Method for detecting mesoscale eddies based on neural network and sea surface height anomaly data

Abstract

The application discloses a mesoscale vortex detection method based on a neural network and sea surface height anomaly data. The method comprises the following steps: acquiring sea surface height data of a preset area; extracting local extreme values in sea surface height anomaly data from the sea surface height data; performing mesoscale vortex detection on the local extreme values by using an isogram method to obtain a first mesoscale vortex detection result; inputting the local extreme values into a mesoscale vortex detection model to obtain a second mesoscale vortex detection result; the mesoscale vortex detection model is obtained based on neural network training; and obtaining a mesoscale vortex detection result of the preset area according to the first mesoscale vortex detection result and the second mesoscale vortex detection result. The application comprehensively uses the two mesoscale vortex detection results, and improves the accuracy of mesoscale vortex detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of mesoscale eddy detection, and in particular to a mesoscale eddy detection method based on neural network and sea surface height anomaly data. BACKGROUND

[0002] Mesoscale eddies in the ocean are an important natural phenomenon, usually with diameters ranging from tens to hundreds of kilometers, formed by complex interactions of fluid dynamics. These eddies not only have important effects on the transport of heat, momentum and matter in the ocean, but also can regulate marine biological productivity and ecosystem stability, and are crucial for understanding the mechanisms of global climate change.

[0003] Accurate detection and identification of mesoscale eddies in the ocean is one of the key tasks of ocean science research. Existing mesoscale eddy detection techniques mainly rely on satellite remote sensing, numerical simulation and field observation, but these methods have problems such as being susceptible to noise, insufficient feature extraction, slow processing speed and sensitivity to outliers, which affect the accuracy of mesoscale eddy detection. SUMMARY

[0004] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.

[0005] The main purpose of the embodiments of the present disclosure is to propose a mesoscale eddy detection method based on neural network and sea surface height anomaly data, which can improve the accuracy of mesoscale eddy detection.

[0006] The first aspect of the embodiments of the present application provides a mesoscale eddy detection method based on neural network and sea surface height anomaly data, the method comprising:

[0007] obtaining sea surface height data of a preset area;

[0008] extracting local extreme values in the sea surface height anomaly data from the sea surface height data;

[0009] detecting mesoscale eddies by contour method on the local extreme values to obtain a first mesoscale eddy detection result;

[0010] inputting the local extreme values into a mesoscale eddy detection model to obtain a second mesoscale eddy detection result; the mesoscale eddy detection model is obtained based on neural network training;

[0011] obtaining a mesoscale eddy detection result of the preset area according to the first mesoscale eddy detection result and the second mesoscale eddy detection result.

[0012] The embodiment of the present application provides a mesoscale vortex detection method based on a neural network and sea surface height anomaly data, local extreme values of sea surface anomaly data in sea surface height data of a preset area are extracted, mesoscale vortex detection is performed according to the local extreme values through an isogram method, a first mesoscale vortex detection result is obtained, mesoscale vortex detection is performed again through a pre-trained mesoscale vortex detection model, a second mesoscale vortex detection result is obtained, and finally, a mesoscale vortex detection result of the preset area is obtained according to the first mesoscale vortex detection result and the second mesoscale vortex detection result. It can be seen that the present application combines the mesoscale vortex detection results twice, and the accuracy of mesoscale vortex detection is improved.

[0013] In some embodiments of the present application, before the local extreme values of the sea surface height anomaly data are extracted from the sea surface height data, the method further comprises:

[0014] According to the meteorological environment data of the preset area, the sea surface height data is corrected to obtain the corrected sea surface height data;

[0015] The local extreme values of the sea surface height anomaly data are extracted from the sea surface height data, comprising:

[0016] The sea surface height anomaly data is extracted from the sea surface height data;

[0017] The local extreme values are determined in the sea surface height anomaly data;

[0018] Wherein, the sea surface height anomaly data is calculated by the following formula:

[0019] SLA(x,y)=H obs (x,y)-H mean (x,y)

[0020] Wherein, SLA(x,y) is sea surface height anomaly data, H obs (x,y) is the corrected sea surface height data, H mean (x,y) is the historical average sea surface height data, x and y represent longitude and latitude.

[0021] In some embodiments of the present application, the local extreme values are determined in the sea surface height anomaly data, comprising:

[0022] The sea surface height anomaly data is divided into a plurality of local windows through a sliding window method;

[0023] The local extreme values of the sea surface height anomaly data are determined in each local window, and the local extreme values include local minimum values and local maximum values.

[0024] In some embodiments of the present application, the mesoscale eddy detection on the local extreme value by the contour method obtains a first mesoscale eddy detection result, comprising:

[0025] contour lines of the local minimum value and the local maximum value are drawn by the contour method;

[0026] whether it is a mesoscale eddy region is determined according to the contour lines;

[0027] if it is the mesoscale eddy region, a specific type of the mesoscale eddy region is determined according to sea surface height anomaly data in the contour lines.

