Satellite remote sensing sea ice detection method and device

Abstract

The application relates to the technical field of remote sensing image processing, and provides a satellite remote sensing sea ice detection method and device. The method comprises the following steps: processing multispectral remote sensing images of a target sea area to obtain image data; extracting multi-dimensional features including edge features, first spectral features and second spectral features from the image data; wherein the edge features are obtained by performing adaptive gradient enhancement and edge probability smoothing processing on the image data, the first spectral features include a sea ice index after correction of the influence of suspended silt, and the second spectral features include a water body index; according to classification thresholds corresponding to each feature in the multi-dimensional features, performing sea ice distribution judgment on each pixel in the image data to generate a sea ice distribution map. The satellite remote sensing sea ice detection method and device provided by the application can realize high-precision identification of sea ice based on the multi-dimensional features including the edge features, the first spectral features and the second spectral features extracted from the image data.

Description

Technical Field

[0001] This application relates to the field of remote sensing image processing technology, and in particular to a satellite remote sensing method and apparatus for detecting sea ice. Background Technology

[0002] Accurate detection of sea ice is of great significance for marine environmental monitoring, shipping safety, and climate research. Currently, sea ice detection methods based on satellite optical remote sensing mainly rely on the differences in spectral characteristics between ice and water. By selecting one or more spectral bands to calculate specific indices and setting global thresholds for pixel classification, ice-water separation can be achieved.

[0003] However, traditional methods that rely on fixed spectral features and global thresholds can achieve good results in open and clean waters, but their detection accuracy and reliability decrease significantly in complex water environments such as nearshore areas and estuaries. The main problems are as follows: First, severe spectral confusion: suspended sediment increases water reflectivity, which overlaps with the spectral features of sea ice (especially thin ice), resulting in a high false detection rate for the fixed threshold method; Second, discontinuous edge extraction: traditional gradient operators are sensitive to noise, and the extracted sea ice contours are broken and rough under complex sea conditions; Third, poor environmental adaptability: a single spectral index is difficult to cope with changes in different lighting, weather, and water color conditions.

[0004] Therefore, there is an urgent need for a sea ice remote sensing method that can effectively overcome the interference of suspended sediment, accurately extract the ice edge, and has strong environmental adaptability, so as to meet the actual needs of high-precision and high-reliability sea ice monitoring in complex water environments. Summary of the Invention

[0005] This application provides a satellite remote sensing sea ice detection method and device to solve the technical problems in the prior art of remote sensing sea ice detection being severely interfered with by suspended sediment, having a high false detection rate in complex water bodies, and having incomplete ice edge extraction, thereby realizing high-precision and high-reliability sea ice monitoring in complex water environments.

[0006] In a first aspect, embodiments of this application provide a satellite remote sensing method for sea ice detection, including:

[0007] Image data is obtained by processing multispectral remote sensing images of the target sea area;

[0008] Multi-dimensional features, including edge features, first spectral features, and second spectral features, are extracted from the image data. The edge features are obtained by adaptive gradient enhancement and edge probability smoothing of the image data. The first spectral features include the sea ice index after correction for the influence of suspended sediment. The second spectral features include the water body index.

[0009] Based on the classification thresholds corresponding to each feature in the multi-dimensional features, the sea ice distribution of each pixel in the image data is determined, and a sea ice distribution map is generated.

[0010] In one embodiment, the sea ice distribution of each pixel in the image data is determined based on the classification threshold corresponding to each feature in the multi-dimensional features, and a sea ice distribution map is generated. Specifically, this can be implemented as follows:

[0011] An adaptive thresholding algorithm is used to determine the classification threshold corresponding to each feature in the multi-dimensional features. Based on the comparison results between the feature values ​​of each pixel in the image data and the corresponding classification threshold, the initial sea ice detection results are generated.

[0012] Based on the initial sea ice detection results, a sea ice distribution map was obtained.

[0013] In another embodiment, an adaptive thresholding algorithm is used to determine the classification threshold corresponding to each feature in the multi-dimensional features. Based on the comparison results between the feature values ​​of each pixel in the image data and the corresponding classification thresholds, an initial sea ice detection result is generated. Specifically, this can be implemented as follows:

[0014] Determine the edge threshold of the edge feature, the first threshold of the first spectral feature, and the second threshold of the second spectral feature;

[0015] For each pixel in the image data, if the value of the edge feature is greater than the edge threshold, the value of the first spectral feature is greater than the first threshold, and the value of the second spectral feature is less than the second threshold, the pixel is determined to be sea ice.

[0016] In yet another embodiment, the image data includes green band data, and the edge features are obtained as follows:

[0017] Gradient calculations are performed on the green band data based on the adaptive gradient operator to obtain the gradient values ​​in the X and Y directions.

[0018] The gradient magnitude map is calculated based on the gradient values ​​in the X and Y directions.

