A high-resolution remote sensing method for tailings dam identification that integrates texture and spectral features

Abstract

This invention discloses a high-resolution remote sensing tailings dam identification method that integrates texture and spectral features, relating to the field of tailings dam identification technology. The method includes the following steps: acquiring multispectral remote sensing images of known tailings dams, named as sample remote sensing images; performing band analysis on the sample remote sensing images to extract surface reflectance in different bands and form sample spectral features; extracting pixel grayscale values ​​from the sample remote sensing images; performing feature extraction on the sample remote sensing images based on pixel grayscale values ​​to obtain sample texture features of the tailings dam; performing preliminary identification of the tailings dam based on the sample spectral features, and then performing final identification of the tailings dam based on the sample texture features. This invention addresses the shortcomings of existing tailings dam identification technologies, which fail to effectively extract tailings dam features and struggle to search and identify tailings dams over large areas, resulting in insufficient accuracy and low practicality in tailings dam identification.

Description

Technical Field

[0001] This invention relates to the field of tailings dam identification technology, specifically a high-resolution remote sensing tailings dam identification method that integrates texture and spectral features. Background Technology

[0002] Tailings dam identification technology refers to a technology that comprehensively utilizes remote sensing technology, geographic information systems, and computer intelligent analysis methods to automatically or semi-automatically detect, extract, identify, and monitor the spatial location, extent, and status information of tailings dams from remote sensing images or other geospatial data.

[0003] Existing tailings dam identification technologies typically employ image recognition. This process requires extracting and learning the image features of the tailings dam to provide a basis for identification. However, current tailings dam identification technologies do not effectively extract these features, making it difficult for deep learning to obtain sufficient training data. Furthermore, existing tailings dam identification technologies struggle to search and identify tailings dams over large areas, only able to identify them within a small range of images. The lack of effective feature extraction also results in low accuracy. For example, in publication number CN11488... Patent application 2375A discloses "a method and apparatus for intelligent identification of tailings dams". This scheme preprocesses remote sensing images and trains a deep learning target detection model, and then identifies tailings dams through the deep learning target detection model. However, it does not provide the learning data used for deep learning, nor does it extract features from the remote sensing images. Therefore, the feasibility of the scheme cannot be guaranteed. Existing tailings dam identification technologies also suffer from the inability to effectively extract the features of tailings dams and the difficulty in searching and identifying tailings dams over a large area, resulting in insufficient accuracy and low practicality in tailings dam identification. Summary of the Invention

[0004] This invention aims to at least partially address one of the technical problems in the existing technology. It acquires multispectral remote sensing images of known tailings ponds, naming them "sample remote sensing images," and then performs band analysis on these images to extract the density (DN) values ​​of different bands. These DN values ​​are then converted to surface reflectance, and the spectral characteristics of the samples are analyzed. Next, pixel grayscale values ​​are extracted from the sample remote sensing images, and feature extraction is performed based on these grayscale values ​​to obtain the individual texture features. The texture features of the tailings ponds are then analyzed based on these individual texture features. Finally, preliminary identification of the tailings ponds is performed based on the spectral features, followed by final identification based on the texture features. This addresses the shortcomings of existing tailings pond identification technologies, such as the ineffective extraction of tailings pond features and the difficulty in searching and identifying tailings ponds over large areas, resulting in insufficient accuracy and low practicality.

[0005] To achieve the above objectives, this application provides a high-resolution remote sensing tailings dam identification method that integrates texture and spectral features, comprising the following steps: Acquire multispectral remote sensing images of known tailings ponds and name them as sample remote sensing images; Band analysis was performed on the sample remote sensing images to extract the surface reflectance of different bands in the sample remote sensing images and form the sample spectral characteristics; Extract pixel grayscale values ​​from the sample remote sensing image; Feature extraction is performed on the remote sensing images of the samples based on pixel grayscale to obtain the sample texture features of the tailings pond; Tailings ponds are initially identified based on sample spectral features, and then finally identified based on sample texture features.

[0006] Furthermore, obtaining multispectral remote sensing images of known tailings ponds specifically involves obtaining multispectral remote sensing images of tailings ponds from mining enterprises, which are named sample remote sensing images. The sample remote sensing images only contain images within the tailings pond area and do not contain images within other background areas.

