A method, device, medium and product for detecting marine oil spills by spaceborne synthetic aperture radar

Abstract

The present application discloses a method, device, medium and product for detecting ocean oil spills by spaceborne synthetic aperture radar, which relates to the field of oil spill detection. The method includes obtaining a spaceborne synthetic aperture radar image of an ocean oil spill area; constructing an ocean oil spill detection model based on the Faster R-CNN model and the UNet++ model; in the ocean oil spill detection model, according to the spaceborne synthetic aperture radar image, using the Faster R-CNN model to determine candidate boxes containing the ocean oil spill area; and according to the spaceborne synthetic aperture radar image corresponding to the candidate boxes, using the UNet++ model to detect the ocean oil spill area to obtain a detection result; according to the spaceborne synthetic aperture radar image of the ocean oil spill area, using the ocean oil spill detection model to obtain the detection result of the ocean oil spill area. The present application can achieve more efficient and accurate detection of spaceborne SAR ocean oil spills.

Description

Technical Field

[0001] This application relates to the field of oil spill detection, and particularly to a spaceborne synthetic aperture radar ocean oil spill detection method, device, medium and product. Background Art

[0002] With the rapid development of marine resource development and the maritime transportation industry, oil spill accidents in the ocean occur frequently, and the pollution caused by oil leakage into the sea can persist in the marine environment for many years. Therefore, it is crucial to develop intelligent, efficient, and rapid oil spill detection technologies to mitigate the harmful effects of oil pollution on the marine environment.

[0003] Compared with ships and airplanes, satellites are low-cost, high-efficiency, and have a large coverage area, making them more suitable for oil spill surveillance. The operation of spaceborne synthetic aperture radar (SAR) is not affected by clouds and light, and can perform high-resolution observations on large areas. This ability makes it a unique microwave instrument for timely monitoring of ocean oil spills. Oil spills appear on the sea surface in the form of floating viscoelastic hydrocarbon layers. Due to the increased surface tension, these hydrocarbon layers dampen the short surface gravity waves and capillary waves generated by the wind. This damping effect results in a reduction in the radar echo from the ocean surface, causing the oil spill to appear as a dark spot area in the synthetic aperture radar image. However, suspected oil spill phenomena such as biological films, low wind speed areas, grease ice, rain cells, wind shadow areas near the coastline, upwelling areas, and internal waves also weaken the radar backscattering, resulting in dark spots in the SAR image. Under certain conditions, multiple types of dark spots may appear simultaneously in the SAR image, making oil spill detection more complex.

[0004] Traditional mathematical models and image processing techniques have been widely used to detect oil spills in single-polarization SAR images. Early studies used a combination of adaptive thresholding and statistical modeling to achieve automatic oil spill detection. During the oil spill detection process, the adaptive thresholding method was used to identify dark spots in the SAR image. However, due to the influence of speckle noise, it is difficult to accurately detect dark spots in the image. For classifiers based on multivariate probability distribution functions, it is crucial to combine prior information about the likelihood of observed oil spills with auxiliary data including wind speed. Image processing techniques including texture analysis and wavelet transform have also been used for oil spill detection. However, when the contrast between the oil spill and the surrounding area is low, these two methods may produce inaccurate detection results.

[0005] The above method relies entirely on the radar echo intensity of the ocean surface measured by single-polarization SAR to detect oil spills. The target scattering matrix measured by full-polarization SAR includes phase and amplitude information. According to the polarization target decomposition theory, various polarization characteristic parameters can be decomposed from the scattering matrix. Since these polarization characteristics can describe and quantify different scattering mechanisms, they have currently been widely used in oil spill detection. However, a previous study showed that the polarization characteristics extracted from full-polarization SAR data are affected by multiplicative and additive system noise, and thus are not suitable for analyzing scattering characteristics and features. In addition, the analysis of airborne full-polarization SAR observation data of the oil spill in the Deepwater Horizon disaster showed that the radar backscattering of the oil-covered area and the clean sea surface is mainly related to surface Bragg scattering. Therefore, the difference in polarization characteristics between oil spills and clean waters may be caused by instrument noise rather than different scattering mechanisms. In addition, the coverage range of full-polarization SAR (20 km to 70 km) is much smaller than that of single-polarization SAR (250 km to 500 km), so it is not suitable for oil spill monitoring.