[0019] A weighting function is constructed based on spatial distance and gradient similarity. The gradient magnitude map is then subjected to weighted smoothing operation based on the weighting function to obtain the edge probability map. The edge probability map is used to reflect the probability that each pixel belongs to the edge of the sea ice. The edge features are the edge probability map.

[0020] In yet another embodiment, the image data further includes shortwave infrared band data, red band data, and near-infrared band data; the above method further includes:

[0021] Dynamic correction factors for suspended sediment concentration were obtained based on red band data and near-infrared band data.

[0022] The sea ice index was obtained based on green band data, shortwave infrared band data, and a dynamic correction factor for suspended sediment concentration.

[0023] In yet another embodiment, the sea ice index satisfies a first preset formula, which is:

[0024]

[0025] in, Indicates the sea ice index; This represents green band data; 'a' represents the dynamic correction factor for suspended sediment concentration. 'b' represents shortwave infrared band data; 'b' represents the first parameter.

[0026] In yet another embodiment, the dynamic correction factor for suspended sediment concentration satisfies a second preset formula, which is:

[0027]

[0028] Where 'a' represents the dynamic correction factor for suspended sediment concentration; Indicates the correction factor; Indicates empirical parameters; Indicates red band data; This represents near-infrared band data.

[0029] In yet another embodiment, the method further includes obtaining a water index based on near-infrared band data and green band data.

[0030] In yet another embodiment, the water index satisfies a third preset formula, which is:

[0031]

[0032] in, Indicates water quality index; Indicates green band data; This represents near-infrared band data.

[0033] Secondly, embodiments of this application provide a satellite remote sensing sea ice detection device, comprising the following modules:

[0034] The data preprocessing module is used to process multispectral remote sensing images of the target sea area to obtain image data;

[0035] The multi-dimensional feature extraction module is used to extract multi-dimensional features from image data, including edge features, first spectral features, and second spectral features. The edge features are obtained by adaptive gradient enhancement and edge probability smoothing of the image data. The first spectral features include the sea ice index after correction for the influence of suspended sediment. The second spectral features include the water body index.

[0036] The classification processing module is used to determine the sea ice distribution of each pixel in the image data based on the classification threshold corresponding to each feature in the multi-dimensional features, and generate a sea ice distribution map.

[0037] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the satellite remote sensing sea ice detection method of the first aspect.

[0038] Fourthly, embodiments of this application provide a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the satellite remote sensing sea ice detection method of the first aspect.

[0039] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the satellite remote sensing sea ice detection method of the first aspect.

[0040] This application provides a satellite remote sensing method and apparatus for sea ice detection. By utilizing edge features obtained through adaptive gradient enhancement and edge probability smoothing, it effectively strengthens the weak edge signal of broken sea ice and suppresses complex background noise, thereby improving the integrity and continuity of sea ice contour extraction. By introducing a sea ice index corrected for the influence of suspended sediment, it eliminates spectral confusion and false detection of ice pixels caused by the high reflectivity of turbid water. Through the fusion of multi-dimensional features, it improves the overall accuracy and reliability of sea ice identification in complex water environments. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a system framework diagram of the satellite remote sensing sea ice detection system provided in the embodiments of this application.

[0043] Figure 2 This is a flowchart illustrating the satellite remote sensing sea ice detection method provided in this application embodiment.

[0044] Figure 3 This is a flowchart illustrating the edge feature extraction method provided in the embodiments of this application.

[0045] Figure 4This is a flowchart illustrating the sea ice index extraction method provided in this application embodiment.

[0046] Figure 5A This is a sea ice distribution map provided in the embodiments of this application under the scenario of thin ice and turbid water.

[0047] Figure 5B This is a sea ice distribution map provided in an embodiment of the present application for a scenario with dense sea ice and containing ice fragments.

[0048] Figure 5C This is a sea ice distribution map provided in an embodiment of this application under a scenario where the sea ice edge is broken and the lighting conditions are complex.

[0049] Figure 6 This is a schematic diagram of the structure of the satellite remote sensing sea ice detection device provided in the embodiments of this application.

[0050] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0052] The following is combined with Figures 1 to 7 This application provides a detailed description of the satellite remote sensing sea ice detection method and apparatus provided in its embodiments.

[0053] The solution provided in this application can be applied to Figure 1 The satellite remote sensing sea ice detection system shown, Figure 1 This is a system framework diagram of the satellite remote sensing sea ice detection system provided in the embodiments of this application.

[0054] For example, the satellite remote sensing sea ice detection system 100 includes a data preprocessing module 110, a multi-dimensional feature extraction module 120, an adaptive classification module 130, a post-processing optimization module 140, and an output module 150.

[0055] The data preprocessing module 110 is used to preprocess the acquired multispectral remote sensing images to generate standardized image data.

[0056] Optionally, preprocessing includes, but is not limited to, radiometric calibration, geometric correction, and land masking.

[0057] The multi-dimensional feature extraction module 120 is used to extract multi-dimensional features from image data in parallel.

[0058] Optionally, the multidimensional features include, but are not limited to, edge features obtained through adaptive gradient enhancement and edge probability smoothing, sea ice index fused with suspended sediment correction factors, and water body index used to enhance water body contrast.

