Geological lithology interpretation method and system for remote sensing image, and storage medium

Abstract

The invention relates to the technical field of remote sensing image interpretation, in particular to a geological lithology interpretation method and system for a remote sensing image and a storage medium, and the method comprises the steps: obtaining the lithology information of a pattern spot of a designated area through a geological map, and carrying out the preliminary lithology clustering analysis through a Landsat-8 remote sensing image in combination with a k-means algorithm; based on the clustering result, when it is determined that multiple categories are covered in the corresponding first pattern spots, multiple reference pattern spots with other known lithology information are introduced, similarity measurement is carried out between the first pattern spots and the reference pattern spots, and the most similar target reference pattern spots are obtained; and based on a maximum likelihood matching method, when the lithology type matching between the target reference pattern spot and the first pattern spot is determined to be successful, spreading the lithology information of the target reference pattern spot to other pattern spots with similar geological backgrounds, and gradually optimizing the lithology interpretation result of the whole region through multiple iterations. By implementing the method, the lithology categories can be accurately matched, and the accuracy and reliability of geological lithology interpretation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image interpretation, and specifically relates to a method, a system and a storage medium for geological lithology interpretation of remote sensing images. Background Art

[0002] Geological lithology interpretation refers to the process of analyzing and discriminating surface lithology types using remote sensing images, which is widely applied in fields such as geological exploration, resource investigation, and disaster warning. Most traditional geological lithology interpretation methods rely on manual or semi-automated processing methods, and require experts to make empirical judgments based on the combination of remote sensing images and geological maps, with low efficiency and limited interpretation accuracy. With the progress of remote sensing technology, automated interpretation has become the key to improving efficiency and accuracy. Although existing automated interpretation methods can achieve rapid and large-scale lithology identification to a certain extent, they still face several challenges, mainly reflected in the following three aspects:

[0003] 1) Insufficient interpretation accuracy: Existing methods mainly rely on simple pixel classification techniques and cannot effectively distinguish different lithology categories. Especially in areas with complex geology, the interpretation results are often affected by image quality and noise, resulting in low accuracy.

[0004] 2) Unclear lithology categories: Although some methods use clustering algorithms (such as the K-means algorithm), the matching relationship between the clustering results and the actual lithology is unclear, resulting in inaccurate assignment of different lithology categories in the classification results.

[0005] 3) Noise interference: There is noise such as clouds, shadows, and illumination changes in remote sensing images. These noise interferences will affect the reliability of the interpretation results. Traditional denoising methods often cannot effectively remove these noises and may lead to misclassification.

[0006] Therefore, how to improve the lithology interpretation accuracy of remote sensing images, especially in the automated interpretation of multi-lithology areas, how to effectively eliminate noise interference, and accurately match the clustering results with the actual lithology, remains a technical problem to be solved urgently in the field of remote sensing geology interpretation. Summary of the Invention

[0007] In order to accurately match lithology categories and improve the accuracy and reliability of geological lithology interpretation, the purpose of the present invention is to provide a method, a system and a storage medium for geological lithology interpretation of remote sensing images. The specific technical solutions adopted are as follows:

[0008] In a first aspect, a method for geological lithology interpretation of remote sensing images disclosed in this application includes:

[0009] S1. Obtain the lithology information of the patches in the specified area from the geological map, and preliminarily conduct lithology clustering analysis by combining the Landsat-8 remote sensing image with the k-means algorithm;

[0010] S2. Based on the clustering results, when it is determined that the corresponding first patch covers multiple categories, introduce multiple reference patches with other known lithology information, and conduct similarity measurement between the first patch and each reference patch to obtain the most similar target reference patch;

[0011] S3. Based on the maximum likelihood matching method, when it is determined that the lithology type matching between the target reference patch and the first patch is successful, spread the lithology information of the target reference patch to other patches with similar geological backgrounds, and through multiple iterations, gradually optimize the lithology interpretation results of the entire area.

