A method for generating simulated oil spill images from marine radar based on backscatter suppression and radar equation constraints

By using a method based on backscatter suppression and radar equation constraints, we can generate simulated oil spill images that conform to the physical laws of marine radar. This solves the problems of insufficient sample size and lack of physical constraints, and enables the expansion of interpretable training data and the improvement of model comparability.

CN123116830APending Publication Date: 2026-11-17DALIAN MARITIME UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611336268.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies lack physical constraints on radar imaging when generating marine radar oil spill image samples, resulting in insufficient sample numbers and difficulty in interpreting the relationship between dark areas of oil slicks and oil-free sea surfaces, making fair comparisons difficult.

Method used

A method based on backscatter suppression and radar equation constraints is adopted to generate simulated images of marine radar oil spills. By acquiring sea surface, radar and oil slick parameters, a random sea surface height field is constructed, the backscatter cross section is calculated, and radar range attenuation and oil slick suppression fields are constructed. Finally, a simulated image that conforms to the physical laws of radar imaging is generated.

Benefits of technology

The generated simulation samples can batch expand the training data, maintain physical constraints, support interpretable training under different sea conditions, oil film morphology and noise conditions, and improve the comparability of the model between different experimental groups.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116830A_ABST
    Figure CN123116830A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for generating simulated oil spill images from marine radar based on backscatter suppression and radar equation constraints. It belongs to the technical fields of marine radar image simulation, marine oil spill monitoring, and artificial intelligence training data generation. The method includes the following steps: acquiring sea state parameters, radar parameters, and oil film parameters; generating a random two-dimensional sea surface height field based on the directional wave spectrum and normalizing it according to the target's significant wave height; calculating the sea surface backscattering cross section according to the sea state type; constructing range attenuation weights based on the radar equations; and locally attenuating the backscattering cross section of the oil-free sea surface to obtain a simulated marine radar image containing an oil spill. This invention can generate simulated samples with controllable sea state, wave direction, significant wave height, dominant wavelength, range attenuation, noise intensity, and oil film suppression intensity in batches, avoiding insufficient physical interpretation caused by simple grayscale superposition. It can be used to expand the training data for marine radar oil spill detection or segmentation models.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the technical fields of marine radar image simulation, sea surface backscattering modeling, marine oil spill monitoring, and artificial intelligence training data generation. It relates to a method for generating marine radar oil spill simulation images based on backscattering suppression and radar equation constraints. Background Technology

[0002] Marine oil spills can severely impact the marine ecosystem, port operational safety, and maritime traffic management. Marine radar, with its advantages of continuous observation, low cost, and suitability for deployment on shipboard or shore-based platforms, has potential applications in near-shore and offshore oil spill monitoring. When oil slicks cover the sea surface, they alter sea surface tension and viscous dissipation characteristics, suppressing capillary waves and short gravity waves. This weakens the short-wave scattering from the sea surface, which is crucial for X-band marine radar, resulting in a low-echo, dark area in radar images relative to the surrounding sea surface.

[0003] In recent years, deep learning methods have achieved good results in object detection, semantic segmentation, and instance segmentation tasks, but their training usually relies on a large number of labeled samples. The number of real marine radar oil spill image samples is limited, and it is difficult to systematically collect oil spill accident scenarios, sea conditions, and oil slick conditions. Training models based on only a small number of real samples is prone to overfitting and fails to cover different sea conditions, oil slick morphologies, distance attenuation conditions, and radar display noise.

[0004] Existing data augmentation methods often employ image enhancement, grayscale darkening, stitching and overlay, or random shape generation. While these methods can increase the number of samples, they often lack the physical constraints of radar imaging, making it difficult to explain the relationship between backscattering from oil-free sea surfaces, radar range attenuation, central blind zone, speckle noise, and dark spots on oil films. If the oil film area is simply darkened directly on the final grayscale image, the generated samples may not reflect the actual imaging mechanism of marine radar, and it is also difficult to make fair comparisons of different oil film suppression levels.