[0007] Further, band analysis is performed on the sample remote sensing image to extract the surface reflectance of different bands in the sample remote sensing image and form the sample spectral features, including the following sub-steps: Band analysis was performed on the sample remote sensing images to extract the DN values ​​of different bands in the sample remote sensing images; The DN values ​​were converted to surface reflectance and the spectral characteristics of the samples were analyzed.

[0008] Further, band analysis is performed on the sample remote sensing image to extract the DN values ​​of different bands in the sample remote sensing image, including the following sub-steps: Each pixel in the multispectral remote sensing image contains DN values ​​for both the visible light band and the near-infrared band, and the DN values ​​of different pixels are different. The visible light bands include red, orange, yellow, green, blue, indigo, and violet bands; The DN values ​​of the red, orange, yellow, green, blue, indigo, violet, and near-infrared bands are encoded and represented by the symbols RB, OB, YB, GB, BB, IB, VB, and UB, respectively.

[0009] Further, converting DN values ​​to surface reflectance and analyzing sample spectral characteristics includes the following sub-steps: Obtain the calibration coefficients of the sample remote sensing image, wherein the calibration coefficients include gain and offset, represented by the symbols GN and OT, respectively; The atmospheric top reflectance corresponding to the DN value is calculated using the formula TOA=α×GN+OT, where TOA is the atmospheric top reflectance and α is RB, OB, YB, GB, BB, IB, VB or UB. The imaging time, location, altitude, aerosol type, and aerosol concentration of the sample remote sensing image are obtained. The imaging time, location, altitude, aerosol type, and aerosol concentration are input into the FLAASH atmospheric correction model, and the error reflectance is output, represented by the symbol ER. Calculate TOA-ER to obtain the surface reflectance of the sample remote sensing image. Calculate the surface reflectance for each pixel in the sample remote sensing image, where each pixel in the sample remote sensing image has a surface reflectance in each visible light band and near-infrared band. The surface reflectances corresponding to RB, OB, YB, GB, BB, IB, VB and UB are respectively labeled as RR, OR, YR, GR, BR, IR, VR and UR. RR, OR, YR, GR, BR, IR, VR and UR are collectively referred to as band reflectances. When analyzing reflectance in any band, it is named the target analytical reflectance. The target analytical reflectance of all pixels in all sample remote sensing images is obtained. Cluster analysis is performed on the target analytical reflectance to obtain different reflectance clusters. The cluster centers are obtained, and the reflectance clusters are sorted and numbered in ascending order of their cluster centers, using the symbol RC. n This indicates that n is a non-zero natural number and n is the index of RC; Obtain the RC value of the band reflectivity n The sequence number n is labeled as SN. SN is combined into a set of numbers in the order of RR, OR, YR, GR, BR, IR, VR and UR, and named as pixel spectral distribution feature. Each pixel in the sample remote sensing image has a pixel spectral distribution feature. The set of all pixel spectral distribution features is the sample spectral feature.

[0010] Further, extracting pixel grayscale values ​​from the sample remote sensing image includes the following sub-steps: The pixel in the i-th row and j-th column of the sample remote sensing image is labeled as P(i,j), where i and j are both non-zero natural numbers and (i,j) is the index of P; Convert the sample remote sensing image into a grayscale image, obtain the grayscale value of P(i,j), name it pixel grayscale, and use the symbol HG(i,j).

[0011] Furthermore, feature extraction based on pixel grayscale of the remote sensing images of the tailings pond samples to obtain sample texture features includes the following sub-steps: Feature extraction is performed on the sample remote sensing image based on pixel grayscale to obtain the individual texture features of the sample remote sensing image; Analysis of sample texture features in tailings ponds based on individual texture features.

[0012] Furthermore, feature extraction is performed on the sample remote sensing image based on pixel grayscale to obtain the individual texture features of the sample remote sensing image, including the following sub-steps: Statistical analysis is performed on HG(i,j), and HG(i,j) with the same value are grouped into a gray-level group. The HG(i,j) in the gray-level group is named the group gray-level value. The grayscale groups are numbered and labeled as GV in ascending order of their grayscale values. m , where m is a non-zero natural number and m is the index of GV; With m as the X-axis, GV m Establish a two-dimensional coordinate system for the Y-axis, named the texture feature distribution map, and then use GV... m Enter the texture feature distribution map according to m, and name the coordinate points in the texture feature distribution map as texture feature distribution points; Linear regression is performed on the texture feature distribution points to obtain the slope of the linear regression function, which is named the individual texture feature. Each of the sample remote sensing images corresponds to one individual texture feature.

