2 to enhance the response to local edges and texture details.

[0039] Fusion Decision: For each pixel (x, y), calculate the firing time of the PCNN neuron corresponding to the high-frequency subband of natural potential and the high-frequency subband of temperature, denoted as . and The fusion decision rule is: select the high-frequency coefficients of the source image with a larger ignition time as the high-frequency coefficients after fusion.

[0040] , Where FH represents the high-frequency fused image, and , respectively, represent the ignition time of the pulse-coupled neural network neurons at (x,y) for the natural potential and temperature images.

[0041] This rule can adaptively preserve clearer and more prominent edge and texture details from any source image, effectively improving the recognition accuracy of fire zone boundaries.

[0042] Step 3. Image Reconstruction and Fire Zone Delineation The inverse NSCT transform INNSCT is used to reconstruct the fused low-frequency subband (FL) and high-frequency subband (FH) to obtain the final fused image (IF). The decomposition filter and directional filter used in the reconstruction process are consistent with those used in the decomposition stage to ensure lossless reconstruction of information.

[0043] IF=INNSCT(FL,FH) The reconstructed fused image IF (see Figure 4 It also includes field structure information of spontaneous potential and thermal anomaly information of temperature. The IF is compared and analyzed with prior data such as borehole verification results.

[0044] To further verify the accuracy of the method for delineating fire zone anomalies in this invention, the reconstructed fused image was overlaid and compared with actual borehole data and on-site investigation of fire points, as follows: Figure 5 As shown. In Figure 5 In the fused image, abnormal regions exceeding the judgment coefficient threshold show a high degree of spatial consistency with the locations of verification boreholes and ignition points deployed in the field. Specifically, high-temperature boreholes located at the center of the high-value anomaly delineation and in the adjacent area (such as...) Figure 5The high-temperature boreholes marked in the diagram exhibit strong fire zone responses in their actual internal temperature measurements, with corresponding fusion image judgment coefficients in extremely high ranges (e.g., >0.35), highly consistent with the combustion state of high temperatures throughout or in localized areas. Conversely, conventional boreholes located outside the comprehensive anomaly delineation area show normal-temperature surrounding rock responses in their actual internal temperature measurements, with no coal seam spontaneous combustion observed. This accurately confirms, from the opposite perspective, the low false alarm rate and high resolution of the method of this invention in fire zone boundary delineation. Furthermore, combined with... Figure 5 The photos on the left, showing the on-site investigation of the ignition point, rock strata tensional fractures, and smoke-emitting areas, reveal a remarkably precise "point-area" spatial correspondence between their corresponding surface spatial coordinates and the core area of ​​the anomaly peak with the highest judgment coefficient and largest area in the fused image obtained by this method. This comparative verification result fully demonstrates that the fusion result of this invention highly matches the temperature scale of deep drilling entities and the macroscopic combustion damage phenomena on the surface, significantly improving the positioning accuracy and reliability of the delineation of underground concealed coal mine fire zones.

[0045] The above embodiments fully demonstrate that the method of the present invention can effectively integrate natural potential and temperature data, significantly improving the ability to identify coal mine fire zones and the accuracy of fire zone delineation.

[0046] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for delineating coal mine fire zone anomalies based on pulse-coupled neural network image fusion, characterized in that: Includes the following steps: (1). Data preprocessing and parameter acquisition: The measured data collected in the exploration area were preprocessed to obtain the spontaneous potential parameters and temperature parameters respectively; then the gridded data was generated and the corresponding spontaneous potential image and temperature image were drawn; (2). The natural potential image and temperature image are decomposed by non-subsampled contour wave transform to obtain their respective low-frequency sub-band images and high-frequency sub-band images; and the low-frequency and high-frequency sub-band images are fused by fusion rules based on dynamic weighting and pulse coupling neural network, respectively. (3) The fused image is reconstructed by inverse transformation of non-subsampled contour wave transform, and the fused image is compared with prior geological information to delineate the prospecting target area based on the anomaly area.

2. The method for anomaly delineation of coal mine fire zones based on pulse-coupled neural network image fusion according to claim 1, characterized in that: The data preprocessing in step (1) includes: The raw measured data is reviewed for quality to remove abnormal data points introduced by instrument, environmental or human factors; interpolation and coordinate transformation techniques are used to unify discrete data points to a grid coordinate system of the same resolution and perform accurate registration. Subsequently, the standardized grid values ​​are configured with different color band mapping relationships according to the preset physical property combination characteristics and converted into color images to enhance specific fire zone anomaly combinations before fusion; Next, the color image is converted to grayscale to prepare for subsequent fusion operations. The grid data is visualized and then converted to grayscale image. Finally, the generated grayscale image is normalized to digitize the image information and generate a standardized source image for image fusion input.