[0059] The adaptive classification module 130 is used to determine the classification threshold for multi-dimensional features using an adaptive threshold algorithm, and to perform pixel-level ice and water classification based on the classification threshold to generate the initial sea ice distribution results.

[0060] The post-processing optimization module 140 is used to perform spatial filtering and isolated small region removal on the initial sea ice distribution results to obtain the sea ice distribution map.

[0061] Output module 150 is used to convert sea ice distribution maps into standard geographic information formats and output them.

[0062] It should be noted that the modules are coupled together via a system bus or a shared buffer queue. Each module acts as an independent processing unit, writing its generated output data (such as preprocessed images) to the bus or buffer queue; downstream modules read the required data from their inputs and then trigger their own processing flow. This loosely coupled connection method based on an intermediary achieves asynchronous decoupling and data buffering between modules.

[0063] Under this architecture, the satellite remote sensing sea ice detection system forms a highly efficient pipeline-style processing link. The upstream module does not need to wait for the downstream module to finish processing before it can start processing the next frame of data or the next task unit, thereby significantly improving the overall throughput and resource utilization of the satellite remote sensing sea ice detection system, which is especially suitable for batch or near real-time processing scenarios of massive remote sensing images.

[0064] Figure 2 This is a flowchart illustrating the satellite remote sensing sea ice detection method provided in the embodiments of this application, as shown below. Figure 2 As shown, the method includes the following:

[0065] Step 201: Process the multispectral remote sensing image of the target sea area to obtain image data.

[0066] The target sea area refers to coastal, estuarine or nearshore sea areas that contain sea ice distribution and have complex optical characteristics (such as high concentration of suspended sediment, strong ocean currents or wind and waves).

[0067] Multispectral remote sensing imagery refers to digital images that contain at least two different spectral bands, acquired by optical sensors mounted on a remote sensing platform.

[0068] Optionally, the remote sensing platform may include, but is not limited to, a satellite platform, an aerial platform, or an unmanned aerial vehicle platform.

[0069] In some embodiments, a satellite platform may be used to acquire multispectral remote sensing images.

[0070] In some embodiments, the multispectral remote sensing imagery includes at least data in the green band, shortwave infrared band, near-infrared band, and red band.

[0071] Specifically, the multispectral remote sensing images of the target sea area are processed, including:

[0072] (1) Radiometric calibration: Convert the raw digital number (DN) values ​​of multispectral remote sensing images into physically meaningful apparent radiance or reflectance data.

[0073] (2) Geometric correction: Geometric correction is performed on multispectral remote sensing images using satellite orbit parameters or ground control points to ensure accurate spatial positioning.

[0074] (3) Land masking: Based on the coastline vector data, the land part in the image is masked, and the pure sea area to be analyzed is preserved.

[0075] After the above processing, image data of each band of the target sea area are obtained.

[0076] For example, image data for each band can be represented as green band data F. G (x,y), red band data F R (x,y), near-infrared band data F NIR (x,y), shortwave infrared band data F SWIR (x,y).

[0077] Step 202: Extract multi-dimensional features from the image data, including edge features, first spectral features, and second spectral features.

[0078] Among them, multidimensional features refer to a set of multiple features extracted from image data that are independent and complementary in the spatial and spectral domains.

[0079] Optionally, the multidimensional features include, but are not limited to, edge features, first spectral features, and second spectral features.

[0080] In this embodiment, the edge features are obtained by adaptive gradient enhancement and edge probability smoothing of the image data.

[0081] Specifically, adaptive gradient enhancement refers to performing convolution operations on green band data using an improved adaptive gradient operator to calculate the gradient components of each pixel in the horizontal and vertical directions, thereby obtaining the gradient magnitude.

[0082] Marginal probability smoothing refers to the weighted smoothing of gradient magnitudes through bilateral filtering to obtain an edge probability map.

[0083] It should be noted that edge probability smoothing considers not only the spatial distance between its neighboring pixels but also the similarity of their gradient magnitudes when calculating the output value of each pixel. This allows it to effectively preserve and enhance the true edge structure while smoothing noise within homogeneous regions. The resulting edge probability map serves as the edge feature, with its value reflecting the probability that the pixel belongs to the edge of sea ice.

[0084] The edge features extracted in this embodiment can enhance the weak signal of the sea ice contour and suppress background noise, thus identifying the potential boundary of the sea ice in terms of spatial morphology.

[0085] In this embodiment of the application, the first spectral feature may include a sea ice index that has been corrected for the effects of suspended sediment.

[0086] Specifically, based on the red band and near-infrared band data from the imagery, the suspended sediment concentration at each pixel location can be obtained. Furthermore, a dynamic correction factor for suspended sediment concentration can be derived based on this concentration.

[0087] It should be noted that the dynamic correction factor for suspended sediment concentration increases with the increase of suspended sediment concentration.