[0005] Therefore, there is a need for a method to generate marine radar oil spill simulation images that unifies the modeling of sea surface backscattering, radar equation range attenuation, and oil film scattering suppression, so that the generated samples can both expand the training data in batches and retain interpretable physical constraints. Summary of the Invention

[0006] To address the issues of insufficient real-world marine radar oil spill samples, difficulties in manual data collection, lack of imaging physical constraints in conventional image enhancement, and challenges in fairly comparing different oil film suppression conditions, this invention employs the following technical solution: a method for generating simulated marine radar oil spill images based on backscatter suppression and radar equation constraints, comprising: S1: Obtain sea surface simulation parameters, radar imaging parameters, and oil slick simulation parameters; S2: Generate a random two-dimensional sea surface height field based on the sea surface simulation parameters, and normalize and scale the random two-dimensional sea surface height field according to the target effective wave height; S3: Calculate the backscattering cross section of the oil-free sea surface based on the normalized and scaled random two-dimensional sea surface height field and sea state type; S4: Construct radar range attenuation weights and a marine radar PPI display mask based on the radar imaging parameters; S5: Obtain the oil spill area mask and construct an oil film suppression field based on the oil film simulation parameters; S6: The oil film suppression field is used to locally attenuate the backscattering cross section of the oil-free sea surface to obtain the oil area in the radar image, and the radar range attenuation weight, the noise term of the radar image and the gray-scale mapping function are combined to generate a simulated marine radar image containing oil spill.

[0007] Furthermore, the sea surface simulation parameters include one or more of the following: target significant wave height, dominant wavelength, wave direction, wind speed, sea state type, image size, spatial resolution, and random seed; The radar imaging parameters include one or more of the following: radar frequency, radar incident angle, radar center position, maximum effective display radius, center blind zone radius, range attenuation control coefficient, and noise intensity. The oil film simulation parameters include one or more of the following: oil film suppression level, dB suppression range, thickness level, edge smoothing coefficient, and shape transformation parameters.

[0008] Furthermore, the process of generating a random two-dimensional sea surface height field is as follows: S21: Construct frequency energy distribution based on JONSWAP wave spectrum, PM wave spectrum or directional wave spectrum; S22: After determining the frequency energy distribution, construct the directional extension function based on the wave direction; S23: After constructing the directional extension function, introduce random phase and time evolution factors into the directional wave spectrum; S24: After introducing a random phase and time evolution factor into the directional wave spectrum, a two-dimensional sea surface height field is generated through inverse fast Fourier transform; and the two-dimensional sea surface height field is scaled to the target effective wave height according to the following formula:

[0009] in, For the effective wave height, This represents the standard deviation of the sea surface height field.

[0010] Furthermore, the calculation of the backscattering cross section of an oil-free sea surface includes: When the sea state is low, the equivalent Bragg scattering model is used to calculate the shortwave backscattering cross section of the first-scale wave; the first-scale wave is ≤ wave scale threshold. When the sea state is high, the second-scale wave modulation backscattering cross section is calculated based on the local sea surface slope, local incident angle, and curvature modulation; the second-scale wave size > wave scale threshold. When the sea state is moderate, the equivalent Bragg scattering and the second-scale wave modulation are weighted and combined to obtain a dual-scale composite backscattering cross section.

[0011] Furthermore, the radar range attenuation weight is constructed based on the radial distance from the pixel to the radar center, and satisfies:

[0012] in: For distance decay weights, The distance from the pixel to the center of the radar. To maximize the effective display radius, This is the range attenuation control coefficient; and the center blind zone and circular PPI display range are set in the radar range attenuation weight.

[0013] Furthermore, the method for obtaining the oil spill area mask includes: Read the actual oil spill labeling image; For real oil spill labeled images, a real oil mask is extracted based on the red channel intensity threshold; the extracted region is then processed by connected component filtering, hole filling, edge smoothing, scaling, rotation or translation to obtain an oil spill area mask for simulation.

[0014] Furthermore, the oil film suppression field is a dB suppression field. The oil film backscattering retention coefficient is calculated according to the following formula.

[0015] in The linear retention coefficient is the one corresponding to the oil film suppression field. The dB suppression field is generated in the oil spill area according to a uniform distribution, a truncated normal distribution, a hierarchical constant distribution, or a spatial distribution modulated by a normalized distance field from the boundary.