[0012] Further, in step S1, the obtaining the lithology information of the patches in the specified area from the geological map and preliminarily conducting lithology clustering analysis by combining the Landsat-8 remote sensing image with the k-means algorithm includes:

[0013] S11. Obtain the geological map of the specified geological interpretation area, and the geological map includes multiple patches indicating different geological units;

[0014] S12. Based on the geological characteristics, determine the lithology types covered in each patch and their corresponding lithology quantities;

[0015] S13. According to the position information of each patch on the geological map, crop the corresponding regional remote sensing image from the Landsat-8 remote sensing image, and preliminarily conduct lithology clustering analysis based on the k-means algorithm on the regional remote sensing image based on the corresponding lithology quantity.

[0016] Further, in step S13, the preliminarily conducting lithology clustering analysis based on the k-means algorithm on the regional remote sensing image based on the corresponding lithology quantity includes:

[0017] S131. Take the lithology quantity in the patch as the initial number of categories K, and randomly select K pixel points from the regional remote sensing image as the initial clustering centers;

[0018] S132. For each of the remaining pixel points in the image, calculate the distance from the pixel point to each clustering center, and assign the pixel point to the category where the nearest clustering center is located to form the corresponding clustering clusters;

[0019] S133. For each clustering cluster, calculate the mean spectral feature of all pixel points in the cluster, and update the initial clustering center based on the mean spectral feature.

[0020] S134. Repeat steps S132 to S133 until the position of the clustering center no longer changes or until the preset maximum number of iterations is reached, and then obtain the corresponding clustering result.

[0021] Further, in step S2, the method further includes: based on the clustering result, when it is determined that only one category is covered in the corresponding second patch, corresponding the category to a known lithology type and assigning a corresponding lithology label.

[0022] Further, in step S2, the introducing multiple reference patches of other known lithology information and performing similarity measurement between the first patch and each reference patch to obtain the most similar target reference patch includes:

[0023] S21. For the first patch and each reference patch, perform dimensionality reduction filtering processing respectively to obtain the dimensionality-reduced feature vectors.

[0024] S22. Based on the dimensionality-reduced feature vectors, calculate the similarity between the first patch and each reference patch by means of cosine similarity measurement, and thereby obtain the most similar target reference patch.

[0025] Further, in step S3, calculate the probability distribution of each category in the first patch under different lithology types through the following formula:

[0026]

[0027] Among them, P(r k |X) represents the probability that the pixel X belonging to the corresponding category in the first patch belongs to the lithology type r k , P(X|r k ) represents the conditional probability of pixel X under the condition of the given lithology type r k , P(r k ) represents the prior probability of the lithology type r k , and P(X) represents the marginal probability of pixel X.

[0028] Further, in step S3, by introducing the normal distribution hypothesis of the spectral characteristics of the image pixels, P(X|r k ) can be expressed as:

[0029]

[0030] Among them, n represents the dimension of the spectral characteristics, μ k represents the mean vector of the lithology type r k , and Σ represents the covariance matrix of the pixel X belonging to the corresponding category.

[0031] Second aspect, the present application also discloses a system for remote sensing image geological lithology interpretation. The system includes a lithology clustering analysis module, a patch similarity measurement module, and a lithology information propagation and optimization module, where:

[0032] The lithology clustering analysis module is used to obtain the lithology information of patches in a specified area through a geological map, and initially perform lithology clustering analysis by combining Landsat-8 remote sensing images with the k-means algorithm;

[0033] The patch similarity measurement module is used to, based on the clustering result, when it is determined that a corresponding first patch contains multiple categories, introduce multiple reference patches with other known lithology information, and perform similarity measurement between the first patch and each reference patch to obtain the most similar target reference patch;

[0034] The lithology information propagation and optimization module is used to, based on the maximum likelihood matching method, when it is determined that the lithology type matching between the target reference patch and the first patch is successful, propagate the lithology information of the target reference patch to other patches with similar geological backgrounds, and through multiple iterations, gradually optimize the lithology interpretation result of the entire area.