[0006] Due to the excellent ability of machine learning techniques to map complex non-linear relationships between inputs and outputs, they have been increasingly applied to remote sensing image classification and segmentation tasks. Various machine learning models, including artificial neural networks, genetic algorithms, and random forests, have been used to detect oil spills in SAR images. Several oil spill detection algorithms based on artificial neural networks use two sets of models: one for segmenting dark spots and the other for classifying dark spots. This dual-model method increases the complexity of the oil spill detection process. In addition, the inputs of artificial neural network models are usually selected subjectively by humans, so this method cannot select the most representative features for oil spill detection. Although random forests and genetic algorithms can eliminate redundant inputs and identify the most important physical, geometric, and texture features, they still require human intervention to achieve oil spill detection.

[0007] Deep learning models consist of multiple processing layers and can effectively learn various features in images. For example, convolutional neural networks can automatically extract multi-level image features from large datasets and fuse information, thus facilitating fast and accurate image classification and segmentation. Various deep learning models such as YOLOv4, Faster R-CNN, and VGG-16 have been used in oil spill detection. These classification or object detection models have been optimized for classifying dark spots or locating oil spills in SAR images in practical applications. In addition, some semantic and instance segmentation models have been used to segment oil spills and extract their extents. However, most segmentation models require sub-images containing oil spills as input. To obtain these sub-images, the location of the oil spill in the entire image must be determined in advance. The original SAR image can be divided into several sub-images, and the oil spill detection process must be repeated for each sub-image, thus increasing the detection time and computational load. In addition, dividing the complete SAR image into sub-images may split large or long and narrow oil spills into several parts, thus increasing the risk of individual oil spills not being detected.

[0008] Based on the development status and existing problems of the above-mentioned prior art, there is an urgent need to provide a new SAR ocean oil spill detection method to solve the problems such as the difficulty in balancing detection accuracy and efficiency, difficult feature extraction, and high false detection rate in the prior art, and to achieve more efficient and accurate detection of spaceborne SAR ocean oil spills, providing strong technical support for marine environmental protection and oil spill accident emergency handling. Summary of the Invention

[0009] The purpose of this application is to provide a spaceborne synthetic aperture radar ocean oil spill detection method, device, medium, and product, which can achieve more efficient and accurate detection of spaceborne SAR ocean oil spills.

[0010] To achieve the above purpose, this application provides the following solutions:

[0011] In the first aspect, this application provides a spaceborne synthetic aperture radar ocean oil spill detection method, and the spaceborne synthetic aperture radar ocean oil spill detection method includes:

[0012] Obtain a spaceborne synthetic aperture radar image of the ocean oil spill area;

[0013] Based on the Faster R-CNN model and the UNet++ model, construct an ocean oil spill detection model; in the ocean oil spill detection model, according to the spaceborne synthetic aperture radar image, use the Faster R-CNN model to determine the candidate boxes containing the ocean oil spill area; and according to the spaceborne synthetic aperture radar image corresponding to the candidate boxes, use the UNet++ model to detect the ocean oil spill area to obtain the detection result; the detection result includes: the boundary, shape, and internal detail features of the ocean oil spill area;

[0014] Based on the spaceborne synthetic aperture radar (SAR) image of the marine oil spill area, using a marine oil spill detection model, the detection result of the marine oil spill area is obtained.