[0016] The oil film suppression field is composed of the oil film thickness field. The oil film thickness field is generated based on the distance field from the pixels inside the oil spill area to the mask boundary, so that the thickness or suppression amount of the oil film in the central region is greater than that in the edge region; and according to

[0017] Calculate the oil film backscattering retention coefficient, where This is the parameter for suppressing oil film thickness.

[0018] Furthermore, the backscattering cross section containing the oil spill satisfies:

[0019] in The backscattering cross section of an oil-free sea surface. The backscattering cross section of the oil-bearing sea surface; the simulated marine radar image containing the oil spill satisfies:

[0020] ,in This is a grayscale mapping or dynamic range compression function. This refers to radar noise, speckle noise, or local clutter.

[0021] Furthermore, the process of obtaining the oil spill area mask and constructing the oil film suppression field based on the oil film simulation parameters is as follows: Extract the shape of the oil spill area from real oil spill annotation images to generate an oil film thickness field or dB suppression field; Then, taking the suppression effect of the oil film on the short-wave roughness of the sea surface and the equivalent electromagnetic echo as constraints, an oil film suppression field is constructed based on the oil film simulation parameters.

[0022] A marine radar oil spill simulation image generation system based on backscatter suppression and radar equation constraints includes: Parameter acquisition module: used to acquire sea surface simulation parameters, radar imaging parameters, and oil slick simulation parameters; Sea surface height field generation module: used to generate a random two-dimensional sea surface height field based on the sea surface simulation parameters, and to normalize and scale the random two-dimensional sea surface height field according to the target effective wave height; Backscattering calculation module: used to calculate the backscattering cross section of oil-free sea surface based on the normalized and scaled random two-dimensional sea surface height field and sea state type; Radar range attenuation module: used to construct radar range attenuation weights and a marine radar PPI display mask based on the radar imaging parameters; Oil film suppression field generation module: acquires the oil spill area mask and constructs the oil film suppression field according to the oil film simulation parameters; Image synthesis module: This module utilizes the oil film suppression field to locally attenuate the backscattering cross-section of the oil-free sea surface, obtaining the oil-bearing region in the radar image. It then combines the radar range attenuation weight, the marine radar PPI display mask, the noise term of the radar image, and the grayscale mapping function to generate a simulated marine radar image containing the oil spill. The PPI display mask is used to constrain the display boundary of the radar image.

[0023] This invention provides a method for generating simulated oil spill images from marine radar based on backscatter suppression and radar equation constraints. The method constrains the generation of dark spots on the oil film within the backscattering domain, avoiding arbitrary darkening of the final grayscale image and giving the simulated sample a clearer physical meaning of radar scattering.

[0024] This invention introduces radar range attenuation, center blind zone, and PPI circular display, making the generated image more consistent with the basic law of echo attenuation and circular display from near to far in marine radar.

[0025] This invention uses real labeled images to extract the shape of the oil film, reducing the unrealistic problems caused by completely randomly generated oil film morphology.

[0026] This invention supports batch generation of different oil film suppression levels, sea states, wave directions, significant wave heights, dominant wavelengths, and noise conditions, and can provide interpretable training samples for target detection, instance segmentation, and semantic segmentation models.

[0027] This invention can pre-generate a random backscattering library from oil-free sea surfaces and reuse it in experiments with different suppression levels, thereby improving the comparability between different experimental groups.

[0028] This invention can generate simulated samples with controllable sea state, wave direction, significant wave height, dominant wavelength, distance attenuation, noise intensity, and oil film suppression intensity in batches, avoiding the insufficient physical interpretation caused by simple grayscale superposition. It can be used to expand the training data of marine radar oil spill detection or segmentation models. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a flowchart of the marine radar oil spill simulation image generation method of the present invention; Figure 2 This is a flowchart of the random sea surface height field generation and backscattering cross section calculation in this invention; Figure 3 This is a schematic diagram of the effect of the dB suppression field or oil film thickness field on the backscattering cross section in this invention; Figure 4 This is a schematic diagram of the simulated image of the marine radar containing the oil spill and the output of the synchronous tag in this invention.