[0035] Third aspect, the present application also discloses a storage medium storing a computer program, which when executed by a processor, implements the above-mentioned method for remote sensing image geological lithology interpretation.

[0036] The present invention has the following beneficial effects:

[0037] 1) Through the combination of the k-means clustering and the patch similarity propagation algorithm, it is possible to achieve accurate matching between the lithology categories in the remote sensing image and the actual lithology types. Compared with traditional methods, the present application can more accurately reflect the lithology distribution of each geological patch, avoiding incorrect interpretation caused by unclear classification results.

[0038] 2) Using remote sensing images and geological maps for automatic interpretation, without relying on manual experience or cumbersome manual correction processes, greatly improves the efficiency and automation degree of lithology interpretation. This has important application value for large-scale geological area interpretation, especially in the fields of resource exploration and geological disaster prediction.

[0039] 3) By optimizing the k-means clustering result through the patch similarity propagation algorithm and the maximum likelihood method, it is possible to effectively eliminate the influence of noise on the interpretation result, avoiding problems of misclassification and inaccurate interpretation. This makes the lithology interpretation result more reliable and more adaptable.

[0040] 4) It can not only interpret the lithology of a single polygon, but also comprehensively consider the similarity between polygons under different regional and different types of geological backgrounds to further optimize the interpretation results. This provides strong support for the automatic interpretation of complex geological regions. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following-described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0042] Figure 1 FIG. is a method flow chart of a method for remotely sensed image geological lithology interpretation provided by an embodiment of the present invention;

[0043] Figure 2 FIG. is an example of a geological map, a remotely sensed image map, and a lithology interpretation map;

[0044] Figure 3 FIG. is a system structure diagram of a system for remotely sensed image geological lithology interpretation provided by an embodiment of the present invention;

[0045] Figure 4 FIG. is a structure diagram of a storage medium provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features, and effects of a method, a system, and a storage medium for remotely sensed image geological lithology interpretation proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0048] The following specifically describes the specific solutions of a method, a system, and a storage medium for remotely sensed image geological lithology interpretation provided by the present invention with reference to the accompanying drawings.

[0049] Please refer to Figure 1 , which shows a method flow chart of a method for remotely sensed image geological lithology interpretation provided by an embodiment of the present invention. The method includes:

[0050] Step S1: Obtain the lithology information of the patches in the specified area from the geological map, and initially perform lithology clustering analysis using the Landsat-8 remote sensing image in combination with the k-means algorithm.

[0051] Specifically, in this application, first, each patch in the specified area, the lithology types and the quantity of lithology it represents will be identified and extracted according to the geological map. Subsequently, for each patch, according to its position information on the geological map, the corresponding image area will be cropped from the Landsat-8 remote sensing image. Then, using the k-means clustering algorithm, the number of clustering centers will be set according to the known quantity of lithology in the patch, and the cropped remote sensing image will be clustered, so as to divide each pixel point in the image into multiple categories that match the quantity of lithology.

[0052] Step S2: Based on the clustering result, when it is determined that the corresponding first patch contains multiple categories, introduce multiple reference patches with other known lithology information, and perform similarity measurement between the first patch and each reference patch to obtain the most similar target reference patch.

[0053] Specifically, when it is determined that the corresponding first patch contains multiple categories, it means that the lithology information of this patch is relatively complex and it is difficult to accurately describe it directly by a single category. Therefore, in order to more accurately identify the lithology information of the first patch, this application further proposes a method based on dimensionality reduction filtering processing and cosine similarity measurement. Specifically, this application will perform dimensionality reduction filtering processing on the first patch and each determined reference patch (such as using dimensionality reduction techniques such as principal component analysis PCA and linear discriminant analysis LDA) to extract the feature vectors after dimensionality reduction (these feature vectors can represent the lithology information of the patch more concisely and effectively); then, using these feature vectors, calculate the similarity degree between the first patch and each reference patch through the cosine similarity measurement method, and finally determine the target reference patch that is most similar to the first patch.

