A hyperspectral remote sensing sea surface oil spill detection method, device, medium and product

Abstract

The application discloses a hyperspectral remote sensing sea surface oil stain detection method, equipment, medium and product, and relates to the technical field of remote sensing, and in particular relates to a hyperspectral remote sensing sea surface oil stain detection method, equipment, medium and product. The method comprises the following steps: performing noise band screening processing on an obtained original hyperspectral remote sensing sea surface image to obtain an optimized first hyperspectral remote sensing sea surface image; performing data dimension reduction on the first hyperspectral remote sensing sea surface image to obtain a second hyperspectral remote sensing sea surface image after dimension reduction; performing weak supervision classification on the second hyperspectral remote sensing sea surface image based on sparse pseudo labels to obtain an initial pixel-level oil stain probability graph; and performing spatial optimization processing on the initial pixel-level oil stain probability graph to obtain a sea surface oil stain detection result of the original hyperspectral remote sensing sea surface image. The technical scheme of the embodiment of the application can significantly improve the accuracy and practicability of sea surface oil stain detection.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a hyperspectral remote sensing method, equipment, medium, and product for detecting oil spills on the sea surface. Background Technology

[0002] With the increasing urgency of global marine environmental protection and emergency response to marine oil spills, high-precision, automated, and reliable marine oil pollution monitoring technologies have become a key component of the marine environmental safety system. Traditional methods such as on-site sampling and shipborne observation have limited spatial and temporal coverage and slow response times, making it difficult to meet the requirements of large-scale, rapid monitoring.

[0003] In existing technologies, the detection of marine oil spills from remote sensing images mainly follows three technical paths: First, methods based on physical models and spectral feature matching, such as constructing detection indices based on the differences in oil and water spectra. While the principle of this method is clear, it is sensitive to imaging conditions and oil spill types, resulting in poor universality. Second, methods based on traditional machine learning and manual feature engineering. Their performance highly depends on prior knowledge in feature design, and they are susceptible to the "curse of dimensionality" when processing high-dimensional spectral data, leading to insufficient stability. Third, methods based on deep learning, while capable of automatically extracting features, rely on a large number of precisely labeled samples. However, marine oil spill samples are scarce and difficult to label, leading to models prone to overfitting and weak generalization ability. Furthermore, their "black box" decision-making process is not conducive to the physical traceability and credibility verification of the results. Summary of the Invention

[0004] This invention provides a hyperspectral remote sensing method, equipment, medium, and product for detecting oil spills on the sea surface, in order to solve the problems of existing methods having strong dependence on labeled data, weak generalization ability, and insufficient interpretability of results.

[0005] According to one aspect of the present invention, a hyperspectral remote sensing method for detecting oil spills on the sea surface is provided, the method comprising:

[0006] The acquired raw hyperspectral remote sensing sea surface image was processed by noise band filtering to obtain the optimized first hyperspectral remote sensing sea surface image.

[0007] The first hyperspectral remote sensing sea surface image is subjected to dimensionality reduction to obtain the second hyperspectral remote sensing sea surface image after dimensionality reduction.

[0008] The second hyperspectral remote sensing sea surface image is weakly supervised to classify based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map.

[0009] Spatial optimization processing is performed on the initial pixel-level oil pollution probability map to obtain the oil pollution detection results of the original hyperspectral remote sensing sea surface image.

[0010] According to another aspect of the present invention, a hyperspectral remote sensing device for detecting oil spills on the sea surface is provided, the device comprising:

[0011] The image optimization module is used to perform noise band filtering on the acquired raw hyperspectral remote sensing sea surface image to obtain the optimized first hyperspectral remote sensing sea surface image.

[0012] The image dimensionality reduction module is used to perform dimensionality reduction on the first hyperspectral remote sensing sea surface image to obtain a dimensionality-reduced second hyperspectral remote sensing sea surface image.

[0013] The probability map generation module is used to perform weakly supervised classification on the second hyperspectral remote sensing sea surface image based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map.

[0014] The output detection result module is used to perform spatial optimization processing on the initial pixel-level oil pollution probability map to obtain the oil pollution detection result of the original hyperspectral remote sensing sea surface image.

[0015] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0016] At least one processor; and

[0017] A memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by the at least one processor, which is then executed by the at least one processor to enable the at least one processor to perform a hyperspectral remote sensing method for detecting oil spills on the sea surface as described in any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement a hyperspectral remote sensing method for detecting oil spills on the sea surface as described in any embodiment of the present invention.

[0020] According to another aspect of the present invention, a computer program product is also provided, including a computer program / instructions that, when executed by a processor, implement the steps of the method as described in any embodiment of the present invention.

[0021] The technical solution of this invention involves filtering noise bands from the acquired original hyperspectral remote sensing sea surface image to obtain an optimized first hyperspectral remote sensing sea surface image; performing dimensionality reduction on the first hyperspectral remote sensing sea surface image to obtain a dimensionality-reduced second hyperspectral remote sensing sea surface image; performing weakly supervised classification on the second hyperspectral remote sensing sea surface image based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map; and performing spatial optimization processing on the initial pixel-level oil pollution probability map to obtain the sea surface oil pollution detection result of the original hyperspectral remote sensing sea surface image. This solution solves the problems of strong dependence on labeled data, weak generalization ability, and insufficient interpretability of results in existing methods, achieving a significant improvement in the accuracy and practicality of sea surface oil pollution detection.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of a hyperspectral remote sensing method for detecting oil spills on the sea surface provided in Embodiment 1 of the present invention;

[0025] Figure 2 This is a flowchart of another hyperspectral remote sensing method for detecting oil spills on the sea surface provided in Embodiment 2 of the present invention;

[0026] Figure 3 This is a schematic diagram of a hyperspectral remote sensing method for detecting oil spills on the sea surface in a specific scenario applicable to an embodiment of the present invention.