[0088] Furthermore, the sea ice index was obtained based on green band data, shortwave infrared band data, and a dynamic correction factor for suspended sediment concentration.

[0089] The sea ice index in this embodiment adaptively modulates the calculation contribution of shortwave infrared data by introducing a dynamic correction factor associated with suspended sediment concentration. This mechanism can specifically correct for changes in water spectral reflectance caused by suspended sediment, effectively suppress spectral confusion, and thus significantly improve the separability of sea ice and turbid water pixels in this index feature.

[0090] In this embodiment of the application, the second spectral feature may include a water index.

[0091] Specifically, the water index can be obtained based on near-infrared band data and green band data.

[0092] The water index in this embodiment can enhance water information and help distinguish water bodies from highly reflective ground features, including sea ice.

[0093] Step 203: Based on the classification thresholds corresponding to each feature in the multi-dimensional features, determine the sea ice distribution of each pixel in the image data and generate a sea ice distribution map.

[0094] Specifically, the classification thresholds corresponding to the edge features, the first spectral features, and the second spectral features are determined by an adaptive threshold algorithm.

[0095] For example, an adaptive thresholding algorithm could be Otsu's Method (Otsu). Its core principle is to adaptively classify feature data into two classes, target (sea ice-related feature responses) and background (water and noise-related feature responses), by maximizing the ratio of inter-class variance to intra-class variance, thereby determining the classification threshold.

[0096] The adaptive threshold algorithm can automatically optimize based on the gray-scale distribution characteristics of each feature parameter, avoiding the problem of insufficient adaptability of the traditional fixed threshold method to complex water environments, and ensuring the matching degree between the classification threshold and the feature data.

[0097] In one embodiment, the classification threshold for the edge feature is the edge threshold Tp, the classification threshold for the first spectral feature (sea ice index) is the first threshold Ti, and the classification threshold for the second spectral feature (water index) is the second threshold Tw.

[0098] Furthermore, for each pixel (x, y) in the image data, its extracted feature value is compared with the corresponding classification threshold, and a judgment is made in combination with the preset decision rules.

[0099] For example, the decision rule could be: for any pixel (x, y), a pixel is considered a sea ice pixel if and only if it simultaneously satisfies the following three conditions:

[0100] Edge eigenvalues ​​G pdf (x,y) is greater than the edge threshold Tp;

[0101] The first spectral eigenvalue NDIIce(x,y) is greater than the first threshold Ti;

[0102] The second spectral feature value NDWI(x,y) is less than the second threshold Tw.

[0103] If any of the above conditions are not met, the pixel is considered a non-sea ice pixel (i.e., a water body).

[0104] Then, by traversing all pixels in the image data, an initial sea ice detection result is generated.

[0105] In other words, the initial sea ice detection result can be determined by the following decision function:

[0106]

[0107] In this context, IceMask(x,y)=1 indicates that the pixel represents sea ice, and IceMask(x,y)=0 indicates that the pixel represents water. This represents the edge feature value at pixel (x,y). Indicates the edge threshold. This represents the sea ice index at pixel (x,y). Indicates the first threshold. This represents the water index at pixel (x,y). This indicates the second threshold.

[0108] Furthermore, the initial sea ice detection results were post-processed and optimized to obtain a sea ice distribution map.

[0109] For example, median filtering can be used to remove salt-and-pepper noise, or morphological opening operations can be used to remove isolated regions with an area smaller than a preset threshold.

[0110] In the embodiments of this application, the sea ice distribution map can be represented in various forms, without any specific limitation. For example, it can be represented in binary raster form, where pixel "1" represents sea ice and pixel "0" represents water. It can also be represented in a standard geographic information format.

[0111] This application provides a satellite remote sensing method and apparatus for sea ice detection. By utilizing edge features obtained through adaptive gradient enhancement and edge probability smoothing, it effectively strengthens the weak edge signal of broken sea ice and suppresses complex background noise, thereby improving the integrity and continuity of sea ice contour extraction. By introducing a sea ice index corrected for the influence of suspended sediment, it eliminates spectral confusion and false detection of ice pixels caused by the high reflectivity of turbid water. Through the fusion of multi-dimensional features, it improves the overall accuracy and reliability of sea ice identification in complex water environments.

[0112] Figure 3 This is a flowchart illustrating the edge feature extraction method provided in the embodiments of this application, as shown below. Figure 3 As shown, the method includes the following:

[0113] Step S301: Perform gradient calculation on the green band data based on the adaptive gradient operator to obtain the gradient values ​​in the X and Y directions.

[0114] The adaptive gradient operator is composed of a gradient enhancement factor α and a smoothing factor σ, which are modulated by the local variance of the image, and is introduced into the classic Sobel operator template.

[0115]

[0116]

[0117] in, This represents the gradient value in the X direction; α represents the gradient value in the Y direction; α represents the gradient enhancement factor; σ represents the smoothing factor.

[0118] Specifically, gradient enhancement factor β is an empirical coefficient with a value of 0.5. Smoothing factor. µ is a scale parameter with a value of 0.1. The local variance of the data within the current operator region. This indicates green band data.