[0013] Furthermore, the analysis of the sample texture features of the tailings dam based on individual texture features includes the following sub-steps: Obtain GV from each sample remote sensing image max(m) Calculate GV max(m) -GV1 marks the calculation result as KC, where max() is the maximum value operator; The range of individual texture features in remote sensing images of all samples with the same KC is statistically analyzed and labeled as KL; The sample texture features are composed of different KCs and their corresponding KLs.

[0014] Furthermore, the preliminary identification of tailings ponds based on sample spectral features, followed by the final identification based on sample texture features, includes the following sub-steps: Acquire multispectral remote sensing images captured in real time and name them as real-time images; Analyze the pixel spectral distribution characteristics of different pixels in the real-time image and name them as real-time spectral distribution characteristics. Check whether the real-time spectral distribution characteristics exist in the sample spectral characteristics. If they do, the pixels are classified as suspicious points; otherwise, the pixels are classified as background points. Consecutive adjacent suspicious points are integrated into a region and named the suspicious region. The individual texture features of the pixels in the suspicious region are analyzed and named the real-time texture features. At the same time, the KC of the real-time texture features is obtained and marked as HK. The KL of the sample texture features with the same KC as HK is found and named the reference range. It is determined whether the real-time texture features are within the reference range. If so, the suspicious region is marked as a tailings pond. Otherwise, the pixels in the suspicious region are classified as background points.

[0015] The beneficial effects of this invention are as follows: This invention acquires multispectral remote sensing images of known tailings ponds, names them as sample remote sensing images, and then performs band analysis on the sample remote sensing images to extract the DN values ​​of different bands in the sample remote sensing images. The DN values ​​are then converted into surface reflectance and the spectral characteristics of the samples are analyzed. The advantage is that the spectral curves of tailings in tailings ponds are significantly different from those of the surrounding natural surface in certain bands. Therefore, tailings and natural surfaces can be distinguished through spectral characteristics, which improves the accuracy and effectiveness of tailings pond identification. This invention extracts pixel grayscale values ​​from sample remote sensing images, then performs feature extraction based on these grayscale values ​​to obtain individual texture features of the sample remote sensing images. These individual texture features are then used to analyze the texture features of tailings ponds. Finally, preliminary identification of the tailings ponds is performed based on sample spectral features, followed by final identification based on sample texture features. The advantage lies in the fact that while preliminary identification using spectral features still has some errors, introducing sample texture features for secondary identification addresses this. Since tailings accumulation is not uniform, different grooves form on the tailings surface, and the tailings themselves differ in color from the surrounding surface. Furthermore, the use of various chemicals in tailings ponds causes specific color changes, further improving the accuracy and effectiveness of tailings pond identification. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of the method of the present invention; Figure 2 This is a schematic diagram of the texture feature distribution map of the present invention; Figure 3 This is a grayscale image of the real-time image of the present invention; Figure 4 This is a schematic diagram of the questionable area of ​​the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1 As shown, this application provides a high-resolution remote sensing tailings dam identification method that integrates texture and spectral features, including the following steps: Step S1: Obtain multispectral remote sensing images of known tailings ponds and name them as sample remote sensing images. Specifically, obtaining multispectral remote sensing images of known tailings ponds involves obtaining multispectral remote sensing images of tailings ponds from mining enterprises and naming them as sample remote sensing images. The sample remote sensing images only contain images within the tailings pond area and do not contain images within other background areas. In practice, the remote sensing images of the samples are provided by the mining companies, and this embodiment will not provide specific details.