[0015] Optionally, the marine oil spill detection model further includes:

[0016] Using the non-maximum suppression algorithm to sort the confidence scores of the candidate boxes containing the marine oil spill area;

[0017] Starting from the candidate box with the highest confidence score, eliminating the candidate boxes whose overlapping area with the current candidate box exceeds the area threshold and whose confidence scores are less than the confidence score threshold, to obtain the candidate boxes after elimination.

[0018] Optionally, the step of obtaining the detection result of the marine oil spill area based on the spaceborne synthetic aperture radar image of the marine oil spill area by using the marine oil spill detection model specifically includes:

[0019] Performing preprocessing operations on the spaceborne synthetic aperture radar image of the marine oil spill area; the preprocessing operations include: speckle noise suppression, incidence angle correction, land masking, and data normalization;

[0020] Based on the preprocessed spaceborne synthetic aperture radar image, using the marine oil spill detection model, the detection result of the marine oil spill area is obtained.

[0021] Optionally, the step of performing preprocessing operations on the spaceborne synthetic aperture radar image of the marine oil spill area specifically includes:

[0022] Using moving average filtering to suppress speckle noise in the spaceborne synthetic aperture radar image;

[0023] Using the CMOD5.N geophysical model function to perform intensity scaling on the spaceborne synthetic aperture radar image after speckle noise suppression in the range direction, to obtain the spaceborne synthetic aperture radar image after incidence angle correction;

[0024] Using the global high-resolution digital elevation model to mask the land area in the spaceborne synthetic aperture radar image after incidence angle correction and enhance the image contrast, to obtain the spaceborne synthetic aperture radar image after land masking;

[0025] Performing data normalization on the spaceborne synthetic aperture radar image after land masking, to obtain the preprocessed spaceborne synthetic aperture radar image.

[0026] Optionally, using the CMOD5.N geophysical model function to perform intensity scaling on the spaceborne synthetic aperture radar image after speckle noise suppression in the range direction, to obtain the spaceborne synthetic aperture radar image after incidence angle correction, specifically includes:

[0027] Using the formula Perform an intensity scaling operation on the spaceborne synthetic aperture radar (SAR) image after speckle noise suppression in the range direction;

[0028] Wherein, is the vertical polarization radar backscattering coefficient, respectively represent wind speed, relative wind direction and incident angle, ssr represents the sea surface roughness after incident angle correction, and CMOD5.N() is the CMOD5.N geophysical model function.

[0029] Optionally, based on the preprocessed spaceborne SAR image, use an oil spill detection model in the ocean to obtain the detection result of the oil spill area in the ocean. After that, it further includes:

[0030] Visualize the detection result.

[0031] In a second aspect, the present application provides a spaceborne synthetic aperture radar oil spill detection device in the ocean. The spaceborne synthetic aperture radar oil spill detection device in the ocean includes:

[0032] An image acquisition module for acquiring a spaceborne synthetic aperture radar image of the oil spill area in the ocean;

[0033] An oil spill detection model construction module for constructing an oil spill detection model in the ocean based on the Faster R-CNN model and the UNet++ model; in the oil spill detection model, according to the spaceborne synthetic aperture radar image, use the Faster R-CNN model to determine the candidate boxes containing the oil spill area in the ocean; and according to the spaceborne synthetic aperture radar image corresponding to the candidate boxes, use the UNet++ model to detect the oil spill area in the ocean to obtain the detection result; the detection result includes: the boundary, shape and internal detail features of the oil spill area in the ocean;

[0034] A detection result determination module for obtaining the detection result of the oil spill area in the ocean by using the oil spill detection model according to the spaceborne synthetic aperture radar image of the oil spill area in the ocean.

[0035] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the computer program to implement the spaceborne synthetic aperture radar oil spill detection method described above.

[0036] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the spaceborne synthetic aperture radar oil spill detection method described above.

[0037] Fifth aspect, the present application provides a computer program product, including a computer program which, when executed by a processor, implements the above-mentioned spaceborne synthetic aperture radar marine oil spill detection method.