[0027] Figure 4 This is a schematic diagram comparing detection results before and after spatial optimization in a specific scenario applicable to an embodiment of the present invention;

[0028] Figure 5 This is a schematic diagram illustrating a specific implementation process in a specific scenario to which this invention applies;

[0029] Figure 6 This is a schematic diagram of the structure of a hyperspectral remote sensing oil spill detection device for the sea surface provided in Embodiment 3 of the present invention;

[0030] Figure 7This is a schematic diagram of the structure of an electronic device for implementing a hyperspectral remote sensing method for detecting oil spills on the sea surface according to an embodiment of the present invention. Detailed Implementation

[0031] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] Example 1

[0034] Figure 1 This is a flowchart of a hyperspectral remote sensing method for detecting oil spills on the sea surface provided in Embodiment 1 of the present invention. This embodiment is applicable to the detection of oil spills on the sea surface from hyperspectral remote sensing images under the condition of limited labeled samples. The method can be executed by a hyperspectral remote sensing oil spill detection device, which can be implemented in hardware and / or software and is generally configured in an electronic device.

[0035] Correspondingly, such as Figure 1 As shown, the method includes:

[0036] S110. The acquired original hyperspectral remote sensing sea surface image is subjected to noise band filtering processing to obtain the optimized first hyperspectral remote sensing sea surface image.

[0037] Among them, hyperspectral remote sensing sea surface image refers to a data cube obtained by imaging the ocean surface through hyperspectral sensors mounted on satellites or airborne platforms, where each pixel contains tens to hundreds of continuous narrow-band spectral information.

[0038] In this embodiment, the acquired raw hyperspectral remote sensing sea surface image is subjected to noise band filtering processing. This includes identifying and removing low-quality bands that are severely affected by atmospheric scattering, sensor errors, or sea surface environmental interference, thereby retaining bands with effective spectral information and high signal-to-noise ratio, forming an optimized first hyperspectral remote sensing sea surface image, providing a cleaner and more reliable data foundation for subsequent analysis.

[0039] Optionally, based on the above embodiments, noise band filtering processing is performed on the acquired original hyperspectral remote sensing sea surface image to obtain an optimized first hyperspectral remote sensing sea surface image, which may include:

[0040] The noise response of the original hyperspectral remote sensing sea surface image was calculated band by band using the Laplacian operator convolution kernel;

[0041] Based on the noise response of each band, the noise intensity of the original hyperspectral remote sensing sea surface image in each band is statistically analyzed.

[0042] Calculate the mean of noise intensity across all bands, and determine an adaptive screening threshold based on the mean.

[0043] The bands with noise intensity higher than the adaptive screening threshold are designated as strong noise bands, and the signals of the strong noise bands in the original hyperspectral remote sensing sea surface image are filtered out to obtain the optimized first hyperspectral remote sensing sea surface image.

[0044] Generally, each band of the original hyperspectral remote sensing sea surface image is convolved using a Laplacian operator kernel to calculate the noise response of the image band by band. This helps to capture high-frequency noise components in the image caused by sensor errors or environmental interference.

[0045] Generally, based on the noise response results of each band, the noise intensity corresponding to each band is evaluated through statistical methods, thereby quantifying the specific degree of noise pollution in each band.

[0046] Generally, the average noise intensity of all bands is calculated, and an adaptive filtering threshold is dynamically determined based on this average value, so as to automatically set the filtering standard according to the noise distribution characteristics of the image itself.

[0047] Generally, bands with noise intensity higher than the adaptive screening threshold are identified as strong noise bands, and the data of these strong noise bands are filtered out from the original hyperspectral remote sensing sea surface image, finally obtaining the optimized first hyperspectral remote sensing sea surface image with significant noise interference removed.

[0048] Specifically, for each single-band image of the hyperspectral imagery Where n is the nth wave segment, and a 3×3 Laplace convolution kernel is used for convolution operation. The noise standard deviation of the nth wave segment is estimated using the following formula: ,in, For the first Standard deviation of band noise This is the noise statistical correction factor. , These represent the height and width of a single-band image, respectively. coordinates Calculate the mean standard deviation of noise across all frequency bands at the noise response value. Set the filter threshold to ,reserve The low-noise band is used to eliminate noise bands.

[0049] S120. Perform dimensionality reduction on the first hyperspectral remote sensing sea surface image to obtain the second hyperspectral remote sensing sea surface image after dimensionality reduction.

[0050] In this embodiment, the first hyperspectral remote sensing sea surface image is subjected to data dimensionality reduction processing. By reducing the number of spectral dimensions, information redundancy and computational burden are reduced, while key spectral feature structures are preserved, thereby obtaining a second hyperspectral remote sensing sea surface image after dimensionality reduction. This helps to highlight the spectral differences between oil spills and the background and improves the efficiency of subsequent processing.

[0051] Optionally, based on the above embodiments, performing dimensionality reduction on the first hyperspectral remote sensing sea surface image to obtain a dimensionality-reduced second hyperspectral remote sensing sea surface image may include:

[0052] The spectral data of the first hyperspectral remote sensing sea surface image were mapped to a nonlinear feature space using the kernel principal component analysis method.

[0053] Extract a predetermined number of principal component features from the nonlinear feature space;

[0054] Based on the preset number of principal component features, the first hyperspectral remote sensing sea surface image is subjected to dimensionality reduction processing to obtain the dimensionality-reduced second hyperspectral remote sensing sea surface image.