[0031] Figure 5 This is a schematic diagram comparing the simulated sample generated by this invention with a real marine radar sample.

[0032] Figure 6 The image shows the simulation results of a marine radar sample containing an oil spill, generated by this invention.

[0033] Figure 7 This is a diagram showing the combined effect of PPI display mask, radial distance attenuation, and low echo darkening of oil film in this invention. (a) PPI without oil background, (b) PPI display mask, (c) mask of oil spill area, and (d) simulated image with oil spill. Detailed Implementation

[0034] It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Figure 1 This is a flowchart of the marine radar oil spill simulation image generation method of the present invention; A method for generating simulated oil spill images from marine radar based on backscatter suppression and radar equation constraints includes: S1: Obtain sea surface simulation parameters, radar imaging parameters, and oil slick simulation parameters; S2: Generate a random two-dimensional sea surface height field based on the sea surface simulation parameters, and normalize and scale the random two-dimensional sea surface height field according to the target effective wave height; S3: Calculate the backscattering cross section of the oil-free sea surface based on the normalized and scaled random two-dimensional sea surface height field and sea state type; Figure 2 This is a flowchart of the random sea surface height field generation and backscattering cross section calculation in this invention; S4: Construct radar range attenuation weights and a marine radar PPI display mask based on the radar imaging parameters; S5: Obtain the oil spill area mask and construct an oil film suppression field based on the oil film simulation parameters; S6: The oil film suppression field is used to locally attenuate the backscattering cross section of the oil-free sea surface to obtain the oil region in the radar image. This is then combined with the radar range attenuation weight, the marine radar PPI display mask, the noise term of the radar image, and the grayscale mapping function to generate a simulated marine radar image containing the oil spill. The PPI display mask is used to constrain the display boundary of the radar image.

[0037] Figure 3 This is a schematic diagram of the effect of the dB suppression field or oil film thickness field on the backscattering cross section in this invention; Furthermore, the sea surface simulation parameters include one or more of the following: target significant wave height, dominant wavelength, wave direction, wind speed, sea state type, image size, spatial resolution, and random seed; The radar imaging parameters include one or more of the following: radar frequency, radar incident angle, radar center position, maximum effective display radius, center blind zone radius, range attenuation control coefficient, and noise intensity. The oil film simulation parameters include one or more of the following: oil film suppression level, dB suppression range, thickness level, edge smoothing coefficient, and shape transformation parameters.

[0038] Furthermore, the process of generating a random two-dimensional sea surface height field is as follows: S21: Construct frequency energy distribution based on JONSWAP wave spectrum, PM wave spectrum or directional wave spectrum; S22: After determining the frequency energy distribution, construct the directional extension function based on the wave direction; S23: After constructing the directional extension function, introduce random phase and time evolution factors into the directional wave spectrum; S24: After introducing a random phase and time evolution factor into the directional wave spectrum, a two-dimensional sea surface height field is generated through inverse fast Fourier transform; and the two-dimensional sea surface height field is scaled to the target effective wave height according to the following formula:

[0039] in, For the effective wave height, Let be the standard deviation of the sea surface height field. Further, the calculation of the backscattering cross section of an oil-free sea surface includes: When the sea state is low, the equivalent Bragg scattering model is used to calculate the shortwave backscattering cross section of the first-scale wave; the first-scale wave is ≤ wave scale threshold. When the sea state is high, the second-scale wave modulation backscattering cross section is calculated based on the local sea surface slope, local incident angle, and curvature modulation; the second-scale wave size > wave scale threshold. When the sea state is moderate, the equivalent Bragg scattering and the second-scale wave modulation are weighted and combined to obtain a dual-scale composite backscattering cross section.

[0040] Furthermore, the radar range attenuation weight is constructed based on the radial distance from the pixel to the radar center, and satisfies:

[0041] in: For distance decay weights, The distance from the pixel to the center of the radar. To maximize the effective display radius, This is the range attenuation control coefficient; and the center blind zone and circular PPI display range are set in the radar range attenuation weight.