[0119] Step S302: Calculate the gradient magnitude map based on the gradient values ​​in the X and Y directions.

[0120] Specifically, the gradient magnitude at pixel (x,y) satisfies the following formula:

[0121]

[0122]

[0123]

[0124] in, Indicates the gradient component in the X direction; Indicates the gradient component in the Y direction; This represents the gradient magnitude at pixel (x,y); Indicates green band data; This represents the gradient value in the X direction; This represents the gradient value in the Y direction; This indicates a convolution operation.

[0125] Step S303: Construct a weight function based on spatial distance and gradient similarity, and perform a weighted smoothing operation on the gradient magnitude map based on the weight function to obtain the marginal probability map.

[0126] Among them, the edge probability map is used to reflect the probability that each pixel belongs to the edge of the sea ice, and the edge feature is the edge probability map.

[0127] Specifically, the weighting function satisfies the following formula:

[0128]

[0129] in, Let represent the filter weights, where i represents the offset of neighboring pixels in the X-row and column directions relative to the center pixel, and j represents the offset of neighboring pixels in the Y-row and column directions relative to the center pixel. This represents the spatial standard deviation (controlling the degree of smoothing; a value of 2.0 is acceptable). This represents the standard deviation of grayscale (controls edge preservation; the value can be 0.1). This represents the gradient magnitude at pixel (x,y). Represents a pixel ( , The gradient magnitude at ().

[0130] Specifically, each cell (x, y) in the marginal probability map satisfies the following formula:

[0131]

[0132] in, This represents the probability quantization value of pixel (x,y) being the edge of sea ice; Indicates the filter weights; Represents a pixel ( , The gradient magnitude at () represents the sliding window half-width, which controls the range of neighborhood calculation.

[0133] In the embodiments of this application, k can be set to 3 (7×7 window) or 4 (9×9 window) as needed. The larger the window k is, the stronger the smoothing effect.

[0134] The satellite remote sensing sea ice detection method provided in this application can extract the green band gradient through an adaptive gradient operator, taking into account both edge enhancement and noise suppression; combined with the gradient magnitude map and a weight function constructed based on spatial distance and gradient similarity, an edge probability map is generated, which not only strengthens the weak contour signal of sea ice, but also accurately preserves edge details, thus significantly improving the positioning accuracy of sea ice boundaries in complex backgrounds.

[0135] Figure 4 This is a flowchart illustrating the sea ice index extraction method provided in the embodiments of this application, as shown below. Figure 4 As shown, the method includes the following:

[0136] Step S401: Obtain the dynamic correction factor for suspended sediment concentration based on red band data and near-infrared band data.

[0137] Specifically, based on the red band and near-infrared band data from the imagery, the suspended sediment concentration at each pixel location can be obtained. Furthermore, a dynamic correction factor for suspended sediment concentration can be derived based on this concentration.

[0138] Specifically, the suspended sediment concentration can satisfy the following formula:

[0139]

[0140] in, Represents a pixel The concentration of suspended sediment at the location; This represents the empirical parameters obtained by fitting experimental data; Indicates red band data; This represents near-infrared band data.

[0141] In the embodiments of this application, This is used to map the combined reflectance values ​​of red band and near-infrared band data to the estimation range of suspended sediment concentration. It can be adaptively adjusted according to the characteristics of the optical sensor and the optical properties of the target water body. For example, It can be 0.18.

[0142] Specifically, the dynamic correction factor for suspended sediment concentration can satisfy the following formula:

[0143]

[0144] Where 'a' represents the dynamic correction factor for suspended sediment concentration; This indicates the preset correction coefficient; This represents the concentration of suspended sediment at pixel (x,y).

[0145] In the embodiments of this application, the correction coefficient Used to adjust the contribution of the estimated suspended sediment concentration to the dynamic correction factor of suspended sediment concentration.

[0146] For example, The possible value is 0.05. It can also be adaptively adjusted according to the optical model used, the inherent optical characteristics of the target sea area, or seasonal changes to optimize the correction effect in different complex water environments.

[0147] Step S402: Obtain the sea ice index based on green band data, shortwave infrared band data, and dynamic correction factor for suspended sediment concentration.

[0148] Specifically, the sea ice index can satisfy the following formula:

[0149]

[0150] in, Indicates the sea ice index; This represents green band data; 'a' represents the dynamic correction factor for suspended sediment concentration. 'b' represents shortwave infrared band data; 'b' represents the first parameter.

[0151] In this embodiment, the specific value of the first parameter b can be adaptively adjusted according to the radiometric unit and numerical calculation requirements of the image data. For example, It can be 0.02.

[0152] It should be noted that the green band and shortwave infrared band data at certain pixels in the image data may be very small, causing the denominator to approach zero, thus leading to overflow or significant errors in the sea ice index calculation. Adding the first parameter 'b' helps maintain the numerical range and reasonableness of the sea ice index output value, preventing extreme values.