[0019] Step S2 involves performing band analysis on the sample remote sensing image, extracting the surface reflectance of different bands in the sample remote sensing image, and constructing the sample spectral features; Step S2 includes the following sub-steps: Step S201: Perform band analysis on the sample remote sensing image and extract the DN values ​​of different bands in the sample remote sensing image; Step S201 includes the following sub-steps: Step S2011: Each pixel in the multispectral remote sensing image contains DN values ​​for the visible light band and the near-infrared band, and the DN values ​​of different pixels are different. Step S2012, the visible light band includes the red band, orange band, yellow band, green band, blue band, indigo band and violet band; Step S2013: Encode the DN values ​​of the red band, orange band, yellow band, green band, blue band, indigo band, violet band and near-infrared band, and represent them by the symbols RB, OB, YB, GB, BB, IB, VB and UB respectively. In practice, the DN value is the brightness value of a pixel in a certain band. It is known data in the multispectral remote sensing image and can be directly obtained. The red band, orange band, yellow band, green band, blue band, indigo band, and violet band are respectively labeled as RB, OB, YB, GB, BB, IB, VB, and UB.

[0020] Step S202: Convert the DN values ​​to surface reflectance and analyze the spectral characteristics of the samples; Step S202 includes the following sub-steps: Step S2021: Obtain the calibration coefficients of the sample remote sensing image. The calibration coefficients include gain and offset, which are represented by the symbols GN and OT, respectively. Step S2022: Calculate the atmospheric top reflectivity corresponding to the DN value using the formula TOA=α×GN+OT, where TOA is the atmospheric top reflectivity and α is RB, OB, YB, GB, BB, I, VB or UB. Step S2023: Obtain the imaging time, location, altitude, aerosol type, and aerosol concentration of the sample remote sensing image. Input the imaging time, location, altitude, aerosol type, and aerosol concentration into the FLAASH atmospheric correction model and output the error reflectance, represented by the symbol ER. Step S2024: Calculate TOA-ER to obtain the surface reflectance of the sample remote sensing image. Calculate the reflectance of each pixel in the sample remote sensing image. Each pixel in the sample remote sensing image has a surface reflectance in each visible light band and near-infrared band. In practice, the calibration coefficients are related to the satellite that captured the remote sensing image of the sample, and are usually inherent parameters of the satellite. The gain GN and offset OT of the obtained sample remote sensing image are 0.00002 and -0.1, respectively. Taking a pixel in a sample remote sensing image as an example, this embodiment names it as the target pixel. The DN value of the target pixel in the blue band is obtained as 15000, i.e., BB=15000. Replacing α with BB, the atmospheric top reflectance TOA is calculated as 15000×0.00002-0.1=0.2. Since the FLAASH atmospheric correction model is a known model used to eliminate radiation errors caused by atmospheric absorption and scattering and to invert the true surface reflectance, it is an existing model and will not be specifically described in this embodiment. After inputting the imaging time, location, altitude, aerosol type, and aerosol concentration into the FLAASH atmospheric correction model, the output error reflectance is 0.05, which means that the atmosphere itself contributes approximately 0.05 to the target pixel. After removing the reflectance, the surface reflectance is equal to TOA-ER=0.2-0.05=0.15, which means that the ground object represented by the target pixel has a reflectance of 15% for blue light. Each pixel in each sample remote sensing image has a surface reflectance in each visible light band and near-infrared band. For example, there are 100 sample remote sensing images, each of which contains 2,000,000 pixels. These 2,000,000 pixels have a surface reflectance in each of the eight bands: red, orange, yellow, green, blue, indigo, violet, and near-infrared.