[0028] In some embodiments of the present application, the training process of the mesoscale eddy detection model comprises:

[0029] historical sea surface height anomaly data and historical remote sensing data of the preset region are obtained;

[0030] labeled mesoscale eddy instances are extracted from the historical sea surface height anomaly data and the historical remote sensing data;

[0031] a detection model based on a neural network is constructed;

[0032] the detection model is trained according to the labeled mesoscale eddy instances to obtain the trained mesoscale eddy detection model.

[0033] In some embodiments of the present application, the detection model is a ResNet network model.

[0034] In some embodiments of the present application, the mesoscale eddy detection result of the preset region is obtained according to the first mesoscale eddy detection result and the second mesoscale eddy detection result, comprising:

[0035] the first mesoscale eddy detection result and the second mesoscale eddy detection result are weighted and fused by the following formula to obtain the mesoscale eddy detection result of the preset region:

[0036] P final (x,y)=α·P CNN (x,y)+(1-α)·P SLA (x,y)

[0037] wherein, P final is the mesoscale eddy detection result of the prediction region, α is a fusion weight parameter, P CNN is the second mesoscale eddy detection result, P SLA is the first mesoscale eddy detection result, and (x,y) is the latitude and longitude of the preset region.

[0038] A second aspect of the embodiment of the present application provides a mesoscale vortex detection device based on a neural network and sea surface height anomaly data, the device comprising:

[0039] a data acquisition module configured to acquire sea surface height data of a preset area;

[0040] a data extraction module configured to extract local extreme values in sea surface height anomaly data from the sea surface height data;

[0041] a first detection module configured to perform mesoscale vortex detection on the local extreme values by using an isogram method to obtain a first mesoscale vortex detection result;

[0042] a second detection module configured to input the local extreme values into a mesoscale vortex detection model to obtain a second mesoscale vortex detection result; the mesoscale vortex detection model is obtained based on neural network training;

[0043] a vortex detection module configured to obtain a mesoscale vortex detection result of the preset area according to the first mesoscale vortex detection result and the second mesoscale vortex detection result.

[0044] A third aspect of the embodiment of the present application provides an electronic device, comprising: at least one control processor and a memory connected in communication with the at least one control processor; the memory stores instructions executable by the at least one control processor, and the instructions are executed by the at least one control processor to enable the at least one control processor to perform the above-mentioned mesoscale vortex detection method based on a neural network and sea surface height anomaly data.

[0045] A fourth aspect of the embodiment of the present application provides a computer readable storage medium, the computer readable storage medium stores computer executable instructions, and the computer executable instructions are used to make a computer execute the above-mentioned mesoscale vortex detection method based on a neural network and sea surface height anomaly data.

[0046] It can be understood that the beneficial effects of the above-mentioned second aspect to fourth aspect and related technologies are the same as the beneficial effects of the above-mentioned first aspect and related technologies, which can be referred to the related description in the first aspect and will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0047] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description, including the accompanying drawings, in which:

[0048] Figure 1 is a flowchart of a mesoscale vortex detection method based on a neural network and sea surface height anomaly data provided by the embodiment of the present application.

[0049] Figure 2 is a sea surface height anomaly data vortex map of a mesoscale vortex detection method based on a neural network and sea surface height anomaly data provided by an embodiment of the present application;

[0050] Figure 3 is a schematic diagram of a mesoscale vortex detection method based on a neural network and sea surface height anomaly data provided by an embodiment of the present application; Figure 2 a result schematic diagram of mesoscale vortex detection;

[0051] Figure 4 is a structural schematic diagram of a mesoscale vortex detection device provided by an embodiment of the present application;

[0052] Figure 5 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0053] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as a limitation of the present application.

[0054] In the description of the present application, if the first, second, etc. are described, it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features or the sequence of the indicated technical features.

[0055] In the description of the present application, it should be understood that the orientation description, such as the orientation or position relationship indicated by up, down, etc. is based on the orientation or position relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as indicating or implying that the device or element indicated must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present application.

[0056] In the description of the present application, it should be noted that, unless otherwise explicitly limited, the words such as setting, installing, connecting, etc. should be broadly understood, and the person skilled in the art can reasonably determine the specific meaning of the above words in the present application in combination with the specific content of the technical solution.

[0057] Oceanic mesoscale eddies are an important phenomenon in the ocean, typically with diameters ranging from tens to hundreds of kilometers, formed from complex interactions of fluid dynamics, which can be driven by factors such as climate change, oceanic circulation, and topographic influences. Common forms include cyclonic and anticyclonic eddies, which not only have an important impact on the transport of heat, momentum, and matter in the ocean, but also regulate the biological productivity and stability of the ecosystem. Their existence is crucial for understanding the mechanisms of global climate change, especially in climate models and weather forecasting.

[0058] Accurate detection and identification of mesoscale eddies in the ocean is an important topic in ocean science research. The existence and changes of mesoscale eddies directly affect the circulation patterns, heat transport, material mixing, and biological productivity of the ocean. By monitoring and identifying mesoscale eddies in real time, scientists can better understand oceanic dynamic processes, predict trends in oceanic environmental changes, and provide scientific basis for ocean resource development and protection. Meanwhile, the detection data of mesoscale eddies are important references for assessing global climate change, marine disaster warning, and fishery resource distribution. In addition, mesoscale eddies can affect the distribution of nutrients in the water, thereby affecting the habitat and habitat of marine organisms. Therefore, by accurately identifying and monitoring mesoscale eddies, decision-makers can develop more scientific ocean management strategies.