[0153] The satellite remote sensing sea ice detection method provided in this application can obtain dynamic correction factors for suspended sediment concentration through red band data and near-infrared band data, accurately quantifying sediment interference; combined with green band and shortwave infrared band data and the factor to calculate the sea ice index, it effectively reduces the spectral confusion of ice and water caused by sediment, and significantly improves the accuracy and robustness of sea ice identification in complex turbid water environments.

[0154] In one embodiment, the above method further includes:

[0155] The water index is derived from near-infrared and green band data.

[0156] Specifically, the water quality index can satisfy the following formula:

[0157]

[0158] in, Indicates water quality index; Indicates green band data; This represents near-infrared band data.

[0159] The satellite remote sensing sea ice detection method provided in this application can calculate the water index based on near-infrared and green band data. It can quickly distinguish water and non-water pixels by utilizing the significant difference in reflectance between water and sea ice in these two bands; it also helps to suppress background interference and improve the accuracy of ice-water separation.

[0160] To verify the effectiveness and adaptability of the method in this application, the embodiments of this application use data from a medium-resolution imaging spectrometer mounted on a satellite as the data source. Based on marine images captured on January 11, February 2, and February 19, 2025, the performance of the method in different scenarios is verified.

[0161] The specific bands used are: green band (center wavelength approximately 565nm), near-infrared band (center wavelength approximately 868nm), short-wave infrared band (center wavelength approximately 1645nm), and red band (center wavelength approximately 650nm).

[0162] This application is combined with Figure 5A , Figure 5B , Figure 5C Three satellite imagery illustrations are attached to illustrate the sea ice extraction results:

[0163] Figure 5AThis paper presents the extraction results in a scenario where thin ice predominates and nearshore waters are turbid (with high suspended sediment concentration). The red areas superimposed in the image represent the automatic sea ice extraction results. The extraction results show that the sea ice index with dynamic suspended sediment correction in this application effectively distinguishes between thin ice and high sediment waters, avoiding the missed detection of thin ice caused by spectral confusion in traditional methods. At the same time, the gradient-probability dual-constraint edge features accurately capture the continuous contour of the thin ice without obvious breaks or roughness, and the extraction results are highly consistent with the actual sea ice distribution.

[0164] Figure 5B The results demonstrate extraction in scenarios with extensive sea ice coverage, uneven thickness, and ice fragmentation caused by ocean currents. The results show that the red sea ice region completely covers both dense ice and ice fragmentation areas. The multi-dimensional feature collaborative decision-making mechanism successfully suppressed noise and current interference around the ice fragments, with no isolated false positives. The adaptive threshold algorithm automatically adjusts the classification criteria based on spectral and edge differences of sea ice of varying thicknesses, achieving accurate integrated identification of thick and ice fragments with no significant false positives or false negatives.

[0165] Figure 5C The extraction results are presented in scenarios with broken sea ice edges and complex image lighting conditions (uneven image brightness due to cloudy weather). The results show that the red sea ice region accurately matches the irregular boundary of the sea ice, and the edge probability map effectively enhances the weak edge signal of the ablation zone, avoiding the edge breakage problem of traditional operators in scenarios with uneven brightness. At the same time, even under the conditions of unstable spectral features caused by changes in lighting and sea ice ablation, high recognition accuracy is maintained, providing reliable support for dynamic sea ice monitoring.

[0166] In summary, the validation results in the three scenarios above demonstrate that the proposed method effectively addresses the challenges of spectral confusion between thin ice and turbid water, noise interference from broken ice and complex sea conditions, and the identification challenges of broken edges and uneven lighting. It achieves high-precision and high-reliability sea ice detection under various complex conditions. The extracted results show continuous and complete sea ice edges, and the outlines of broken ice floes are clearly defined and rich in detail, effectively solving the drawbacks of traditional methods such as broken and rough outlines. This indicates that the proposed solution overcomes spectral interference and noise effects in complex water environments through multi-dimensional feature collaboration, achieving the goal of high-precision and high-reliability sea ice detection, and providing efficient and feasible technical support for dynamic sea ice monitoring in complex nearshore waters.

[0167] The foregoing mainly describes the solution provided in this application. Accordingly, this application also provides a satellite remote sensing sea ice detection device for implementing the above-described method embodiments.

[0168] The satellite remote sensing sea ice detection device provided in the embodiments of this application is described below. The satellite remote sensing sea ice detection device described below and the satellite remote sensing sea ice detection method described above can be referred to each other.

[0169] Figure 6 This is a schematic diagram of the structure of the satellite remote sensing sea ice detection device provided in an embodiment of this application. Figure 6 As shown, the satellite remote sensing sea ice detection device 600 includes the following:

[0170] The data preprocessing module 601 is used to process multispectral remote sensing images of the target sea area to obtain image data;

[0171] The multi-dimensional feature extraction module 602 extracts multi-dimensional features from the image data, including edge features, first spectral features, and second spectral features. The edge features are obtained by adaptive gradient enhancement and edge probability smoothing of the image data. The first spectral features include the sea ice index after correction for the influence of suspended sediment. The second spectral features include the water body index.