[0021] Step S2025: Mark the surface reflectances corresponding to RB, OB, YB, GB, BB, IB, VB and UB as RR, OR, YR, GR, BR, IR, VR and UR respectively, and refer to RR, OR, YR, GR, BR, IR, VR and UR collectively as band reflectances; Step S2026: When analyzing reflectance in any band, name it the target analysis reflectance. Obtain the target analysis reflectance of all pixels in all sample remote sensing images. Perform cluster analysis on the target analysis reflectance to obtain different reflectance clusters. Obtain the cluster centers of the reflectance clusters and sort and number the reflectance clusters in ascending order of their cluster centers, using the symbol RC. n This indicates that n is a non-zero natural number and n is the index of RC; Step S2027: Obtain the RC value of the band reflectivity. n The sequence number n is labeled as SN. SN is combined into a set of numbers in the order of RR, OR, YR, GR, BR, IR, VR and UR, and named as pixel spectral distribution feature. Each pixel in the sample remote sensing image has a pixel spectral distribution feature. The set of all pixel spectral distribution features is the sample spectral feature. In specific implementation, for example, if the surface reflectance of the target pixel in the blue band is 0.15, it can be labeled as BR=0.15. RR, OR, YR, GR, IR, VR, and UR are similarly defined and will not be further explained in this embodiment. Assuming the target reflectance is BR, this means performing cluster analysis on the surface reflectance of 2,000,000 pixels in the blue band for each of the 100 sample remote sensing images. Since the surfaces of different areas within the tailings dam have different states and various characteristic compositions, cluster analysis is needed to distinguish the characteristic compositions under different states. For example, the surface reflectance of iron-rich tailings slag is higher in the red band but lower in the blue band, while a certain agent is higher in the blue band but lower in the red band. Within the blue band, the surface reflectance is relatively low. By analyzing the surface reflectance in eight different bands, different characteristic compositions can be obtained, allowing for accurate identification of tailings dams. Cluster analysis yields five clusters, indicating that the surface reflectance of tailings dams in the blue band typically clusters within these five clusters, numbered RC1 to RC5. The cluster center is the mean of all sample points within a cluster, which can be directly obtained. RC1 to RC5 can then be approximated as extremely low, low, medium, high, and extremely high levels, respectively. When the surface reflectance of a target pixel in the blue band falls within RC1, it indicates that the surface reflectance of the corresponding land feature in the blue band is extremely low. Similarly, the RC values ​​in the red, orange, yellow, green, blue, indigo, purple, and near-infrared bands can be analyzed. nFor example, the RC values ​​of the target pixel points RR, OR, YR, GR, BR, IR, VR, and UR listed in this embodiment. n The corresponding numbers are RC3, RC6, RC2, RC4, RC1, RC3, RC5, and RC3, with corresponding SNs of 3, 6, 2, 4, 1, 3, 5, and 3, respectively. Combining them in order, the pixel spectral distribution feature is 36241353. By statistically analyzing the pixel spectral distribution features of 2,000,000 pixels in each of the 100 sample remote sensing images and forming a set, the sample spectral features can be obtained, representing the composition characteristics of the surface reflectance of the tailings dam surface in different spectral bands.

[0022] Step S3: Extract pixel grayscale values ​​from the sample remote sensing image; Step S3 includes the following sub-steps: Step S301: Mark the pixel in the i-th row and j-th column of the sample remote sensing image as P(i,j), where i and j are both non-zero natural numbers and (i,j) is the index of P; Step S302: Convert the sample remote sensing image into a grayscale image, obtain the grayscale value of P(i,j), name it pixel grayscale, and use the symbol HG(i,j); In practice, for example, if a pixel is located in the 2nd row and 326th column, corresponding to P(2,326), and the grayscale value of that pixel is 128, then HG(2,326) is 128. Step S4 involves extracting features from the remote sensing image of the tailings pond based on pixel grayscale values ​​to obtain the sample texture features. Step S4 includes the following sub-steps: Step S401: Extract features from the sample remote sensing image based on pixel grayscale to obtain the individual texture features of the sample remote sensing image; Step S401 includes the following sub-steps: Step S4011: Statistically analyze HG(i,j), group HG(i,j) with the same value into a gray group, and name the HG(i,j) in the gray group as the group gray value. Step S4012: Number the grayscale groups and label them as GV according to the grayscale values ​​of the groups from smallest to largest. m , where m is a non-zero natural number and m is the index of GV; Please see Figure 2 As shown, in step S4013, with m as the X-axis, GV m Establish a two-dimensional coordinate system for the Y-axis, named the texture feature distribution map, and then use GV... m Enter the texture feature distribution map according to m, and name the coordinate points in the texture feature distribution map as texture feature distribution points; Step S4014: Perform linear regression on the texture feature distribution points, obtain the slope of the linear regression function, and name it as a single texture feature. Each sample remote sensing image corresponds to a single texture feature. In practice, statistically grouping gray values ​​actually means identifying which gray values ​​exist in the remote sensing image of a sample. For example, if a sample remote sensing image contains a total of 26 gray values, that is, 26 groups of gray values, these groups are labeled in ascending order of gray value to obtain GV1 to GV2. 26 The texture feature distribution map is constructed as follows: Figure 2 As shown, the regression yielded a single-unit texture feature of 5.4797.