[0054] Step S3: Based on the maximum likelihood matching method, when it is determined that the lithology type matching between the target reference patch and the first patch is successful, spread the lithology information of the target reference patch to other patches with similar geological backgrounds, and through multiple iterations, gradually optimize the lithology interpretation result of the entire area.

[0055] Specifically, the present application uses the maximum likelihood method to calculate the probability distribution of each category in the first image patch under different lithology types, and obtains the most likely lithology label for each category by maximizing the likelihood function. Afterwards, the most likely lithology label for each category in the first image patch is matched with the confirmed lithology label represented by the corresponding category in the target reference image patch. If the match is successful, the matching result will be introduced into the similarity propagation mechanism, which will comprehensively consider the geological background, spatial position relationship and similarity of lithology characteristics, and effectively propagate the lithology information of the target reference image patch to other images with similar geological backgrounds. The final interpretation result can be referred to. Figure 2 Among them, the leftmost one is the geological map, the middle one is the remote sensing image map, and the rightmost one is the lithological interpretation map. It can be seen from the figure that in the lithological interpretation map generated by this application, the lithological types of each category are marked with different color area blocks. The final interpretation results can be used for geological exploration, resource assessment and mineral distribution analysis.

[0056] It should be noted that this propagation process is not completed in one go, but needs to be continuously refined through multiple iterations. In each iteration, the lithology label of each patch will be updated and optimized based on the lithology label of the current patch and the lithology information of similar patches. This process will continue until the lithology interpretation results become stable or the preset upper limit of iterations is reached.

[0057] As can be seen from the above, the present application discloses a method for interpreting geological lithology of remote sensing images. On the one hand, through the combination of k-means clustering and the image similarity propagation algorithm, it is possible to achieve accurate matching of the lithology category in the remote sensing image with the actual lithology type. Compared with the traditional method, the present application can more accurately reflect the lithology distribution of each geological image, avoiding the erroneous interpretation caused by unclear classification results. On the other hand, remote sensing images and geological maps are used for automated interpretation, without relying on manual experience or cumbersome manual correction process, which greatly improves the efficiency and automation of lithology interpretation. This has important application value in large-scale geological regional interpretation, especially in the fields of resource exploration and geological disaster prediction. Finally, the k-means clustering results are optimized by the image similarity propagation algorithm and the maximum likelihood method, which can effectively eliminate the influence of noise on the interpretation results and avoid the problems of misclassification and inaccurate interpretation. This makes the lithology interpretation results more reliable and more adaptable. It should be noted that this application can not only interpret the lithology of a single image patch, but also comprehensively consider the similarities between image patches in different regions and different types of geological backgrounds to further optimize the interpretation results. This provides strong support for the automated interpretation of complex geological areas.

[0058] In one embodiment, in step S1, the lithology information of the patches in the specified area is obtained through the geological map, and the lithology clustering analysis is preliminarily carried out by combining the Landsat-8 remote sensing image with the k-means algorithm, including:

[0059] Step S11, obtain the geological map of the specified geological interpretation area, and the geological map includes multiple patches indicating different geological units.

[0060] Among them, each patch in the geological map represents a geological unit, and the lithology types and lithology quantities therein can provide necessary prior knowledge for subsequent classification work.

[0061] Specifically, based on the geological characteristics of the specified geological interpretation area, this application obtains the geological map of this area through professional geological surveys and remote sensing data processing techniques.

[0062] It should be noted that the lithology type information of each patch is provided in the geological map, which is the conclusion drawn based on previous geological explorations and field surveys.

[0063] Step S12, based on the geological characteristics, determine the lithology types covered in each patch and their corresponding lithology quantities.