[0172] The classification processing module 603 determines the sea ice distribution of each pixel in the image data based on the classification threshold corresponding to each feature in the multi-dimensional features, and generates a sea ice distribution map.

[0173] In one embodiment, the classification processing module 603 is specifically used for:

[0174] An adaptive thresholding algorithm is used to determine the classification threshold corresponding to each feature in the multi-dimensional features. Based on the comparison results between the feature values ​​of each pixel in the image data and the corresponding classification threshold, the initial sea ice detection results are generated.

[0175] Based on the initial sea ice detection results, a sea ice distribution map was obtained.

[0176] In yet another embodiment, the classification processing module 603 described above is specifically used for:

[0177] Determine the edge threshold of the edge feature, the first threshold of the first spectral feature, and the second threshold of the second spectral feature;

[0178] For each pixel in the image data, if the value of the edge feature is greater than the edge threshold, the value of the first spectral feature is greater than the first threshold, and the value of the second spectral feature is less than the second threshold, the pixel is determined to be sea ice.

[0179] In yet another embodiment, the multi-dimensional feature extraction module 602 described above is specifically used for:

[0180] Gradient calculations are performed on the green band data based on the adaptive gradient operator to obtain the gradient values ​​in the X and Y directions.

[0181] The gradient magnitude map is calculated based on the gradient values ​​in the X and Y directions.

[0182] A weighting function is constructed based on spatial distance and gradient similarity. The gradient magnitude map is then subjected to weighted smoothing operation based on the weighting function to obtain the edge probability map. The edge probability map is used to reflect the probability that each pixel belongs to the edge of the sea ice. The edge features are the edge probability map.

[0183] In yet another embodiment, the aforementioned sea ice index satisfies a first preset formula, which is:

[0184]

[0185] in, Indicates the sea ice index; This represents green band data; 'a' represents the dynamic correction factor for suspended sediment concentration. 'b' represents shortwave infrared band data; 'b' represents the first parameter.

[0186] In yet another embodiment, the aforementioned dynamic correction factor for suspended sediment concentration satisfies a second preset formula, which is:

[0187]

[0188] Where 'a' represents the dynamic correction factor for suspended sediment concentration; Indicates the correction factor; Indicates empirical parameters; Indicates red band data; This represents near-infrared band data.

[0189] In yet another embodiment, the multi-dimensional feature extraction module 602 described above is specifically used for:

[0190] The water index is derived from near-infrared and green band data.

[0191] In yet another embodiment, the aforementioned water index satisfies a third preset formula, which is:

[0192]

[0193] in, Indicates water quality index; Indicates green band data; This represents near-infrared band data.

[0194] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7As shown, the electronic device may include a processor 710, a communications interface 720, a memory 730, and a communication bus 740. The processor 710, communications interface 720, and memory 730 communicate with each other via the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute a satellite remote sensing sea ice detection method. This method includes: processing multispectral remote sensing images of the target sea area to obtain image data; extracting multidimensional features from the image data, including edge features, a first spectral feature, and a second spectral feature; wherein the edge features are obtained by adaptive gradient enhancement and edge probability smoothing of the image data, the first spectral feature includes a sea ice index corrected for the influence of suspended sediment, and the second spectral feature includes a water body index; and determining the sea ice distribution of each pixel in the image data based on the classification thresholds corresponding to each feature in the multidimensional features to generate a sea ice distribution map.

[0195] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0196] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the satellite remote sensing sea ice detection method provided by the above methods. The method includes: processing multispectral remote sensing images of a target sea area to obtain image data; extracting multidimensional features from the image data, including edge features, a first spectral feature, and a second spectral feature; wherein the edge features are obtained by adaptive gradient enhancement and edge probability smoothing processing of the image data, the first spectral feature includes a sea ice index after correction for the influence of suspended sediment, and the second spectral feature includes a water body index; and judging the sea ice distribution of each pixel in the image data according to the classification thresholds corresponding to each feature in the multidimensional features to generate a sea ice distribution map.

[0197] In another aspect, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the satellite remote sensing sea ice detection method provided by the above methods. The method includes: processing multispectral remote sensing images of a target sea area to obtain image data; extracting multidimensional features from the image data, including edge features, a first spectral feature, and a second spectral feature; wherein the edge features are obtained by adaptive gradient enhancement and edge probability smoothing processing of the image data, the first spectral feature includes a sea ice index after correcting for the influence of suspended sediment, and the second spectral feature includes a water body index; and judging the sea ice distribution of each pixel in the image data according to the classification thresholds corresponding to each feature in the multidimensional features, thereby generating a sea ice distribution map.