[0042] Furthermore, the method for obtaining the oil spill area mask includes: Read the actual oil spill labeling image; For real oil spill labeled images, a real oil mask is extracted based on the red channel intensity threshold; the extracted region is then processed by connected component filtering, hole filling, edge smoothing, scaling, rotation or translation to obtain an oil spill area mask for simulation.

[0043] Furthermore, the oil film suppression field is a dB suppression field. The oil film backscattering retention coefficient is calculated according to the following formula.

[0044] in The linear retention coefficient is the one corresponding to the oil film suppression field. The dB suppression field is generated in the oil spill area according to a uniform distribution, a truncated normal distribution, a hierarchical constant distribution, or a spatial distribution modulated by a normalized distance field from the boundary.

[0045] The oil film suppression field is composed of the oil film thickness field. The oil film thickness field is generated based on the distance field from the pixels inside the oil spill area to the mask boundary, so that the thickness or suppression amount of the oil film in the central region is greater than that in the edge region; and according to

[0046] Calculate the oil film backscattering retention coefficient, where This is the parameter for suppressing oil film thickness.

[0047] Furthermore, the backscattering cross section containing the oil spill satisfies:

[0048] in The backscattering cross section of an oil-free sea surface. The backscattering cross section of the oil-bearing sea surface; the simulated marine radar image containing the oil spill satisfies:

[0049] ,in This is a grayscale mapping or dynamic range compression function. This refers to radar noise, speckle noise, or local clutter.

[0050] Furthermore, the process of obtaining the oil spill area mask and constructing the oil film suppression field based on the oil film simulation parameters is as follows: Extract the shape of the oil spill area from real oil spill annotation images to generate an oil film thickness field or dB suppression field; Then, taking the suppression effect of the oil film on the short-wave roughness of the sea surface and the equivalent electromagnetic echo as constraints, an oil film suppression field is constructed based on the oil film simulation parameters.

[0051] A marine radar oil spill simulation image generation system based on backscatter suppression and radar equation constraints includes: Parameter acquisition module: used to acquire sea surface simulation parameters, radar imaging parameters, and oil slick simulation parameters; Sea surface height field generation module: used to generate a random two-dimensional sea surface height field based on the sea surface simulation parameters, and to normalize and scale the random two-dimensional sea surface height field according to the target effective wave height; Backscattering calculation module: used to calculate the backscattering cross section of oil-free sea surface based on the normalized and scaled random two-dimensional sea surface height field and sea state type; Radar range attenuation module: used to construct radar range attenuation weights and a marine radar PPI display mask based on the radar imaging parameters; Oil film suppression field generation module: acquires the oil spill area mask and constructs the oil film suppression field according to the oil film simulation parameters; Image synthesis module: used to locally attenuate the backscattering cross section of the oil-free sea surface using the oil film suppression field to obtain the oil area in the radar image, and combine the radar range attenuation weight, the marine radar PPI display mask, the noise term of the radar image and the grayscale mapping function to generate a marine radar simulation image containing oil spill.

[0052] The method also includes pre-generating a random backscattering library of oil-free sea surfaces in batches, and reusing or randomly selecting the random backscattering library of oil-free sea surfaces in simulation experiments with different oil film suppression levels, so as to maintain comparability of different oil film suppression experiments in terms of sea surface background distribution.

[0053] The simulated marine radar image containing the oil spill is output synchronously with the corresponding oil spill area mask, semantic segmentation label, instance segmentation label or target detection label, which is used to train or evaluate the oil spill detection model, instance segmentation model or semantic segmentation model.

[0054] The method further includes superimposing non-uniform near-field highlights, central blind zones, fine-particle speckle noise, local strong echo clutter, or radar display circular boundaries onto the generated image, while ensuring that the backscattering suppression relationship within the oil spill area is not canceled out when superimposing near-field highlights.

[0055] Example 1: Method for Generating Simulated Images of Marine Radar Containing Oil Spills In one embodiment, the image size is set to 1024×1024 pixels, the radar display area is a circular PPI area, and a blind zone is set in the center. Sea surface parameters include the target's significant wave height. Dominant wavelength , wave direction And wind speed U. The radar frequency can be set to the X-band, and the radar range attenuation control coefficient is set according to the display range.