[0064] Specifically, this application counts the lithology types in each patch based on the geological characteristics. The geological characteristics include texture, color, shape, etc. These characteristics can reflect the physical and chemical properties of different rocks in the geological map, thus helping to distinguish different lithologies. Then, based on the statistical results of the lithology types, the corresponding lithology quantities are further determined. The "lithology quantity" mentioned here refers to the total number of lithology types contained in the patch. Taking the patches numbered A and B in the geological map as examples, when it is statistically determined that the lithology types covered in them are "basalt - sandstone - coarse sand" and "sandstone - fine sand - carbonate rock", the corresponding lithology quantities can be further determined to be both 3 (that is, containing three different lithologies). For details, please refer to Table 1 below:

[0065] Table 1 Statistical table of lithology types and quantities in different patches

[0066] Map Spot Number Lithology Type Lithology Quantity A Basalt - Sandstone - Coarse Sand 3 B Sandstone - Fine Sand - Carbonate Rock 3

[0067] Step S13, according to the position information of each patch on the geological map, crop the corresponding regional remote sensing image from the Landsat-8 remote sensing image, and preliminarily carry out lithology clustering analysis based on the k-means algorithm for the regional remote sensing image based on the corresponding lithology quantity.

[0068] It should be noted that the Landsat-8 remote sensing image provides spectral data for a specified geological interpretation area. Due to its high resolution and rich spectral information, this data can be used for subsequent lithology classification.

[0069] Specifically, this application will use the number of lithologies corresponding to each patch as an input parameter of the k-means algorithm to perform lithology clustering analysis based on the k-means algorithm in the remote sensing image of the area corresponding to the patch. Specifically, this application will, according to the spectral characteristics of the pixels in the remote sensing image, cluster the pixels into a specified number of categories, and the number of these categories matches the number of lithologies in the patch. Among them, each category corresponds to a lithology, thus realizing the automatic identification of lithologies in the remote sensing image.

[0070] In one embodiment, in step S13, the preliminary lithology clustering analysis of the regional remote sensing image based on the corresponding number of lithologies using the k-means algorithm includes:

[0071] Step S13l: Use the number of lithologies in the patch as the initial number of categories K, and randomly select K pixel points from the regional remote sensing image as the initial clustering centers.

[0072] Specifically, this application will determine the initial number of categories K of the k-means algorithm according to the number of lithologies in the patches in the geological map, and randomly select K pixel points from the corresponding regional remote sensing image as the initial clustering centers of the algorithm.

[0073] Step S132: For each of the remaining pixel points in the image, calculate the distance from this pixel point to each clustering center, and assign this pixel point to the category where the nearest clustering center is located to form the corresponding clustering clusters.

[0074] Specifically, for each of the remaining pixel points in the image (i.e., all pixel points except the initial clustering centers), calculate the Euclidean distance from this pixel point to each clustering center, and assign this pixel point to the category k where the nearest clustering center is located to form preliminary clustering clusters. The following two formulas can be specifically referred to:

[0075]

[0076] where M represents the total number of bands in the image, p ij represents the value of pixel point p i in the j-th band, and C kj represents the value of clustering center C k in the j-th band.

[0077] Step S133: For each clustering cluster, calculate the mean value of the spectral characteristics of all pixel points in the cluster, and update the initial clustering center based on the mean value of the spectral characteristics.

[0078] Specifically, in the present application, the mean value of the spectral features corresponding to all pixel points in each clustering cluster is calculated through the following formula, and the initial clustering center is updated based on the obtained mean value of the spectral features to form a new clustering center:

[0079]

[0080] where S k represents the set of pixels belonging to class k, |S k | represents the number of all pixel points in class k, p i represents the pixel point belonging to class k, and C k ′ represents the new clustering center formed after updating.

[0081] Step S134: Repeat steps S132 to S133 until the position of the clustering center no longer changes or when the preset maximum number of iterations is reached, and the corresponding clustering result is obtained.

[0082] Specifically, in each iteration, the present application will reassign pixel points to the nearest clustering center according to the current clustering center and calculate the new clustering center. This process will be repeated until the position of the clustering center no longer changes or when the preset maximum number of iterations is reached, and it is considered that the clustering algorithm converges, and at this time, the iteration will stop.