[0055] Generally, kernel principal component analysis (KPI) is used to transform the spectral data of the first hyperspectral remote sensing sea surface image from the original space to a higher-dimensional nonlinear feature space. This transformation utilizes kernel function techniques to effectively capture and characterize the complex nonlinear structure and intrinsic correlations in the original spectral data, thus making it easier to distinguish the spectral features of oil spills from seawater in subsequent analyses.

[0056] Generally, in the nonlinear feature space obtained through kernel trick mapping, a predetermined number of principal component features are calculated and extracted. These principal components are the directions in which the variance of the data contributes the most in this nonlinear space, and can retain the key information and differences of the original high-dimensional data to the greatest extent with fewer data dimensions.

[0057] Generally, based on the extracted preset number of principal component features, a projection transformation is performed on all pixels of the first hyperspectral remote sensing sea surface image, and the original high-dimensional spectral information of each pixel is represented by a few principal component feature coefficients. Through this process, data dimensionality reduction is achieved, resulting in a second hyperspectral remote sensing sea surface image with significantly reduced data volume while retaining the core spectral separability information.

[0058] Specifically, for the screened low-noise hyperspectral data, a kernel matrix is ​​constructed using the RBF (Radial Basis Function) kernel function. The kernel matrix is ​​then centered, and the eigenvalues ​​and eigenvectors of the kernel matrix are solved. The first 25 principal components are selected to form a feature mapping matrix, which maps the original high-dimensional spectral data to a 25-dimensional nonlinear feature space, thus completing the dimensionality reduction process. This reduces data redundancy and computational complexity while preserving the spectral discrimination features.

[0059] S130. The second hyperspectral remote sensing sea surface image is subjected to weakly supervised classification based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map.

[0060] Sparse pseudo-labels can be understood as using a very small amount of imprecise labeling information as a supervisory signal, and then automatically inferring or generating labels for model training through algorithms. Weakly supervised classification can be understood as a machine learning paradigm that uses the aforementioned incomplete and imprecise supervisory information to train a classification model when the training data lacks a large amount of fine-grained labeling.

[0061] In this embodiment, the second hyperspectral remote sensing sea surface image is classified under weak supervision based on sparse pseudo-labels. Sparse pseudo-labels refer to the use of a small amount of prior knowledge or automatically generated coarse labeling information to train the classification model through a weak supervision learning algorithm. This generates an initial pixel-level oil pollution probability map that reflects the possibility of each pixel being an oil pollution without relying on a large number of finely labeled samples, effectively alleviating the constraint of insufficient labeled data.

[0062] Specifically, using sparse pseudo-labels as training samples, an RBF kernel SVM binary classification model is constructed, and the model optimization objective function is: The constraints are: ,in, For the classification hyperplane normal vector, For bias, As slack variables, The penalty coefficient is... For kernel mapping eigenvectors, For sample pseudo-labels, The summation index variable is m, where m is the total number of training samples. The value is the square of the modulus. After training, global pixels are classified and predicted, and the posterior probability of each pixel belonging to the oily area is output, thus obtaining the initial oily probability map.

[0063] S140. Spatial optimization processing is performed on the initial pixel-level oil pollution probability map to obtain the oil pollution detection results of the original hyperspectral remote sensing sea surface image.

[0064] In this embodiment, the initial pixel-level oil pollution probability map is spatially optimized. By introducing spatial context information or applying post-processing techniques, isolated points and irregular areas in the probability map are smoothed or corrected, enhancing the spatial continuity and consistency of the detection results. Ultimately, accurate and reliable oil pollution detection results are obtained, improving the visualization quality and practicality of the final output map.

[0065] The technical solution of this invention involves filtering noise bands from the acquired original hyperspectral remote sensing sea surface image to obtain an optimized first hyperspectral remote sensing sea surface image; performing dimensionality reduction on the first hyperspectral remote sensing sea surface image to obtain a dimensionality-reduced second hyperspectral remote sensing sea surface image; performing weakly supervised classification on the second hyperspectral remote sensing sea surface image based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map; and performing spatial optimization processing on the initial pixel-level oil pollution probability map to obtain the sea surface oil pollution detection result of the original hyperspectral remote sensing sea surface image. This solution solves the problems of strong dependence on labeled data, weak generalization ability, and insufficient interpretability of results in existing methods, achieving a significant improvement in the accuracy and practicality of sea surface oil pollution detection.

[0066] Example 2

[0067] Figure 2 This is a flowchart of a hyperspectral remote sensing method for detecting oil spills on the sea surface provided in Embodiment 2 of the present invention. This embodiment is based on and optimized from the above embodiments. Specifically, the step of "performing spatial optimization processing on the initial pixel-level oil spill probability map to obtain the oil spill detection result of the original hyperspectral remote sensing sea surface image" has been refined.

[0068] Correspondingly, such as Figure 2 As shown, the method includes:

[0069] S210. The acquired original hyperspectral remote sensing sea surface image is subjected to noise band filtering processing to obtain the optimized first hyperspectral remote sensing sea surface image.

[0070] S220. Perform dimensionality reduction on the first hyperspectral remote sensing sea surface image to obtain the dimensionality-reduced second hyperspectral remote sensing sea surface image.

[0071] S230. The second hyperspectral remote sensing sea surface image is subjected to weakly supervised classification based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map.