[0056] A two-dimensional sea surface height field z(x,y,t) is generated based on the directional wave spectrum. After generation, the standard deviation of the height field is calculated. The effective wave height is determined according to the following formula:

[0057] If the current significant wave height is inconsistent with the target significant wave height, the height field is scaled as a whole according to the ratio of the target significant wave height to the current significant wave height. By changing the random phase, random seed, or time evolution factor, multiple different random sea surface images can be generated under the same sea state parameters.

[0058] For low sea states, calculate the equivalent Bragg scattering term. For high sea states, calculate the large-scale tilt-curvature modulation term. For moderate sea states, the following dual-scale composite approach is adopted:

[0059] in, The backscattering cross section of an oil-free sea surface. Slope modulation factor, These are the weighting coefficients. The above model can be implemented as an equivalent model based on sea state type and radar display requirements.

[0060] Construct distance attenuation weights based on the radial distance R from the pixel to the radar center:

[0061] in, To maximize the effective display radius, This is the distance attenuation control coefficient. For pixels within the central blind zone, the display value is set to a preset low value or an invalid value; for pixels outside the PPI circular display area, they are set to the background value.

[0062] Read the actual oil spill label image and extract the labeled area based on the red channel threshold. One possible extraction condition is:

[0063] Where Rc, Gc, and Bc are the red, green, and blue channel values, respectively; TR is the red channel threshold; and Tc is the red relative dominance threshold. The extracted mask can be used for connected component filtering, hole filling, edge smoothing, scaling, and rotation.

[0064] In one implementation, the oil film suppression field uses a dB suppression amount. The oil film backscattering retention coefficient is:

[0065] For example, weak suppression corresponds to about 1-2 dB, moderate suppression corresponds to about 3 dB, and strong suppression corresponds to about 4-5 dB. The oil film area can be characterized by a uniform distribution, a truncated normal distribution, or a spatial distribution that varies with distance from the boundary.

[0066] In another implementation, the oil film thickness field is first constructed based on the distance from the pixels inside the oil spill area to the boundary. Then calculate the retention factor according to the following exponential relationship:

[0067] in, This is the oil film thickness suppression strength parameter. The center of the oil film region can be set to be relatively thick, and the edges relatively thin, so that the simulated oil film dark spots have a difference between the center and the edges.

[0068] The oil-containing backscattering cross section is obtained by the following formula:

[0069] The final simulated marine radar image containing the oil spill is as follows:

[0070] in, This is a grayscale mapping or dynamic range compression function. This refers to speckle noise, local clutter, or radar display noise. If near-range high brightness or strong central echo is superimposed, the backscattering suppression relationship within the oil film area should not be canceled out.

[0071] Example 2: Reuse of a random, oil-free sea surface backscattering library In one embodiment, a random backscattering library of oil-free sea surfaces is generated in batches according to multiple combinations of sea state parameters. For example, each sea state combination generates several backscattering maps of oil-free sea surfaces, and the corresponding sea state parameters, random seeds, and backscattering model parameters are recorded.

[0072] When comparing different oil film suppression levels, background data is extracted from the same random backscattering library of oil-free sea surfaces, and then different oil film suppression fields are applied to each. This avoids unfair comparisons caused by different background distributions in different oil film suppression experiments.

[0073] Figure 4 This is a schematic diagram of the simulated image of the marine radar containing the oil spill and the output of the synchronous tag in this invention.

[0074] Figure 5 This is a schematic diagram comparing the simulated sample generated by this invention with a real marine radar sample.

[0075] Example 3: Synchronous Label Output and Model Training Applications While generating simulated marine radar images containing oil spills, the invention can simultaneously output semantic segmentation masks, instance segmentation polygon labels, target detection bounding box labels, or COCO format labels. These labels are obtained from the oil spill area mask through scale transformation, coordinate normalization, and format conversion.

[0076] The generated data can be combined with a small number of real marine radar oil spill images to train YOLO series instance segmentation models, DeepLabV3+, UNet, or other semantic segmentation models. This training purpose is not limited to a specific model structure; the core is that the simulation samples are generated jointly by sea surface backscattering, radar range attenuation, and oil film scattering suppression.