[0083] In one embodiment, in step S2, the method further includes: based on the clustering result, when it is determined that only one class is covered in the corresponding second patch, corresponding this class to the known lithology types and assigning the corresponding lithology label.

[0084] Specifically, the present application will also check whether there is a single-class situation in each patch, that is, when it is determined that the second patch only contains one class (that is, there is only one class in the clustering result) after k-means clustering, directly correspond this class to the known lithology types one by one. At this time, there is no need to perform further similarity propagation analysis. The present application will directly apply the currently assigned lithology label to this second patch and consider that the lithology of this second patch is the lithology represented by this label.

[0085] In one embodiment, in step S2, introducing a plurality of reference patches with other known lithology information and performing similarity measurement between the first patch and each reference patch to obtain the most similar target reference patch includes:

[0086] Step S21: Perform dimensionality reduction filtering processing on the first patch and each reference patch respectively to obtain the dimensionality-reduced feature vectors.

[0087] Specifically, this application will use appropriate dimensionality reduction techniques (such as principal component analysis PCA, linear discriminant analysis LDA, or other dimensionality reduction algorithms) to perform dimensionality reduction on these high-dimensional features. It should be noted that the purpose of dimensionality reduction is to reduce the complexity of high-dimensional image features, thereby improving the efficiency and accuracy of matching.

[0088] Step S22: Based on the dimensionality-reduced feature vectors, calculate the similarity between the first patch and each reference patch through the cosine similarity metric, and accordingly obtain the most similar target reference patch.

[0089] Specifically, this application calculates the similarity between the first patch and the i-th reference patch through the following formula:

[0090]

[0091] where the symbol "·" represents the dot product of vectors, the symbol "∥·∥" represents the L2 norm of vectors, n represents the feature dimension, x i represents the feature vector of the i-th reference patch after dimensionality reduction, x j represents the feature vector of the first patch after dimensionality reduction, x i,d represents the feature vector of the i-th reference patch in the d-th feature dimension, x j,d represents the feature vector of the first patch in the d-th feature dimension.

[0092] It should be noted that this application will calculate the dot product of the corresponding feature vector pairs in each dimension and accumulate and sum these dot product values. Then, divide the cumulative dot product sum by the product of the norms of the first patch feature vector and the reference patch feature vector to obtain the cosine similarity. Further, it should be noted that this application will use the cosine similarity to evaluate the similarity between the first patch and each reference patch, and select the reference patch with the largest similarity value as the target reference patch.

[0093] In one embodiment, in step S3, calculate the probability distribution of each category in the first patch under different lithology types through the following formula:

[0094]

[0095] where P(r k |X) represents the probability that the pixel X belonging to the corresponding category in the first patch belongs to the lithology type r k , P(X|r k ) represents the conditional probability of pixel X given the lithology type r k , P(r k ) represents the prior probability of the lithology type r k , and P(X) represents the marginal probability of pixel X.

[0096] In one embodiment, in step S3, by introducing the normal distribution hypothesis of the spectral features of the image pixels, P(X|r k ) can be expressed as:

[0097]

[0098] where n represents the dimension of the spectral features, μ k represents the mean vector of the lithology type r k , and Σ represents the covariance matrix of the pixels X belonging to the corresponding class.

[0099] Please refer to Figure 3 , a remote sensing image geological lithology interpretation system disclosed in the present application, the system includes a lithology clustering analysis module, a patch similarity measurement module, and a lithology information propagation and optimization module, wherein:

[0100] The lithology clustering analysis module is used to obtain the lithology information of the patches in the specified area through the geological map, and initially perform lithology clustering analysis by combining the Landsat-8 remote sensing image with the k-means algorithm.

[0101] The patch similarity measurement module is used to, based on the clustering result, when it is determined that a plurality of categories are covered in the corresponding first patch, introduce a plurality of reference patches with other known lithology information, and perform similarity measurement between the first patch and each reference patch to obtain the most similar target reference patch.