[0072] Optionally, based on the above embodiments, the second hyperspectral remote sensing sea surface image is subjected to weakly supervised classification based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map, which may include:

[0073] Based on the spectral angle mapping method, the spectral angle between the spectral vector of each pixel in the second hyperspectral remote sensing sea surface image and the preset oil pollution reference spectrum is calculated, and all pixels are initially divided into a potential oil pollution candidate set and a background seawater candidate set according to the global spectral angle threshold.

[0074] K-means clustering algorithm is executed on two candidate sets respectively. The cluster center point of each cluster and several of its nearest neighbor pixels are selected to form the oil stain key point candidate set and the background key point candidate set respectively. The total number of selected pixels must meet the preset sparsity ratio.

[0075] All pixels in the candidate set of oil stain key points are assigned pseudo-labels belonging to the oil stain area, while all pixels in the candidate set of background key points are assigned pseudo-labels not belonging to the oil stain area, thus obtaining the sparse pseudo-labels.

[0076] Obtain sparse pseudo-labels corresponding to at least one key pixel in the second hyperspectral remote sensing sea surface image; wherein, the sparse pseudo-labels include: belonging to the oil spill area and not belonging to the oil spill area;

[0077] Based on the pixel spectral data corresponding to each key pixel and sparse pseudo-labels, a lightweight training sample is constructed, and the support vector machine (SVM) binary classifier is trained using the lightweight training sample to obtain an oil pollution classification model for the second hyperspectral remote sensing sea surface image.

[0078] Based on the oil pollution classification model, all pixels in the second hyperspectral remote sensing sea surface image are classified and predicted to obtain the initial pixel-level oil pollution probability map, which is used to describe the confidence level of each pixel in the second hyperspectral remote sensing sea surface image belonging to the oil pollution area.

[0079] Generally, the process begins by calculating the angle between the spectral vector of each pixel in the reduced-dimensional second hyperspectral remote sensing sea surface image and a preset oil pollution reference spectrum, based on the spectral angle mapping method. Then, according to a globally set spectral angle threshold, all pixels are initially divided into a potential oil pollution candidate set and a background seawater candidate set, thereby achieving coarse-grained binary classification of pixels.

[0080] Generally, the K-means clustering algorithm is executed independently on the initially divided potential oil pollution candidate set and background seawater candidate set to group the pixels in each set. Then, the cluster center point and several pixels closest to the spectral features of the center point are selected from each cluster and merged to form the oil pollution key point candidate set and the background key point candidate set, respectively. The total number of selected pixels must meet a preset sparsity requirement with a low proportion.

[0081] Generally, all pixels in the candidate set of oil stain key points are uniformly assigned pseudo-labels as belonging to the oil stain area, while all pixels in the candidate set of background key points are uniformly assigned pseudo-labels as not belonging to the oil stain area. Through this non-global labeling method based on representative key points, the generated sparse pseudo-labels are finally obtained.

[0082] Generally, from the second hyperspectral remote sensing sea surface image, sparse pseudo-labels corresponding to at least one key pixel selected in the previous step are obtained. These labels contain two categories of information: "belongs to the oil spill area" and "does not belong to the oil spill area".

[0083] Generally, a lightweight training sample set with a small amount of data is constructed based on the pixel spectral data corresponding to each key pixel and the sparse pseudo-labels they carry. This sample set is then used to train a support vector machine binary classifier, thereby obtaining a customized oil pollution classification model for the current second hyperspectral remote sensing sea surface image.

[0084] Generally, based on the trained oil pollution classification model, all pixels in the second hyperspectral remote sensing sea surface image are classified and predicted. The probability value of each pixel belonging to the oil pollution category is calculated, and finally an initial pixel-level oil pollution probability map covering the entire image is obtained. This map is used to quantitatively describe the confidence level of each pixel in the image belonging to the oil pollution area.

[0085] S240. Based on the second hyperspectral remote sensing sea surface image, construct an undirected topological graph model with pixels as nodes and the spectral similarity between pixels as edge weights.

[0086] Among them, the undirected topological graph model can be understood as a graph structure that uses nodes to represent pixels and edge weights to represent the spectral similarity between pixels, and is used to describe the spatial and spectral relationships of pixels in hyperspectral remote sensing images.

[0087] In this embodiment, based on the dimension-reduced second hyperspectral remote sensing sea surface image, an undirected topological graph model is constructed with pixels as nodes and the spectral similarity between pixels as edge weights. This model can quantitatively represent the proximity relationship between any two pixels in the image in the feature space, thereby associating the global spatial structure of the image with spectral similarity.

[0088] Optionally, based on the above embodiments, an undirected topological graph model is constructed using the second hyperspectral remote sensing sea surface image, with pixels as nodes and spectral similarity between pixels as edge weights, including:

[0089] Using each pixel of the second hyperspectral remote sensing sea surface image as a graph node, the edges of the graph model are defined by the spatial adjacency relationship between pixels, and an undirected topological graph model is initialized.

[0090] Calculate the spectral vector difference between each pair of adjacent graph nodes connected by an edge, where,

[0091] The spectral vector is a multidimensional array composed of the spectral data corresponding to each pixel in the second hyperspectral remote sensing sea surface image;

[0092] Based on the spectral vector difference, the weight of the edge between the corresponding graph nodes is calculated using an exponential decay function, where the smaller the spectral vector difference, the larger the edge weight.

[0093] Based on all graph nodes and the calculated edge weights, construct the adjacency matrix or weight matrix of the undirected topological graph model to complete the construction of the undirected topological graph model.

[0094] Generally, each pixel of the second hyperspectral remote sensing sea surface image after dimensionality reduction is taken as the basic unit and regarded as a graph node. Then, according to the spatial position of the pixel in the image, the graph nodes corresponding to adjacent pixels are connected by edges, thus initially defining the basic skeleton structure of an undirected topological graph model.