[0038] According to the specific embodiments provided by the present application, the present application has the following technical effects:

[0039] The present application provides a spaceborne synthetic aperture radar marine oil spill detection method, device, medium and product. Based on the Faster R-CNN model and the UNet++ model, a marine oil spill detection model is constructed. Based on Faster R-CNN, preliminary target positioning and recognition of the marine oil spill area are carried out, and the characteristics of quickly and accurately detecting target objects are used to determine the possible range of the oil spill area. At the same time, with the help of the advantages of UNet++ in semantic segmentation, more refined segmentation and feature extraction are carried out on the located area, so as to accurately depict information such as the boundary, shape and internal detail features of the oil spill area. By using the marine oil spill detection model to detect marine oil spills, the present application can solve the problems in the prior art such as the difficulty in balancing detection accuracy and efficiency, difficult feature extraction and high false detection rate, and realize more efficient and accurate detection of spaceborne SAR marine oil spills, providing strong technical support for marine environmental protection and oil spill accident emergency treatment. Description of the Drawings

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0041] Figure 1 It is a schematic flow chart of a spaceborne synthetic aperture radar marine oil spill detection method in an embodiment of the present application. Detailed Embodiments

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present application.

[0043] To make the above objects, features and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0044] In an exemplary embodiment, as Figure 1As shown in the figure, a method for detecting ocean oil spills using spaceborne synthetic aperture radar is provided. This method includes the following steps S101 to S103. Among them:

[0045] S101, obtain a spaceborne synthetic aperture radar image of the ocean oil spill area;

[0046] S102, based on the Faster R-CNN model and the UNet++ model, construct an ocean oil spill detection model; in the ocean oil spill detection model, according to the spaceborne synthetic aperture radar image, use the Faster R-CNN model to determine the candidate boxes containing the ocean oil spill area; and according to the spaceborne synthetic aperture radar image corresponding to the candidate boxes, use the UNet++ model to detect the ocean oil spill area to obtain the detection result; the detection result includes: the boundary, shape and internal detail features of the ocean oil spill area;

[0047] The Faster R-CNN (Fast Region-based Convolutional Neural Network) model includes a feature extraction network (such as using VGG-16 as the basic network), a Region Proposal Network (RPN), and a classification and regression head. The Faster R-CNN model is trained using the labeled training set data. During the training process, the feature extraction network is responsible for extracting the deep semantic features in the SAR image, the RPN network generates a series of candidate boxes that may contain the ocean oil spill area based on these features, and the classification and regression head classifies (judges whether it is an oil spill area) and regresses (precisely adjusts the position and size of the candidate boxes) these candidate boxes. The weight parameters of the model are continuously adjusted through the backpropagation algorithm to minimize the loss function, and the loss function comprehensively considers the classification loss and the regression loss.

[0048] The nested U-shaped structure of the UNet++ model is one of its core advantages. Compared with the traditional U-shaped structure, this structure is not a simple single-layer encoding-decoding design, but contains multiple encoding-decoding paths. The existence of multiple paths enables the model to capture image features at different scales and levels simultaneously. When facing complex and variable marine oil spill images, whether it is a large-area oil spill patch or a tiny oil spill spot, it can be accurately perceived. Features at different levels are fused and transmitted in these paths, forming a rich feature information network. For example, in the encoding part, as the image is gradually downsampled, the spatial resolution of the image gradually decreases, but the semantic information is continuously enhanced. Each layer can extract feature information with different levels of abstraction, from low-level features such as edges and textures to more semantic shape and region features. The decoding part then cleverly upsamples these feature information back to the original image resolution and fuses features from different levels during this process. This cross-level feature fusion mechanism greatly improves the model's ability to perceive the boundaries and internal details of the oil spill area. For example, for those pixel points at the boundary of the oil spill area that are blurred and have a subtle transition with the background, by fusing low-level edge features and high-level semantic features, it is possible to more accurately judge their belonging categories, thus finely depicting the boundary contour of the oil spill area. For internal details, the distribution uniformity, shape changes, etc. can also be better identified based on multi-scale features. During the training process, the UNet++ model is trained using the labeled training set data. The model specifically performs semantic segmentation on the possibly oil spill areas preliminarily located by Faster R-CNN. The model parameters are adjusted by minimizing the cross-entropy loss function, which can effectively measure the difference between the model prediction result and the true label, prompting the model to continuously optimize the parameters to improve the prediction accuracy. The trained model can assign labels to each pixel in the image (usually two categories: oil spill area and background in marine oil spill detection). This pixel-level accurate classification ability significantly improves the detection accuracy and detail level compared with the traditional method that can only make a rough judgment on the area, laying a solid foundation for subsequent work such as oil spill quantification assessment and precise definition of the pollution range.