3. The method for anomaly delineation of coal mine fire zones based on pulse-coupled neural network image fusion according to claim 1, characterized in that: The specific steps of step (2) include: (2.1). Using non-subsampled contourlet transform, the natural potential image I is transformed into... u and temperature image I t Decomposed into low-frequency sub-band images and high-frequency sub-band images; (2.2). A dynamic weighted fusion rule is used to fuse the low-frequency subbands of the two source images; (2.3). The high-frequency subbands of the two source images are fused using a fusion rule based on pulse-coupled neural network.

4. The method for anomaly delineation of coal mine fire zones based on pulse-coupled neural network image fusion according to claim 3, characterized in that: In step (2.1), the natural potential image I is... u The non-subsampled contour wave transform decomposition process is defined as D NSCT (I u The expression is as follows: ， Temperature image I t The non-subsampled contour wave transform decomposition process is defined as D NSCT (I t The expression is as follows: ， in, D NSCT (*) indicates a multi-scale, multi-directional decomposition operation of non-subsampled contour wave transform; k This indicates the number of multiscale decomposition layers in the non-subsampled contour wave transform. H u and L u These represent the natural potential images. I u The high-frequency directional sub-band image set and the low-frequency sub-band image set are obtained after non-subsampled contour wave transform decomposition; and The images of natural potentials at the 1st and 2nd are respectively represented. k High-frequency subband images and low-frequency subband images corresponding to the scale decomposition layer; H t and L t Temperature images are shown respectively. I t The high-frequency directional sub-band image set and the low-frequency sub-band image set are obtained after non-subsampled contour wave transform decomposition; and The temperature images are respectively represented at the 1st... k High-frequency subband images and low-frequency subband images corresponding to the scale decomposition layer.

5. The method for anomaly delineation of coal mine fire zones based on pulse-coupled neural network image fusion according to claim 3, characterized in that: In step (2.2), the low-frequency sub-band fusion rule adopts a dynamic weighting method, and the formula is as follows: ， Wherein, FL represents the fused low-frequency subband image; This represents the low-frequency sub-band image corresponding to the k-th scale decomposition layer of the natural potential image; The temperature image represents the low-frequency sub-band image corresponding to the k-th scale decomposition layer; w1 and w2 represent the fusion weighting coefficients of the natural potential low-frequency sub-band image and the temperature low-frequency sub-band image, respectively; the weighting coefficients w1 and w2 are set according to the physical property response characteristics of the target anomaly.

6. The method for delineating coal mine fire zone anomalies based on pulse-coupled neural network image fusion according to claim 3, characterized in that: In step (2.3), the definition of the pulse-coupled neural network fusion rule is as follows: ， in, The coefficients are calculated using a sliding window. grayscale value, kl This represents the extent of connection between the pulse-coupled neural network and surrounding neurons. This represents the feedback input of the neuron, equal to ; It is a connection matrix; For the input connection, it equals the total firing time within the neuron's connection range. α L and V L These are its attenuation coefficient and amplification coefficient, respectively; Indicates the overall internal activity items; β It refers to connection strength; It is a threshold. α θ Its attenuation coefficient, V θ It is its amplification factor; n represents the number of calculation iterations; when At this time, the neuron will generate a pulse. , Ignition time; This represents the pulse output state of the neuron during the nth iteration.

7. The method for anomaly delineation of coal mine fire zones based on pulse-coupled neural network image fusion according to claim 6, characterized in that: During the iteration process, important information in the image is extracted by analyzing the ignition time of the coefficients; the high-frequency subband fusion decision rule is as follows: ， Where FH represents the high-frequency fused image, and The images of spontaneous potential and temperature are respectively in ( x , y ) ignition time of the pulse-coupled neural network neuron at point; , These represent the firing times of PCNN neurons corresponding to the high-frequency subbands of natural potential and temperature, respectively.

8. The method for anomaly delineation of coal mine fire zones based on pulse-coupled neural network image fusion according to claim 1, characterized in that: In step (3), the fused low-frequency subband FL and high-frequency subband FH are reconstructed into the final fused image IF using the inverse non-subsampled contourlet transform, i.e.: IF=INNSCT(FL, FH) INNSCT stands for Inverse Non-Subsampled Profile Wave Transform, and the filter used in the reconstruction process is consistent with that used in the decomposition stage.

Images