[0023] Step S402: Analyze the sample texture features of the tailings pond based on individual texture features; Step S402 includes the following sub-steps: Step S4021: Obtain the GV value in each sample remote sensing image. max(m) Calculate GV max(m) -GV1 marks the calculation result as KC, where max() is the maximum value operator; Step S4022: Calculate the range of individual texture features of all remote sensing images with the same KC, and label them as KL; Step S4023: Sample texture features are composed of different KCs and their corresponding KLs; In specific implementation, Figure 2 For example, Figure 2 GV in max(m) That is, GV 26 Among them, GV1 is 112, and GV 26 The value is 240, and KC is calculated to be 128, which means that the grayscale value range in the sample remote sensing image is 240. Because the larger the grayscale value range, the richer the colors in the tailings pond are. The individual texture features calculated for different ranges are quite different. Therefore, it is necessary to perform independent statistics on different KC values. For example, when KC=128, the KL value for 100 sample remote sensing images with KC equal to 128 is [5.2368, 5.5642]. By calculating the KL value corresponding to different KC values, the texture features of the samples can be obtained.

[0024] Step S5 involves preliminary identification of the tailings pond based on sample spectral features, followed by final identification based on sample texture features. Step S5 includes the following sub-steps: Please see Figure 3 As shown, in step S501, the multispectral remote sensing image obtained in real time is acquired and named as a real-time image; Step S502: Analyze the pixel spectral distribution characteristics of different pixels in the real-time image, name them as real-time spectral distribution characteristics, and check whether the real-time spectral distribution characteristics exist in the sample spectral characteristics. If they do, the pixels are classified as suspicious points; if not, the pixels are classified as background points. Please see Figure 4 As shown, in step S503, consecutive adjacent suspicious points are integrated into a region and named a suspicious region. The individual texture features of the pixels in the suspicious region are analyzed and named real-time texture features. At the same time, the KC of the real-time texture features is obtained and marked as HK. The KL of the KC that is the same as HK in the sample texture features is found and named as the reference range. It is determined whether the real-time texture features are within the reference range. If so, the suspicious region is marked as a tailings pond; otherwise, the pixels in the suspicious region are classified as background points. In practical implementation, for example, the grayscale image of the captured real-time image, such as... Figure 3 As shown, pixel spectral distribution features of different pixels in the real-time image are extracted to obtain real-time spectral distribution features. For example, the real-time spectral distribution feature of a certain pixel is 36241353, which exists within the sample spectral features. Therefore, this pixel is classified as a suspicious point. All pixels are analyzed, and consecutively adjacent suspicious points are integrated into a suspicious region. If a suspicious point is located in the eight-neighborhood of another suspicious point, the two suspicious points are considered adjacent. The suspicious region obtained from the analysis is as follows: Figure 4 As shown, analysis Figure 4 The individual texture features were analyzed, resulting in a real-time texture feature value of 5.3357. Figure 4 The corresponding KC is 128, i.e., HK=128. The KL when KC equals 128 is found to be [5.2368, 5.5642], i.e., the reference range is [5.2368, 5.5642]. Since the real-time texture features are within the reference range, the questionable area is marked as a tailings pond.

[0025] Example 2: This application provides an electronic device, which may include a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The memory stores computer-readable instructions, and the processor can call these instructions. When the processor executes a computer-readable instruction, it performs steps such as those in the high-resolution remote sensing tailings dam identification method that fuses texture and spectral features, to achieve the following functions: acquiring a multispectral remote sensing image of a known tailings dam, named a sample remote sensing image; performing band analysis on the sample remote sensing image, extracting the surface reflectance of different bands in the sample remote sensing image, and forming sample spectral features; extracting pixel grayscale values ​​from the sample remote sensing image; performing feature extraction on the sample remote sensing image based on pixel grayscale values ​​to obtain sample texture features of the tailings dam; performing preliminary identification of the tailings dam based on the sample spectral features, and then performing final identification of the tailings dam based on the sample texture features.