[0049] The trained Faster R-CNN model and UNet++ model are fused using a high-performance computer. The specific fusion method is as follows: The preprocessed spaceborne synthetic aperture radar (SAR) image is provided as input data to the trained Faster R-CNN model. The Faster R-CNN model utilizes a feature extraction network to deeply mine and learn features such as complex textures and gray-scale variations in the SAR image. After extracting rich feature information, the Region Proposal Network (RPN) constructs a series of candidate boxes that may contain oil spill areas based on these features. Each candidate box contains location information and a corresponding confidence score, which reflects the likelihood of the candidate box containing an oil spill area.

[0050] To further improve the detection efficiency and reduce the interference of redundant information, the Non-Maximum Suppression (NMS) algorithm is introduced. The NMS algorithm sorts all candidate boxes according to the confidence scores, and then starts from the candidate box with the highest confidence score. It suppresses the candidate boxes whose overlapping area with it exceeds the area threshold (e.g., 0.5) and whose confidence scores are less than the confidence score threshold. This effectively filters out the most representative and reliable candidate boxes for oil spill areas, thus greatly reducing the amount of data for subsequent processing and improving the accuracy and efficiency of detection.

[0051] The spaceborne synthetic aperture radar image corresponding to the candidate box of the oil spill area obtained after NMS processing (i.e., the oil spill dark spot sub-image) is used as the input to the UNet++ model. The UNet++ model performs fine oil spill segmentation based on the received spaceborne synthetic aperture radar image. Its encoding part first performs downsampling operations on the input oil spill dark spot sub-image. Through a combination of a series of convolutional layers and pooling layers, the resolution of the image gradually decreases, but different levels of feature information are extracted at each layer, from relatively shallow local features such as edges and textures to deep semantic features. When the encoding part finishes the downsampling operation, the decoding part starts to work. The decoding process is a process of gradually upsampling to restore the image resolution, and during this process, the UNet++ model fuses and transfers the feature information extracted at different levels in the encoding part through a special skip connection mechanism. Specifically, each decoding layer fuses the feature information from the corresponding encoding layer and more shallow encoding layers, so that the model can make full use of multi-scale feature information to perform more accurate segmentation and feature extraction of the oil spill area while restoring the image resolution.

[0052] After completing model fusion and obtaining preliminary detection results, using the validation set data to optimize and adjust the fused model is a crucial step in improving the model's performance and reliability. As independent data samples from the training set, the validation set data can truly reflect the model's performance when facing unknown data. By applying the fused model to the validation set data, various performance indicators of the model on the validation set are comprehensively evaluated. Among them, accuracy is the proportion of the number of samples correctly predicted as oil spill areas and correctly predicted as non-oil spill areas to the total number of samples, which reflects the overall accuracy of the model; recall is the proportion of the number of samples that are actually oil spill areas and are correctly predicted as oil spill areas by the model to the total number of samples in the actual oil spill area, which reflects the model's ability to capture oil spill areas. Based on the evaluation results of the performance indicators, the hyperparameters of the model are finely adjusted. Through rounds of validation set evaluation and hyperparameter adjustment processes, the fused model is continuously optimized to maintain a high detection speed while continuously improving the detection accuracy and generalization ability, and ultimately be able to adapt to the oil spill detection task of spaceborne SAR images in various complex marine environments, providing strong technical support and guarantee for marine environmental protection and oil spill accident emergency handling.