[0102] The lithology information propagation and optimization module is used to, based on the maximum likelihood matching method, when it is determined that the lithology type matching between the target reference patch and the first patch is successful, propagate the lithology information of the target reference patch to other patches with similar geological backgrounds, and through multiple iterations, gradually optimize the lithology interpretation result of the entire area.

[0103] In one embodiment, the above-mentioned modules are also used to implement the remote sensing image geological lithology interpretation method described in any one of the foregoing method embodiments, and the present application does not limit this.

[0104] As can be seen from the above, a geological lithology interpretation system for remote sensing images disclosed in this application, on the one hand, through the combination of k-means clustering and patch similarity propagation algorithm, can achieve accurate matching between the lithology categories in remote sensing images and the actual lithology types. Compared with traditional methods, this application can more accurately reflect the lithology distribution of each geological patch, avoiding incorrect interpretation caused by unclear classification results. On the other hand, it uses remote sensing images and geological maps for automatic interpretation, without relying on manual experience or cumbersome manual correction processes, greatly improving the efficiency and automation level of lithology interpretation. This has important application value for large-scale geological area interpretation, especially in fields such as resource exploration and geological disaster prediction. Finally, the k-means clustering results are optimized through the patch similarity propagation algorithm and the maximum likelihood method, which can effectively eliminate the influence of noise on the interpretation results and avoid problems of misclassification and inaccurate interpretation. This makes the lithology interpretation results more reliable and more adaptable. It should be noted that this application can not only interpret the lithology of a single patch, but also comprehensively consider the similarity between patches under different regions and different types of geological backgrounds to further optimize the interpretation results. This provides strong support for the automatic interpretation of complex geological areas.

[0105] Please refer to Figure 4 , a storage medium disclosed in this application stores a computer program, and when the computer program is executed by a processor, it implements the above-mentioned method for interpreting the geological lithology of remote sensing images.

[0106] As can be seen from the above, a storage medium disclosed in this application, on the one hand, through the combination of k-means clustering and patch similarity propagation algorithm, can achieve accurate matching between the lithology categories in remote sensing images and the actual lithology types. Compared with traditional methods, this application can more accurately reflect the lithology distribution of each geological patch, avoiding incorrect interpretation caused by unclear classification results. On the other hand, it uses remote sensing images and geological maps for automatic interpretation, without relying on manual experience or cumbersome manual correction processes, greatly improving the efficiency and automation level of lithology interpretation. This has important application value for large-scale geological area interpretation, especially in fields such as resource exploration and geological disaster prediction. Finally, the k-means clustering results are optimized through the patch similarity propagation algorithm and the maximum likelihood method, which can effectively eliminate the influence of noise on the interpretation results and avoid problems of misclassification and inaccurate interpretation. This makes the lithology interpretation results more reliable and more adaptable. It should be noted that this application can not only interpret the lithology of a single patch, but also comprehensively consider the similarity between patches under different regions and different types of geological backgrounds to further optimize the interpretation results. This provides strong support for the automatic interpretation of complex geological areas.

[0107] It should be noted that: the above order of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0108] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

Claims

1. A method for interpreting geological lithology of remote sensing images, characterized in that: The method comprises: S1. Obtain lithology information of the designated area through geological maps, and use Landsat-8 remote sensing images combined with k-means algorithm to conduct preliminary lithology clustering analysis; S2. Based on the clustering result, when it is determined that the corresponding first image spot contains multiple categories, multiple reference image spots with other known lithology information are introduced, and similarity measurement is performed between the first image spot and each reference image spot to obtain the most similar target reference image spot; S3. Based on the maximum likelihood matching method, when it is determined that the lithology type between the target reference spot and the first spot is successfully matched, the lithology information of the target reference spot is propagated to other spots with similar geological backgrounds, and the lithology interpretation results of the entire area are gradually optimized through multiple iterations.