[0026] Furthermore, when the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0027] Example 3: This application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the high-resolution remote sensing tailings dam identification method that integrates texture and spectral features provided by the above methods. The method includes: acquiring a multispectral remote sensing image of a known tailings dam and naming it a sample remote sensing image; performing band analysis on the sample remote sensing image to extract the surface reflectance of different bands in the sample remote sensing image and form sample spectral features; extracting pixel grayscale values ​​from the sample remote sensing image; performing feature extraction on the sample remote sensing image based on pixel grayscale values ​​to obtain sample texture features of the tailings dam; performing preliminary identification of the tailings dam based on the sample spectral features, and then performing final identification of the tailings dam based on the sample texture features.

[0028] Example 4: This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it performs the steps of the above-described high-resolution remote sensing tailings dam identification method that integrates texture and spectral features to achieve the following functions: acquiring multispectral remote sensing images of known tailings dams and naming them sample remote sensing images; performing band analysis on the sample remote sensing images to extract the surface reflectance of different bands in the sample remote sensing images and composing sample spectral features; extracting pixel grayscale values ​​from the sample remote sensing images; performing feature extraction on the sample remote sensing images based on pixel grayscale values ​​to obtain sample texture features of the tailings dams; performing preliminary identification of the tailings dams based on the sample spectral features, and then performing final identification of the tailings dams based on the sample texture features.

[0029] Based on the above description of the embodiments, the embodiments of the present invention can be provided as methods, systems, or computer program products. Based on this understanding, the above technical solutions, in essence or in terms of their contribution to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or certain parts of the embodiments.

[0030] In the embodiments provided in this application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules or units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces. The indirect coupling or communication connection between systems, modules, and units may be electrical, mechanical, or other forms.

[0031] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A high-resolution remote sensing method for identifying tailings ponds by fusing texture and spectral features, characterized in that, Includes the following steps: Acquire multispectral remote sensing images of known tailings ponds and name them as sample remote sensing images; Band analysis was performed on the sample remote sensing images to extract the surface reflectance of different bands in the sample remote sensing images and form the sample spectral characteristics; Extract pixel grayscale values ​​from the sample remote sensing image; Feature extraction is performed on the remote sensing images of the samples based on pixel grayscale to obtain the sample texture features of the tailings pond; Tailings ponds are initially identified based on sample spectral features, and then finally identified based on sample texture features. Performing band analysis on the sample remote sensing image to extract the surface reflectance of different bands and compose the sample spectral features includes the following sub-steps: Band analysis is performed on the sample remote sensing image to extract the DN values ​​of different bands in the sample remote sensing image; including: each pixel in the multispectral remote sensing image contains DN values ​​of the visible light band and the near-infrared band, and the DN values ​​of different pixels are different; the visible light band includes red band, orange band, yellow band, green band, blue band, indigo band, and violet band; the DN values ​​of the red band, orange band, yellow band, green band, blue band, indigo band, violet band, and near-infrared band are encoded and represented by the symbols RB, OB, YB, GB, BB, IB, VB, and UB, respectively; Converting DN values ​​to surface reflectance and analyzing sample spectral characteristics includes: obtaining calibration coefficients for remote sensing images of the samples, wherein the calibration coefficients include gain and offset, represented by the symbols GN and OT, respectively; The atmospheric top reflectance corresponding to the DN value is calculated using the formula TOA=α×GN+OT, where TOA is the atmospheric top reflectance and α is RB, OB, YB, GB, BB, IB, VB or UB. The imaging time, location, altitude, aerosol type, and aerosol concentration of the sample remote sensing image are obtained. The imaging time, location, altitude, aerosol type, and aerosol concentration are input into the FLAASH atmospheric correction model, and the error reflectance is output, represented by the symbol ER. Calculate TOA-ER to obtain the surface reflectance of the sample remote sensing image. Calculate the surface reflectance for each pixel in the sample remote sensing image, where each pixel in the sample remote sensing image has a surface reflectance in each visible light band and near-infrared band. The surface reflectances corresponding to RB, OB, YB, GB, BB, IB, VB and UB are respectively labeled as RR, OR, YR, GR, BR, IR, VR and UR. RR, OR, YR, GR, BR, IR, VR and UR are collectively referred to as band reflectances. When analyzing reflectance in any band, it is named the target analytical reflectance. The target analytical reflectance of all pixels in all sample remote sensing images is obtained. Cluster analysis is performed on the target analytical reflectance to obtain different reflectance clusters. The cluster centers are obtained, and the reflectance clusters are sorted and numbered in ascending order of their cluster centers, using the symbol RC. n This indicates that n is a non-zero natural number and n is the index of RC; Obtain the RC value of the band reflectivity n The sequence number n is labeled as SN. SN is combined into a set of numbers in the order of RR, OR, YR, GR, BR, IR, VR and UR, and named as pixel spectral distribution feature. Each pixel in the sample remote sensing image has a pixel spectral distribution feature. The set of all pixel spectral distribution features is the sample spectral feature.