[0053] S103. According to the spaceborne synthetic aperture radar image of the marine oil spill area, use the marine oil spill detection model to obtain the detection result of the marine oil spill area.

[0054] S103 specifically includes:

[0055] S31. Perform preprocessing operations on the spaceborne synthetic aperture radar image of the marine oil spill area; the preprocessing operations include: speckle noise suppression, incidence angle correction, land masking, and data normalization;

[0056] In a specific embodiment, a server equipped with an Intel i9-10900x CPU, 256 GB of RAM, and an NVIDIA RTX 3090 graphics card is used to perform preprocessing operations on the spaceborne synthetic aperture radar image of the marine oil spill area;

[0057] S31 specifically includes:

[0058] S1. Use moving average filtering to perform speckle noise suppression on the spaceborne synthetic aperture radar image; the spaceborne synthetic aperture radar image after speckle noise suppression is filtered and smoothed to a 100-meter pixel interval;

[0059] S2. Use the CMOD5.N geophysical model function to perform intensity scaling operations on the spaceborne synthetic aperture radar image after speckle noise suppression in the range direction to reduce the variation of radar backscattering with the incidence angle;

[0060] Specifically, use the formula Perform an intensity scaling operation on the spaceborne synthetic aperture radar (SAR) image after speckle noise suppression in the range direction;

[0061] Among them, is the vertical polarization radar backscattering coefficient, respectively represent wind speed, relative wind direction, and incident angle. ssr represents the sea surface roughness after incident angle correction, and CMOD5.N() is the CMOD5.N geophysical model function.

[0062] S3. Use the global high-resolution digital elevation model to mask the land area in the spaceborne SAR image after incident angle correction and enhance the image contrast, eliminating the influence of land dark targets on oil spill detection, especially for spaceborne SAR images obtained in coastal areas;

[0063] S4. Normalize the data of the spaceborne SAR image after land masking to avoid outliers that may dominate the results.

[0064] S32. According to the preprocessed spaceborne SAR image, adopt an oil spill detection model in the ocean to obtain the detection results of the oil spill area in the ocean.

[0065] After S103, it further includes:

[0066] Visualize the detection results, such as marking the oil spill area on the map, generating a detection report, etc., so that relevant departments can timely and accurately grasp the situation of ocean oil spills and take corresponding countermeasures.

[0067] This application combines the object detection ability of the Faster R-CNN model with the semantic segmentation advantage of the UNet++ model. The Faster R-CNN model can quickly locate the possible range of ocean oil spill areas by using its feature extraction network and region proposal network, providing a basis for subsequent precise detection. The UNet++ model, through its unique nested U-shaped structure and multi-path encoding-decoding feature fusion mechanism, can finely segment the areas located by the Faster R-CNN model, accurately depicting the boundaries, shapes, and internal detail features of the oil spill areas. For example, when detecting some areas with irregular shapes and uneven oil spill distributions, the UNet++ model can effectively utilize feature information at different levels for comprehensive judgment, avoiding segmentation errors caused by single-level features, thereby significantly improving the overall detection accuracy, and the overall detection accuracy exceeds 89%. Through the preliminary positioning accuracy of the Faster R-CNN model for the target area and the fine segmentation ability of the UNet++ model for the located area, the combined effect of the two significantly improves the detection accuracy.