[0095] Generally, for each pair of adjacent graph nodes directly connected by an edge in an undirected topological graph, the difference in spectral vectors between their corresponding two pixels is calculated. Here, the spectral vector specifically refers to the feature array composed of the spectral data of each pixel in the second hyperspectral remote sensing sea surface image after dimensionality reduction. This difference value is used to measure the similarity of two adjacent pixels in spectral features.

[0096] Generally, based on the calculated difference in spectral vectors between each pair of adjacent graph nodes, an exponential decay function is used to calculate and determine the weight of the connection edge between them. This function is designed so that the smaller the difference in spectral vectors between two nodes, the larger the calculated edge weight value, thus visually representing a stronger connection between pixels with similar spectral features on the graph.

[0097] Generally, after determining all graph nodes and the weights of the edges connecting them, the entire undirected topological graph model is formally described and stored using an adjacency matrix or weight matrix. The rows and columns of the matrix correspond to all graph nodes, and the element values ​​in the matrix represent the weights of the edges between corresponding node pairs. This completes the construction of the undirected topological graph model for subsequent spatial continuity constraints.

[0098] S250. Based on the undirected topological graph model, a spatial continuity constraint is applied to the initial pixel-level oil pollution probability map to obtain the oil pollution detection results of the original hyperspectral remote sensing sea surface image.

[0099] Optionally, based on the above embodiments, a spatial continuity constraint is applied to the initial pixel-level oil pollution probability map based on the undirected topological graph model to obtain the sea surface oil pollution detection result of the original hyperspectral remote sensing sea surface image, which may include:

[0100] Construct the graph Laplacian matrix corresponding to the undirected topological graph model;

[0101] An optimization objective function is constructed with spatial smoothness as the optimization objective and the initial pixel-level oil contamination probability map as the fidelity constraint.

[0102] Solving the objective function yields an optimized oil pollution probability map, which serves as the oil pollution detection result for the original hyperspectral remote sensing sea surface image.

[0103] Generally, based on a completed undirected topological graph model, the graph Laplacian matrix corresponding to the undirected topological graph model is constructed by calculating the degree of each graph node (i.e., the sum of the weights of all edges connected to that node) and the edge weight relationships between nodes. This matrix is ​​a mathematical representation of the graph structure, which can reflect the connection strength between nodes in the graph and the global topological properties.

[0104] Generally, an optimization objective function is constructed with spatial smoothness as the core optimization goal. This function encourages neighboring nodes with larger edge weights (i.e., pixels with similar spectra) in the undirected topological graph model to obtain similar oil contamination probability values. At the same time, the initial pixel-level oil contamination probability map is introduced into the objective function as a fidelity constraint to ensure that the optimization result does not deviate excessively from the initial classification prediction, thereby achieving a balance between pursuing spatial continuity and smoothness and respecting the initial classification confidence.

[0105] Generally, by solving the above-defined objective function using mathematical methods, we obtain an oil pollution probability solution that minimizes the value of the objective function and has a more continuous and consistent spatial distribution. This solution is the optimized oil pollution probability map, which is ultimately used as the oil pollution detection result of the original hyperspectral remote sensing sea surface image.

[0106] Specifically, the pixels of the second hyperspectral remote sensing sea surface image are used as graph nodes to construct an undirected topological graph model. Where V represents the node combination and E represents the edge set, the edge weights are calculated based on the spectral differences between adjacent pixels. Where exp(.) is the exponential function, These are the weighting coefficients. and Let represent the spectral vectors of the i-th and j-th pixels, respectively. Construct the graph Laplacian matrix L and establish the spatial optimization objective function:

[0107]

[0108] Where P is the optimized oil contamination probability map, and O is the initial oil contamination probability map. To ensure fidelity, Let be the confidence diagonal matrix, and min be the minimization operation; then, solve the linear equation: The optimized oil pollution probability map is obtained, isolated noise points and fragmented false targets are removed, and spatially continuous oil pollution detection results are obtained.

[0109] The technical solution of this invention involves filtering noise bands from the acquired original hyperspectral remote sensing sea surface image to obtain an optimized first hyperspectral remote sensing sea surface image; performing dimensionality reduction on the first hyperspectral remote sensing sea surface image to obtain a dimensionality-reduced second hyperspectral remote sensing sea surface image; performing weakly supervised classification on the second hyperspectral remote sensing sea surface image based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map; constructing an undirected topological graph model based on the second hyperspectral remote sensing sea surface image, with pixels as nodes and spectral similarity between pixels as edge weights; and applying spatial continuity constraints to the initial pixel-level oil pollution probability map based on the undirected topological graph model to obtain the sea surface oil pollution detection results of the original hyperspectral remote sensing sea surface image. This solution solves the problems of strong dependence on labeled data, weak generalization ability, and insufficient interpretability of results in existing methods, achieving a significant improvement in the accuracy and practicality of sea surface oil pollution detection.

[0110] In this specific embodiment, in order to address the challenges of limited detection accuracy, weak generalization ability, and low automation in existing hyperspectral remote sensing methods for detecting oil spills on the sea surface, which suffer from high-dimensional data redundancy, imaging noise interference, and strong dependence on a large number of accurately labeled samples, this invention presents a complete hyperspectral remote sensing method for detecting oil spills on the sea surface based on weak supervision and graph optimization.