[0198] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0199] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0200] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A satellite remote sensing method for sea ice detection, characterized in that, The method includes: Multispectral remote sensing images of the target sea area are processed to obtain image data, which includes green band data, shortwave infrared band data, red band data, and near-infrared band data. Extract multi-dimensional features, including edge features, first spectral features, and second spectral features, from the image data; Based on the classification thresholds corresponding to each feature in the multi-dimensional features, the sea ice distribution of each pixel in the image data is determined, and a sea ice distribution map is generated. The edge features are obtained by performing adaptive gradient enhancement and edge probability smoothing on the image data, and are acquired in the following manner: Gradient calculations are performed on the green band data based on the adaptive gradient operator to obtain the gradient values ​​in the X and Y directions. A gradient magnitude map is calculated based on the gradient values ​​in the X and Y directions. A weight function is constructed based on spatial distance and gradient similarity. The gradient magnitude map is then subjected to a weighted smoothing operation based on the weight function to obtain an edge probability map. The edge probability map is used to reflect the probability that each pixel belongs to the edge of sea ice. The edge feature is the edge probability map. The first spectral feature includes a sea ice index corrected for the influence of suspended sediment, which is obtained in the following manner: Based on the red band data and the near-infrared band data, a dynamic correction factor for suspended sediment concentration is obtained. The sea ice index is obtained based on the green band data, the shortwave infrared band data, and the dynamic correction factor for suspended sediment concentration. The second spectral feature includes the water index.

2. The satellite remote sensing sea ice detection method according to claim 1, characterized in that, The step of determining the sea ice distribution of each pixel in the image data based on the classification threshold corresponding to each feature in the multi-dimensional features and generating a sea ice distribution map includes: The classification threshold corresponding to each feature in the multi-dimensional features is determined by an adaptive threshold algorithm. Based on the comparison results between the feature values ​​of each pixel in the image data and the corresponding classification threshold, an initial sea ice detection result is generated. Based on the initial sea ice detection results, the sea ice distribution map is obtained.

3. The satellite remote sensing sea ice detection method according to claim 2, characterized in that, The step involves determining the classification threshold corresponding to each feature in the multi-dimensional features using an adaptive thresholding algorithm, and generating initial sea ice detection results based on the comparison results between the feature values ​​of each pixel in the image data and the corresponding classification thresholds, including: Determine the edge threshold of the edge feature, the first threshold of the first spectral feature, and the second threshold of the second spectral feature; For each pixel in the image data, if the value of the edge feature is greater than the edge threshold, the value of the first spectral feature is greater than the first threshold, and the value of the second spectral feature is less than the second threshold, then the pixel is determined to be sea ice.

4. The satellite remote sensing sea ice detection method according to claim 1, characterized in that, The sea ice index satisfies a first preset formula, which is: ； in, This indicates the sea ice index; The green band data is represented by 'a'; 'a' represents the dynamic correction factor for suspended sediment concentration. b represents the shortwave infrared band data; b represents the first parameter; The dynamic correction factor for suspended sediment concentration satisfies a second preset formula, which is: ； Where 'a' represents the dynamic correction factor for suspended sediment concentration; Indicates the correction factor; Indicates empirical parameters; This refers to the red band data; This refers to the near-infrared band data.

5. The satellite remote sensing sea ice detection method according to claim 1, characterized in that, The method further includes: The water index is obtained based on the near-infrared band data and the green band data; The water index satisfies a third preset formula, which is: ； in, This refers to the water body index; This refers to the green band data; This refers to the near-infrared band data.

6. A satellite remote sensing sea ice detection device, characterized in that, include: The data preprocessing module is used to process multispectral remote sensing images of the target sea area to obtain image data, which includes green band data, shortwave infrared band data, red band data, and near-infrared band data. A multi-dimensional feature extraction module is used to extract multi-dimensional features, including edge features, first spectral features, and second spectral features, from the image data. The classification processing module is used to determine the sea ice distribution of each pixel in the image data based on the classification threshold corresponding to each feature in the multi-dimensional features, and generate a sea ice distribution map. The edge features are obtained by performing adaptive gradient enhancement and edge probability smoothing on the image data, and are acquired in the following manner: Gradient calculations are performed on the green band data based on the adaptive gradient operator to obtain the gradient values ​​in the X and Y directions. A gradient magnitude map is calculated based on the gradient values ​​in the X and Y directions. A weight function is constructed based on spatial distance and gradient similarity. The gradient magnitude map is then subjected to a weighted smoothing operation based on the weight function to obtain an edge probability map. The edge probability map is used to reflect the probability that each pixel belongs to the edge of sea ice. The edge feature is the edge probability map. The first spectral feature includes a sea ice index corrected for the influence of suspended sediment, which is obtained in the following manner: Based on the red band data and the near-infrared band data, a dynamic correction factor for suspended sediment concentration is obtained. The sea ice index is obtained based on the green band data, the shortwave infrared band data, and the dynamic correction factor for suspended sediment concentration. The second spectral feature includes the water index.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the satellite remote sensing sea ice detection method as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the satellite remote sensing sea ice detection method as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the satellite remote sensing sea ice detection method as described in any one of claims 1 to 5.

Images