[0068] Due to the adoption of the Faster R-CNN model for rapid target localization in the early stage, it can quickly screen out possible oil spill areas from the massive data of spaceborne SAR, reducing the amount of data for subsequent processing. Compared with the traditional method of analyzing each full image one by one, the detection time is greatly shortened. For example, when processing large-area ocean monitoring data, the Faster R-CNN model can quickly exclude a large number of non-oil spill areas, and then the UNet++ model only needs to perform fine processing on these small amounts of suspicious areas, avoiding unnecessary waste of computing resources and significantly improving the detection efficiency. The time to process one scene of SAR image is no more than one minute. Through the ability of the RPN network in the Faster R-CNN model to quickly generate candidate boxes and the overall optimization of the ocean oil spill detection model, the detection efficiency is improved.

[0069] In the data preprocessing stage of this application, speckle noise suppression, incidence angle correction, land masking, and data normalization operations are carried out according to the characteristics of spaceborne SAR data. And in the training stage of the ocean oil spill detection model, a suitable loss function is used to optimize the training of the model. The dual optimization of the data and the ocean oil spill detection model enables the ocean oil spill detection model to better adapt to the complexity and variability of spaceborne SAR data. Whether it is data obtained in different sea areas and different climate conditions, or various noises and geometric deformation interferences existing in the data, the model can maintain good detection performance. Through the collaborative optimization and adjustment of the data and the ocean oil spill detection model, the adaptability and robustness of the model are enhanced.

[0070] Based on the same inventive concept, the embodiment of this application also provides a spaceborne synthetic aperture radar ocean oil spill detection device for implementing the above-mentioned spaceborne synthetic aperture radar ocean oil spill detection method. The implementation solutions provided by this device to solve problems are similar to those recorded in the above method. Therefore, the specific limitations in one or more embodiments of the following spaceborne synthetic aperture radar ocean oil spill detection device can refer to the limitations on the spaceborne synthetic aperture radar ocean oil spill detection method in the above text, and will not be elaborated here.

[0071] In an exemplary embodiment, a spaceborne synthetic aperture radar ocean oil spill detection device is provided, including:

[0072] An image acquisition module, configured to acquire a spaceborne synthetic aperture radar image of an ocean oil spill area;

[0073] An ocean oil spill detection model construction module is used to construct an ocean oil spill detection model based on the Faster R-CNN model and the UNet++ model. In the ocean oil spill detection model, according to the spaceborne synthetic aperture radar image, the Faster R-CNN model is used to determine the candidate boxes containing the ocean oil spill areas; and according to the spaceborne synthetic aperture radar image corresponding to the candidate boxes, the UNet++ model is used to detect the ocean oil spill areas to obtain the detection results. The detection results include: the boundaries, shapes and internal detail features of the ocean oil spill areas.

[0074] A detection result determination module is used to obtain the detection results of the ocean oil spill areas by using the ocean oil spill detection model according to the spaceborne synthetic aperture radar images of the ocean oil spill areas.

[0075] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements a method for detecting ocean oil spills by spaceborne synthetic aperture radar.

[0076] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0077] In an exemplary embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0078] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with relevant regulations.