[0111] Specifically, Figure 3 This is a schematic diagram of the hyperspectral remote sensing method for detecting oil spills on the sea surface provided in an embodiment of the present invention, as shown below. Figure 3 As shown, this process covers the entire chain from raw image input to final detection result output. First, raw hyperspectral remote sensing images of the sea surface are acquired and preprocessed to eliminate noise and distortion introduced by the sensor itself. Then, a specific image denoising algorithm is used to clean the image, suppressing background interference such as sea surface ripples and uneven illumination, resulting in a relatively clean denoised image. Based on this, classifiers such as support vector machines are used to perform pixel-level classification of the denoised image, initially extracting potential oil pollution areas on the sea surface and generating a discrete initial oil pollution probability map. Finally, to improve the completeness and continuity of the detection results, a spatial optimization algorithm is introduced to process the initial probability map, smoothing spatial noise and filling internal holes, outputting a final optimized oil pollution detection result with a clear spatial contour.

[0112] Specifically, Figure 4 This is a schematic diagram comparing the detection results before and after spatial optimization provided in an embodiment of the present invention, such as... Figure 4 As shown, this comparison intuitively demonstrates the significant improvement effect of the spatial optimization mechanism on the initial classification results. The left side shows the initial oil spill probability map without spatial optimization. It can be seen that due to the high dimensionality and susceptibility to noise interference of hyperspectral data, the initial results exhibit high fragmentation and discreteness, with the effective signal being masked by a large number of artifacts and isolated noise points. The right side shows the oil spill probability map after spatial optimization. By introducing the graph Laplacian matrix and spatial smoothness constraints, the optimized result successfully eliminates isolated small noise points, significantly enhances the connectivity of the oil spill area, and makes the oil spill patches present a complete, continuous, and closed spatial morphology, which is more consistent with the diffusion and aggregation characteristics of real sea surface oil spills in physical space.

[0113] Specifically, Figure 5 This is a schematic diagram illustrating a specific implementation process provided in the embodiments of the present invention, such as... Figure 5As shown, this example intuitively presents the complete evolution process from a real sea surface scene to the final accurate extraction of oil slick areas. The first row shows the processing effect of the real remote sensing image. The left side is the original image containing complex sea surface textures, and the right side is the smoother and clearer image after denoising, intuitively demonstrating the effect of the preprocessing stage in suppressing background interference. The second row shows the evolution of the classification probability map during the algorithm processing. The left side is the initial image after support vector machine processing, where dense yellow bright spots represent the suspected oil slick pixels initially identified by the algorithm, but there is obvious scattered noise. The right side is the final image after ERW (Graph Optimization-Based Spatial Regularization) processing. The red area is continuous and smooth, successfully aggregating discrete suspected points into a complete oil slick with clear boundaries, verifying the effectiveness and robustness of the embodiment of the present invention in real complex sea surface scenes.

[0114] Example 3

[0115] Figure 6 This invention provides a hyperspectral remote sensing device for detecting oil spills on the sea surface, as described in Embodiment 3. Figure 6 As shown, the device includes:

[0116] The image optimization module 610 is used to perform noise band filtering processing on the acquired original hyperspectral remote sensing sea surface image to obtain the optimized first hyperspectral remote sensing sea surface image.

[0117] The image dimensionality reduction module 620 is used to perform dimensionality reduction on the first hyperspectral remote sensing sea surface image to obtain a dimensionality-reduced second hyperspectral remote sensing sea surface image.

[0118] The probability map generation module 630 is used to perform weakly supervised classification on the second hyperspectral remote sensing sea surface image based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map.

[0119] The output detection result module 640 is used to perform spatial optimization processing on the initial pixel-level oil pollution probability map to obtain the oil pollution detection result of the original hyperspectral remote sensing sea surface image.

[0120] The technical solution of this invention involves filtering noise bands from the acquired original hyperspectral remote sensing sea surface image to obtain an optimized first hyperspectral remote sensing sea surface image; performing dimensionality reduction on the first hyperspectral remote sensing sea surface image to obtain a dimensionality-reduced second hyperspectral remote sensing sea surface image; performing weakly supervised classification on the second hyperspectral remote sensing sea surface image based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map; and performing spatial optimization processing on the initial pixel-level oil pollution probability map to obtain the sea surface oil pollution detection result of the original hyperspectral remote sensing sea surface image. This solution solves the problems of strong dependence on labeled data, weak generalization ability, and insufficient interpretability of results in existing methods, achieving a significant improvement in the accuracy and practicality of sea surface oil pollution detection.

[0121] Based on the above embodiments, the image module 610 is optimized and is specifically used for:

[0122] The noise response of the original hyperspectral remote sensing sea surface image was calculated band by band using the Laplacian operator convolution kernel;

[0123] Based on the noise response of each band, the noise intensity of the original hyperspectral remote sensing sea surface image in each band is statistically analyzed.

[0124] Calculate the mean of noise intensity across all bands, and determine an adaptive screening threshold based on the mean.

[0125] The bands with noise intensity higher than the adaptive screening threshold are designated as strong noise bands, and the signals of the strong noise bands in the original hyperspectral remote sensing sea surface image are filtered out to obtain the optimized first hyperspectral remote sensing sea surface image.

[0126] Based on the above embodiments, the image dimensionality reduction module 620 is specifically used for:

[0127] The spectral data of the first hyperspectral remote sensing sea surface image were mapped to a nonlinear feature space using the kernel principal component analysis method.

[0128] Extract a predetermined number of principal component features from the nonlinear feature space;

[0129] Based on the preset number of principal component features, the first hyperspectral remote sensing sea surface image is subjected to dimensionality reduction processing to obtain the dimensionality-reduced second hyperspectral remote sensing sea surface image.