[0077] like Figure 6 As shown, the generated simulated samples can be combined with real marine radar oil spill samples to form training data, which can be used to improve the adaptability of oil spill detection or segmentation models to real-world scenarios. This result figure is used to illustrate the output format of the present invention and does not limit the specific number of training sets, image size, or model structure.

[0078] like Figure 7 As shown, (a) PPI without oil background, (b) PPI display mask, (c) mask of the oil spill area, and (d) simulated image containing oil spill. Following the above steps, multiple sets of simulated marine radar images containing oil spills can be generated. These images simultaneously demonstrate PPI circular display, radial distance attenuation, background sea clutter texture, and low-echo darkening effects in the oil slick area. Different random seeds, sea state parameters, and oil slick suppression levels correspond to different image background and dark spot intensities.

[0079] It should be noted that the wave spectrum type, backscattering model, range attenuation form, oil film suppression function, and label format in the above embodiments can all be replaced or combined according to the application scenario. As long as the basic idea is to jointly model backscattering from an oil-free sea surface, radar range attenuation, and oil film scattering suppression in marine radar oil spill simulation, and output marine radar simulation images containing oil spills and their labels, they should all fall within the protection scope of this invention.

[0080] The oil film suppression parameters described in this invention are used for simulation image generation and training data augmentation, and are not equivalent to the inversion results of the actual oil film thickness, dielectric constant, or emulsification degree.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention 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 or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating simulated oil spill images from marine radar based on backscatter suppression and radar equation constraints, characterized in that, include: S1: Obtain sea surface simulation parameters, radar imaging parameters, and oil slick simulation parameters; S2: Generate a random two-dimensional sea surface height field based on the sea surface simulation parameters, and normalize and scale the random two-dimensional sea surface height field according to the target effective wave height; S3: Calculate the backscattering cross section of the oil-free sea surface based on the normalized and scaled random two-dimensional sea surface height field and sea state type; S4: Construct radar range attenuation weights and a marine radar PPI display mask based on the radar imaging parameters; S5: Obtain the oil spill area mask and construct an oil film suppression field based on the oil film simulation parameters; S6: The backscattering cross section of the oil-free sea surface is locally attenuated using the oil film suppression field to obtain the oil region in the radar image. The simulated marine radar image containing the oil spill is generated by combining the radar range attenuation weight, the marine radar PPI display mask, the noise term of the radar image, and the grayscale mapping function.

2. The method for generating simulated marine radar oil spill images based on backscatter suppression and radar equation constraints according to claim 1, characterized in that, The sea surface simulation parameters include one or more of the following: target significant wave height, dominant wavelength, wave direction, wind speed, sea state type, image size, spatial resolution, and random seed. The radar imaging parameters include one or more of the following: radar frequency, radar incident angle, radar center position, maximum effective display radius, center blind zone radius, range attenuation control coefficient, and noise intensity. The oil film simulation parameters include one or more of the following: oil film suppression level, dB suppression range, thickness level, edge smoothing coefficient, and shape transformation parameters.

3. The method for generating simulated marine radar oil spill images based on backscatter suppression and radar equation constraints according to claim 1, characterized in that, The process of generating a random two-dimensional sea surface height field is as follows: S21: Construct frequency energy distribution based on JONSWAP wave spectrum, PM wave spectrum or directional wave spectrum; S22: After the frequency energy distribution, construct the directional extension function based on the wave direction; S23: After constructing the directional extension function, random phase and time evolution factors are introduced into the directional wave spectrum; S24: After introducing a random phase and time evolution factor into the directional wave spectrum, a two-dimensional sea surface height field is generated through inverse fast Fourier transform; and the two-dimensional sea surface height field is scaled to the target effective wave height according to the following formula: in, For the effective wave height, This represents the standard deviation of the sea surface height field.