[0079] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0080] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0081] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0082] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for detecting marine oil spills using spaceborne synthetic aperture radar, characterized in that: The spaceborne synthetic aperture radar ocean oil spill detection method comprises: Acquire spaceborne synthetic aperture radar images of marine oil spill areas; Based on the Faster R-CNN model and the UNet++ model, a marine oil spill detection model is constructed; in the marine oil spill detection model, the Faster R-CNN model is used according to the satellite-borne synthetic aperture radar image to determine the candidate frame containing the marine oil spill area; and according to the satellite-borne synthetic aperture radar image corresponding to the candidate frame, the UNet++ model is used to detect the marine oil spill area to obtain the detection result; the detection result includes: the boundary, shape and internal detail features of the marine oil spill area; According to the satellite-borne synthetic aperture radar image of the marine oil spill area, the marine oil spill detection model is used to obtain the detection result of the marine oil spill area; The detection result of the marine oil spill area is obtained by using the marine oil spill detection model based on the satellite-borne synthetic aperture radar image of the marine oil spill area, specifically including: Performing preprocessing operations on the satellite-borne synthetic aperture radar image of the marine oil spill area; the preprocessing operations include: speckle noise suppression, incident angle correction, land masking and data normalization; According to the pre-processed spaceborne synthetic aperture radar images, the marine oil spill detection model is used to obtain the detection results of the marine oil spill area; The preprocessing operation of the satellite-borne synthetic aperture radar image of the marine oil spill area specifically includes: Using moving mean filtering to suppress speckle noise in spaceborne synthetic aperture radar images; The CMOD5.N geophysical model function is used to perform intensity scaling operation on the spaceborne synthetic aperture radar image after speckle noise suppression in the distance direction to obtain the spaceborne synthetic aperture radar image after incident angle correction. The global high-resolution digital elevation model is used to mask the land area and enhance the image contrast of the spaceborne synthetic aperture radar image after incident angle correction, so as to obtain the spaceborne synthetic aperture radar image after land masking. The data of the spaceborne synthetic aperture radar image after land masking is normalized to obtain the preprocessed spaceborne synthetic aperture radar image.

2. The method for detecting marine oil spills using spaceborne synthetic aperture radar according to claim 1, characterized in that: The marine oil spill detection model also includes: The confidence scores of candidate boxes containing the marine oil spill area are sorted using the non-maximum suppression algorithm; Starting from the candidate box with the highest confidence score, the candidate boxes whose overlapping area with the current candidate box exceeds the area threshold and whose confidence scores are less than the confidence score threshold are eliminated to obtain the eliminated candidate boxes.

3. The spaceborne synthetic aperture radar ocean oil spill detection method according to claim 1, characterized in that: The CMOD5.N geophysical model function is used to perform intensity scaling operations on the spaceborne synthetic aperture radar image after speckle noise suppression in the distance direction to obtain the spaceborne synthetic aperture radar image after incident angle correction, which specifically includes: Using the formula Perform intensity scaling operation in the distance direction on the spaceborne synthetic aperture radar image after speckle noise suppression; in, is the vertical polarization radar backscatter coefficient, represent wind speed, relative wind direction and incident angle respectively, ssr represents the sea surface roughness after incident angle correction, and CMOD5.N() is the CMOD5.N geophysical model function.

4. The method for detecting marine oil spills using spaceborne synthetic aperture radar according to claim 1, characterized in that: According to the pre-processed spaceborne synthetic aperture radar image, the marine oil spill detection model is used to obtain the detection results of the marine oil spill area, which also includes: Visualize the detection results.

5. A spaceborne synthetic aperture radar ocean oil spill detection device, used to implement the spaceborne synthetic aperture radar ocean oil spill detection method according to any one of claims 1 to 4, characterized in that: The spaceborne synthetic aperture radar ocean oil spill detection equipment comprises: An image acquisition module, used to acquire satellite-borne synthetic aperture radar images of the marine oil spill area; The marine oil spill detection model construction module is used to construct a marine oil spill detection model based on the Faster R-CNN model and the UNet++ model; in the marine oil spill detection model, the Faster R-CNN model is used to determine the candidate frame containing the marine oil spill area according to the satellite-borne synthetic aperture radar image; and the UNet++ model is used to detect the marine oil spill area according to the satellite-borne synthetic aperture radar image corresponding to the candidate frame to obtain the detection result; the detection result includes: the boundary, shape and internal detail features of the marine oil spill area; The detection result determination module is used to obtain the detection result of the marine oil spill area based on the satellite-borne synthetic aperture radar image of the marine oil spill area and the marine oil spill detection model.

6. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the spaceborne synthetic aperture radar marine oil spill detection method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the spaceborne synthetic aperture radar ocean oil spill detection method described in any one of claims 1 to 4 is implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the spaceborne synthetic aperture radar ocean oil spill detection method described in any one of claims 1 to 4 is implemented.

Images