[0130] Based on the above embodiments, the probability map generation module 630 is specifically used for:

[0131] Based on the spectral angle mapping method, the spectral angle between the spectral vector of each pixel in the second hyperspectral remote sensing sea surface image and the preset oil pollution reference spectrum is calculated, and all pixels are initially divided into a potential oil pollution candidate set and a background seawater candidate set according to the global spectral angle threshold.

[0132] K-means clustering algorithm is executed on two candidate sets respectively. The cluster center point of each cluster and several of its nearest neighbor pixels are selected to form the oil stain key point candidate set and the background key point candidate set respectively. The total number of selected pixels must meet the preset sparsity ratio.

[0133] All pixels in the candidate set of oil stain key points are assigned pseudo-labels belonging to the oil stain area, while all pixels in the candidate set of background key points are assigned pseudo-labels not belonging to the oil stain area, thus obtaining the sparse pseudo-labels.

[0134] Obtain sparse pseudo-labels corresponding to at least one key pixel in the second hyperspectral remote sensing sea surface image; wherein, the sparse pseudo-labels include: belonging to the oil spill area and not belonging to the oil spill area;

[0135] Based on the pixel spectral data corresponding to each key pixel and sparse pseudo-labels, a lightweight training sample is constructed, and the support vector machine (SVM) binary classifier is trained using the lightweight training sample to obtain an oil pollution classification model for the second hyperspectral remote sensing sea surface image.

[0136] Based on the oil pollution classification model, all pixels in the second hyperspectral remote sensing sea surface image are classified and predicted to obtain the initial pixel-level oil pollution probability map, which is used to describe the confidence level of each pixel in the second hyperspectral remote sensing sea surface image belonging to the oil pollution area.

[0137] Furthermore, based on the above embodiments, the output detection result module 640 may further include:

[0138] The graph model construction module is used to construct an undirected topological graph model based on the second hyperspectral remote sensing sea surface image, with pixels as nodes and the spectral similarity between pixels as edge weights.

[0139] The spatial constraint module is used to apply spatial continuity constraints to the initial pixel-level oil pollution probability map based on the undirected topological graph model, so as to obtain the oil pollution detection results of the original hyperspectral remote sensing sea surface image.

[0140] Based on the above embodiments, the graph model construction module is specifically used for:

[0141] Using each pixel of the second hyperspectral remote sensing sea surface image as a graph node, the edges of the graph model are defined by the spatial adjacency relationship between pixels, and an undirected topological graph model is initialized.

[0142] Calculate the spectral vector difference between each pair of adjacent graph nodes connected by an edge, wherein the spectral vector is a multidimensional array composed of the spectral data corresponding to each pixel in the second hyperspectral remote sensing sea surface image;

[0143] Based on the spectral vector difference, the weight of the edge between the corresponding graph nodes is calculated using an exponential decay function, where the smaller the spectral vector difference, the larger the edge weight.

[0144] Based on all graph nodes and the calculated edge weights, construct the adjacency matrix or weight matrix of the undirected topological graph model to complete the construction of the undirected topological graph model.

[0145] Based on the above embodiments, the space constraint module is specifically used for;

[0146] Construct the graph Laplacian matrix corresponding to the undirected topological graph model;

[0147] An optimization objective function is constructed with spatial smoothness as the optimization objective and the initial pixel-level oil contamination probability map as the fidelity constraint.

[0148] Solving the objective function yields an optimized oil pollution probability map, which serves as the oil pollution detection result for the original hyperspectral remote sensing sea surface image.

[0149] The hyperspectral remote sensing marine oil pollution detection device provided in this embodiment of the invention can execute the hyperspectral remote sensing marine oil pollution detection method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0150] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0151] Example 4

[0152] Figure 7 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0153] like Figure 7 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0154] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0155] Processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as performing a hyperspectral remote sensing method for detecting oil spills on the sea surface as described in any embodiment of the present invention, namely:

[0156] The acquired raw hyperspectral remote sensing sea surface image was processed by noise band filtering to obtain the optimized first hyperspectral remote sensing sea surface image.

[0157] The first hyperspectral remote sensing sea surface image is subjected to dimensionality reduction to obtain the second hyperspectral remote sensing sea surface image after dimensionality reduction.

[0158] The second hyperspectral remote sensing sea surface image is weakly supervised to classify based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map.

[0159] Spatial optimization processing is performed on the initial pixel-level oil pollution probability map to obtain the oil pollution detection results of the original hyperspectral remote sensing sea surface image.

[0160] In some embodiments, a hyperspectral remote sensing method for detecting oil spills on the sea surface as described in any one of the embodiments of the present invention can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the hyperspectral remote sensing method for detecting oil spills on the sea surface as described above as described in any one of the embodiments of the present invention can be performed. Alternatively, in other embodiments, processor 11 can be configured by any other suitable means (e.g., by means of firmware) to perform the hyperspectral remote sensing method for detecting oil spills on the sea surface as described in any one of the embodiments of the present invention.