2. The high-resolution remote sensing tailings dam identification method based on the fusion of texture and spectral features according to claim 1, characterized in that, Obtaining multispectral remote sensing images of known tailings ponds specifically involves obtaining multispectral remote sensing images of tailings ponds from mining enterprises, which are named sample remote sensing images. The sample remote sensing images only contain images within the tailings pond area and do not contain images of other background areas.

3. The high-resolution remote sensing tailings dam identification method based on the fusion of texture and spectral features according to claim 2, characterized in that, Extracting pixel grayscale from a sample remote sensing image includes the following sub-steps: The pixel in the i-th row and j-th column of the sample remote sensing image is labeled as P(i,j), where i and j are both non-zero natural numbers and (i,j) is the index of P; Convert the sample remote sensing image into a grayscale image, obtain the grayscale value of P(i,j), name it pixel grayscale, and represent it by the symbol HG(i,j).

4. The high-resolution remote sensing tailings dam identification method based on the fusion of texture and spectral features according to claim 3, characterized in that, Extracting features from remote sensing images of tailings ponds based on pixel grayscale includes the following sub-steps: Feature extraction is performed on the sample remote sensing image based on pixel grayscale to obtain the individual texture features of the sample remote sensing image; Analysis of sample texture features in tailings ponds based on individual texture features.

5. The high-resolution remote sensing tailings dam identification method based on the fusion of texture and spectral features according to claim 4, characterized in that, Extracting features from the sample remote sensing image based on pixel grayscale to obtain the individual texture features of the sample remote sensing image includes the following sub-steps: Statistical analysis is performed on HG(i,j), and HG(i,j) with the same value are grouped into a gray-level group. The HG(i,j) in the gray-level group is named the group gray-level value. The grayscale groups are numbered and labeled as GV in ascending order of their grayscale values. m , where m is a non-zero natural number and m is the index of GV; With m as the X-axis, GV m Establish a two-dimensional coordinate system for the Y-axis, named the texture feature distribution map, and then use GV... m Enter the texture feature distribution map according to m, and name the coordinate points in the texture feature distribution map as texture feature distribution points; Linear regression is performed on the texture feature distribution points to obtain the slope of the linear regression function, which is named the individual texture feature. Each of the sample remote sensing images corresponds to one individual texture feature.

6. The high-resolution remote sensing tailings dam identification method based on the fusion of texture and spectral features according to claim 5, characterized in that, Analyzing the texture features of tailings pond samples based on individual texture features includes the following sub-steps: Obtain GV from each sample remote sensing image max(m) Calculate GV max(m) -GV1 marks the calculation result as KC, where max() is the maximum value operator; The range of individual texture features in remote sensing images of all samples with the same KC is statistically analyzed and labeled as KL; The sample texture features are composed of different KCs and their corresponding KLs.

7. The high-resolution remote sensing tailings dam identification method based on the fusion of texture and spectral features according to claim 6, characterized in that, The preliminary identification of tailings ponds based on sample spectral features, followed by the final identification based on sample texture features, includes the following sub-steps: Acquire multispectral remote sensing images captured in real time and name them as real-time images; Analyze the pixel spectral distribution characteristics of different pixels in the real-time image and name them as real-time spectral distribution characteristics. Check whether the real-time spectral distribution characteristics exist in the sample spectral characteristics. If they do, the pixels are classified as suspicious points; otherwise, the pixels are classified as background points. Consecutive adjacent suspicious points are integrated into a region and named the suspicious region. The individual texture features of the pixels in the suspicious region are analyzed and named the real-time texture features. At the same time, the KC of the real-time texture features is obtained and marked as HK. The KL of the sample texture features with the same KC as HK is found and named the reference range. It is determined whether the real-time texture features are within the reference range. If so, the suspicious region is marked as a tailings pond. Otherwise, the pixels in the suspicious region are classified as background points.

Images