4. The method for generating simulated marine radar oil spill images based on backscatter suppression and radar equation constraints according to claim 1, characterized in that, The calculation of the backscattering cross section of an oil-free sea surface includes: When the sea state is low, the equivalent Bragg scattering model is used to calculate the shortwave backscattering cross section of the first-scale wave; the first-scale wave is ≤ wave scale threshold. When the sea state is high, the second-scale wave modulation backscattering cross section is calculated based on the local sea surface slope, local incident angle, and curvature modulation; the second-scale wave size > wave scale threshold. When the sea state is moderate, the equivalent Bragg scattering and the second-scale wave modulation are weighted and combined to obtain a dual-scale composite backscattering cross section.

5. The method for generating simulated marine radar oil spill images based on backscatter suppression and radar equation constraints according to claim 1, characterized in that, The radar range attenuation weight is constructed based on the radial distance from the pixel to the radar center, and satisfies: in, For distance decay weights, The distance from the pixel to the center of the radar. To maximize the effective display radius, This is the range attenuation control coefficient; and the center blind zone and circular PPI display range are set in the radar range attenuation weight.

6. The method for generating simulated marine radar oil spill images based on backscatter suppression and radar equation constraints according to claim 1, characterized in that, The mask for obtaining the oil spill area includes: Read the actual oil spill label image; For real oil spill labeled images, a real oil mask is extracted based on the red channel intensity threshold; the extracted regions are then processed by sequentially filtering connected components, filling holes, smoothing edges, scaling, rotating or translating to obtain an oil spill area mask for simulation.

7. The method for generating simulated marine radar oil spill images based on backscatter suppression and radar equation constraints according to claim 1, characterized in that, The oil film suppression field is a dB suppression field. The oil film backscattering retention coefficient is calculated according to the following formula. in The linear retention coefficient is the one corresponding to the oil film suppression field. The dB suppression field is generated in the oil spill area according to a uniform distribution, a truncated normal distribution, a hierarchical constant distribution, or a spatial distribution modulated by a normalized distance field from the boundary. The oil film suppression field is composed of the oil film thickness field. The oil film thickness field is generated based on the distance field from the pixels inside the oil spill area to the mask boundary, so that the thickness or suppression amount of the oil film in the central region is greater than that in the edge region; and the oil film backscattering retention coefficient is calculated according to the following formula: in: This is the parameter for suppressing oil film thickness.

8. The method for generating simulated marine radar oil spill images based on backscatter suppression and radar equation constraints according to claim 1, characterized in that, The backscattering cross section containing the oil spill satisfies: in: The backscattering cross section of an oil-free sea surface. The backscattering cross section of the oil-bearing sea surface; the simulated marine radar image containing the oil spill satisfies: in: This is a grayscale mapping or dynamic range compression function. This refers to radar noise, speckle noise, or local clutter.

9. The method for generating simulated marine radar oil spill images based on backscatter suppression and radar equation constraints according to claim 1, characterized in that, The process of obtaining the oil spill area mask and constructing the oil film suppression field based on the oil film simulation parameters is as follows: Extract the shape of the oil spill area from real oil spill annotation images to generate an oil film thickness field or dB suppression field; Then, taking the suppression effect of the oil film on the short-wave roughness and equivalent electromagnetic echo of the sea surface as constraints, an oil film suppression field is constructed based on the oil film simulation parameters.

10. A marine radar oil spill simulation image generation system based on backscatter suppression and radar equation constraints, characterized in that, include: Parameter acquisition module: used to acquire sea surface simulation parameters, radar imaging parameters, and oil slick simulation parameters; Sea surface height field generation module: used to generate a random two-dimensional sea surface height field based on the sea surface simulation parameters, and to normalize and scale the random two-dimensional sea surface height field according to the target effective wave height; Backscattering calculation module: used to calculate the backscattering cross section of oil-free sea surface based on the normalized and scaled random two-dimensional sea surface height field and sea state type; Radar range attenuation module: used to construct radar range attenuation weights and a marine radar PPI display mask based on the radar imaging parameters; Oil film suppression field generation module: acquires the oil spill area mask and constructs the oil film suppression field according to the oil film simulation parameters; Image synthesis module: used to locally attenuate the backscattering cross section of the oil-free sea surface using the oil film suppression field to obtain the oil area in the radar image, and combine the radar range attenuation weight, the marine radar PPI display mask, the noise term of the radar image and the grayscale mapping function to generate a marine radar simulation image containing oil spill.