[0161] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0162] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0163] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0164] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0165] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0166] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0167] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0168] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A hyperspectral remote sensing method for detecting oil spills on the sea surface, characterized in that, The method includes: The acquired original hyperspectral remote sensing sea surface image was subjected to noise band screening processing, and the bands with noise standard deviation less than half of the mean noise standard deviation of all bands were retained to obtain the optimized first hyperspectral remote sensing sea surface image. The first hyperspectral remote sensing sea surface image is subjected to dimensionality reduction to obtain the second hyperspectral remote sensing sea surface image after dimensionality reduction. The second hyperspectral remote sensing sea surface image is weakly supervised to classify based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map. Spatial optimization processing is performed on the initial pixel-level oil pollution probability map to obtain the oil pollution detection results of the original hyperspectral remote sensing sea surface image; Specifically, the second hyperspectral remote sensing sea surface image is subjected to weakly supervised classification based on sparse pseudo-labels to obtain an initial pixel-level oil pollution probability map, including: Based on the spectral angle mapping method, the spectral angle between the spectral vector of each pixel in the second hyperspectral remote sensing sea surface image and the preset oil pollution reference spectrum is calculated, and all pixels are initially divided into a potential oil pollution candidate set and a background seawater candidate set according to the global spectral angle threshold. K-means clustering algorithm is executed on two candidate sets respectively. The cluster center point of each cluster and several of its nearest neighbor pixels are selected to form the oil stain key point candidate set and the background key point candidate set respectively. The total number of selected pixels must meet the preset sparsity ratio. All pixels in the candidate set of oil stain key points are assigned pseudo-labels belonging to the oil stain area, while all pixels in the candidate set of background key points are assigned pseudo-labels not belonging to the oil stain area, thus obtaining the sparse pseudo-labels. Obtain sparse pseudo-labels corresponding to at least one key pixel in the second hyperspectral remote sensing sea surface image; wherein, the sparse pseudo-labels include: belonging to the oil spill area and not belonging to the oil spill area; Based on the pixel spectral data corresponding to each key pixel and sparse pseudo-labels, a lightweight training sample is constructed, and the support vector machine (SVM) binary classifier is trained using the lightweight training sample to obtain an oil pollution classification model for the second hyperspectral remote sensing sea surface image. Based on the oil pollution classification model, all pixels in the second hyperspectral remote sensing sea surface image are classified and predicted to obtain the initial pixel-level oil pollution probability map, which is used to describe the confidence level of each pixel in the second hyperspectral remote sensing sea surface image belonging to the oil pollution area.

2. The method according to claim 1, characterized in that, The acquired raw hyperspectral remote sensing sea surface image was processed by noise band filtering to obtain the optimized first hyperspectral remote sensing sea surface image, including: The noise response of the original hyperspectral remote sensing sea surface image was calculated band by band using the Laplacian operator convolution kernel; Based on the noise response of each band, the noise intensity of the original hyperspectral remote sensing sea surface image in each band is statistically analyzed. Calculate the mean of noise intensity across all bands, and determine an adaptive screening threshold based on the mean. The bands with noise intensity higher than the adaptive screening threshold are designated as strong noise bands, and the signals of the strong noise bands in the original hyperspectral remote sensing sea surface image are filtered out to obtain the optimized first hyperspectral remote sensing sea surface image.

3. The method according to claim 1, characterized in that, The first hyperspectral remote sensing sea surface image is subjected to dimensionality reduction to obtain a second hyperspectral remote sensing sea surface image, including: The spectral data of the first hyperspectral remote sensing sea surface image were mapped to a nonlinear feature space using the kernel principal component analysis method. Extract a predetermined number of principal component features from the nonlinear feature space; Based on the preset number of principal component features, the first hyperspectral remote sensing sea surface image is subjected to dimensionality reduction processing to obtain the dimensionality-reduced second hyperspectral remote sensing sea surface image.

4. The method according to claim 1, characterized in that, Spatially optimize the initial pixel-level oil pollution probability map to obtain the sea surface oil pollution detection results of the original hyperspectral remote sensing sea surface image, including: Based on the second hyperspectral remote sensing sea surface image, an undirected topological graph model is constructed with pixels as nodes and the spectral similarity between pixels as edge weights. Based on the undirected topological graph model, a spatial continuity constraint is applied to the initial pixel-level oil pollution probability map to obtain the oil pollution detection results of the original hyperspectral remote sensing sea surface image.

5. The method according to claim 4, characterized in that, Based on the second hyperspectral remote sensing sea surface image, an undirected topological graph model is constructed, with pixels as nodes and the spectral similarity between pixels as edge weights, including: Using each pixel of the second hyperspectral remote sensing sea surface image as a graph node, the edges of the graph model are defined by the spatial adjacency relationship between pixels, and an undirected topological graph model is initialized. Calculate the spectral vector difference between each pair of adjacent graph nodes connected by an edge, wherein the spectral vector is a multidimensional array composed of the spectral data corresponding to each pixel in the second hyperspectral remote sensing sea surface image; Based on the spectral vector difference, the weight of the edge between the corresponding graph nodes is calculated using an exponential decay function, where the smaller the spectral vector difference, the larger the edge weight. Based on all graph nodes and the calculated edge weights, construct the adjacency matrix or weight matrix of the undirected topological graph model to complete the construction of the undirected topological graph model.

6. The method according to claim 4, characterized in that, Based on the undirected topological graph model, a spatial continuity constraint is applied to the initial pixel-level oil pollution probability map to obtain the oil pollution detection results of the original hyperspectral remote sensing sea surface image, including: Construct the graph Laplacian matrix corresponding to the undirected topological graph model; An optimization objective function is constructed with spatial smoothness as the optimization objective and the initial pixel-level oil contamination probability map as the fidelity constraint. Solving the objective function yields an optimized oil pollution probability map, which serves as the oil pollution detection result for the original hyperspectral remote sensing sea surface image.

7. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the hyperspectral remote sensing method for detecting oil spills on the sea surface according to any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the hyperspectral remote sensing method for detecting oil spills on the sea surface as described in any one of claims 1-6.

9. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the hyperspectral remote sensing method for detecting oil spills on the sea surface according to any one of claims 1-6.

Images