2. The method according to claim 1, characterized in that In step S1, the lithology information of the designated area is obtained through the geological map, and the Landsat-8 remote sensing image is combined with the k-means algorithm to perform preliminary lithology cluster analysis, including: S11, obtaining a geological map of a designated geological interpretation area, wherein the geological map includes a plurality of spots indicating different geological units; S12. Based on the geological characteristics, determine the lithology types and corresponding lithology quantities covered in each map block; S13. According to the location information of each image spot on the geological map, the corresponding regional remote sensing image is cropped from the Landsat-8 remote sensing image, and based on the corresponding lithology quantity, a preliminary lithology clustering analysis based on the k-means algorithm is performed on the regional remote sensing image.

3. The method according to claim 2, characterized in that In step S13, the lithology clustering analysis based on the corresponding lithology quantity is preliminarily performed on the regional remote sensing image based on the k-means algorithm, including: S131, taking the number of lithologies in the patch as the initial number of categories K, and randomly selecting K pixels from the remote sensing image of the region as initial cluster centers; S132, for each remaining pixel point in the image, calculate the distance between the pixel point and each cluster center, and assign the pixel point to the category where the nearest cluster center is located, to form a corresponding cluster; S133, for each cluster, calculating the spectral feature mean of all pixels in the cluster, and updating the initial cluster center based on the spectral feature mean; S134, repeating steps S132 to S133 until the position of the cluster center no longer changes or when a preset maximum number of iterations is reached, a corresponding clustering result is obtained.

4. The method according to claim 1, characterized in that: In step S2, the method further includes: based on the clustering result, when it is determined that the corresponding second image patch only covers one category, the category is matched with a known lithology type and a corresponding lithology label is assigned.

5. The method according to claim 1, characterized in that In step S2, the introduction of multiple reference spots with other known lithology information and the measurement of similarity between the first spot and each reference spot to obtain the most similar target reference spot include: S21, performing dimensionality reduction filtering processing on the first image spot and each reference image spot respectively to obtain a feature vector after dimensionality reduction; S22. Based on the feature vectors after dimensionality reduction, the similarity between the first image spot and each reference image spot is calculated by using the cosine similarity measurement method, and the most similar target reference image spot is obtained accordingly.

6. The method according to claim 1, characterized in that In step S3, the probability distribution of each category in the first image patch under different lithology types is calculated by the following formula: Among them, P(r k |X) indicates that the pixel X in the first image spot belongs to the corresponding category of lithology type r k The probability of P(X|r k ) represents a given lithology type r k The conditional probability of pixel X under the condition of k ) represents the lithology type r k The prior probability of , P(X) represents the marginal probability of pixel X.

7. The method according to claim 1, characterized in that In step S3, by introducing the normal distribution assumption of the spectral characteristics of image pixels, P(X|r k ) can be expressed as: Where n represents the dimension of the spectral feature, μ k Indicates the lithology type r k , Σ represents the covariance matrix of pixels X belonging to the corresponding category.

8. A system for interpreting geological lithology of remote sensing images, characterized in that: The system includes a lithology clustering analysis module, a pattern similarity measurement module, and a lithology information dissemination and optimization module, wherein: The lithology cluster analysis module is used to obtain lithology information of the designated area through the geological map, and to perform preliminary lithology cluster analysis using Landsat-8 remote sensing images combined with the k-means algorithm; The spot similarity measurement module is used to introduce multiple reference spots with other known lithology information when determining that the corresponding first spot covers multiple categories based on the clustering result, and perform similarity measurement between the first spot and each reference spot to obtain the most similar target reference spot; The lithology information propagation and optimization module is used to propagate the lithology information of the target reference spot to other spots with similar geological backgrounds based on the maximum likelihood matching method when it is determined that the lithology type between the target reference spot and the first spot is successfully matched, thereby gradually optimizing the lithology interpretation results of the entire area through multiple iterations.

9. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for interpreting geological lithology of remote sensing images is implemented.

Images