[0059] Existing mesoscale eddy detection and identification techniques mainly rely on satellite remote sensing, numerical simulation, and field observation methods. Satellite remote sensing can efficiently obtain ocean information over a large area using data such as sea surface temperature, salinity, and sea level anomaly. With the development of earth observation technology, Sea Level Anomaly (SLA) data obtained from satellite altimeters is currently a common solution for extracting mesoscale eddies in the ocean. However, due to factors such as satellite orbit parameter deviations, sensor system errors, and atmospheric conditions, SLA data is prone to noise information. The generation of SLA data also requires the use of statistical parameters over many years, which affects the accuracy of extracting mesoscale eddies. Furthermore, the Rossby wave deformation radius scale caused by Earth's rotation and changes in latitude differences is similar to the scale of mesoscale eddies, further interfering with the identification and extraction of mesoscale eddies.

[0060] Mesoscale eddies are divided into two types: cyclonic eddies (counterclockwise rotation in the Northern Hemisphere, also known as "cold eddies") and anticyclonic eddies (counterclockwise rotation in the Southern Hemisphere, also known as "warm eddies"). Cyclonic eddies exhibit negative extreme values of height anomaly, while anticyclonic eddies exhibit positive extreme values of height anomaly. The most intuitive representation of mesoscale eddies is the closed sea surface height anomaly contour, where the SLA maximum point corresponds to an anticyclonic eddy, and the SLA minimum point corresponds to a cyclonic eddy.

[0061] Therefore, with reference to Figure 1The embodiment of the application provides a mesoscale vortex detection method based on a neural network and sea surface height anomaly data, the method is applied to a controller, the controller can be a server, can be an electronic device, can also be a mobile terminal, and the like, and here is not specifically limited, and the method comprises the following steps S101 to S105.

[0062] In step S101, sea surface height data of a preset area is acquired.

[0063] In this step, the preset area is a marine area to be subjected to mesoscale vortex detection, and the sea surface height data is acquired by satellite remote sensing technology, specifically, the sea surface height is measured by a satellite radar altimeter.

[0064] In step S102, local extreme values in sea surface height anomaly data are extracted from the sea surface height data.

[0065] In this step, meteorological and environmental data of the preset area are acquired by satellite remote sensing technology, the sea surface height data is corrected according to the meteorological and environmental data, the sea surface height anomaly data (SLA data) is calculated based on the corrected sea surface height data, and the local extreme values of the sea surface height anomaly data are determined by a sliding window technology, so as to eliminate noise data interference in the sea surface height anomaly data, wherein the local extreme values include local minimum values and local maximum values.

[0066] In step S103, the local extreme values are subjected to mesoscale vortex detection by an isogram method, and a first mesoscale vortex detection result is obtained.

[0067] In this step, the isogram method is a method for extracting lines and surfaces with the same value from a two-dimensional or three-dimensional data field, in the embodiment, an isogram chart is drawn around each local minimum value and local maximum value by the isogram method, the isogram closed area is regarded as a possible mesoscale vortex area, the sea surface height anomaly data in the mesoscale vortex area is extracted, the nature of the mesoscale vortex area is further confirmed in combination with the characteristics of the mesoscale vortex, and the first mesoscale vortex detection result is obtained.

[0068] In step S104, the local extreme values are input into a mesoscale vortex detection model, and a second mesoscale vortex detection result is obtained, wherein the mesoscale vortex detection model is obtained based on neural network training.

[0069] In this step, based on the neural network model, the mesoscale vortex detection model is pre-trained, the deep features are automatically learned by the mesoscale vortex detection model, and the accuracy of marine image recognition and data processing can be improved by the mesoscale vortex detection model, when the local extreme values of the preset area are acquired, the local extreme values are input into the mesoscale vortex detection model, and the second mesoscale vortex detection result is obtained.

[0070] Step S105, obtaining the mesoscale vortex detection result of the preset area according to the first mesoscale vortex detection result and the second mesoscale vortex detection result.

[0071] In this step, the first mesoscale vortex detection result and the second mesoscale vortex detection result are weighted and fused, and according to the result of the weighted and fused, the mesoscale vortex detection result of the preset area is obtained, and by comprehensively combining the two mesoscale vortex detection results, the accuracy of the mesoscale vortex detection is further improved.

[0072] Unlike the prior art, the embodiment of the application extracts the local extreme value of the sea surface abnormal data in the sea surface height data of the preset area, performs mesoscale vortex detection according to the local extreme value by the contour method to obtain the first mesoscale vortex detection result, and performs mesoscale vortex detection again by the pre-trained mesoscale vortex detection model to obtain the second mesoscale vortex detection result, and finally obtains the mesoscale vortex detection result of the preset area according to the first mesoscale vortex detection result and the second mesoscale vortex detection result. It can be seen that the embodiment of the application comprehensively combines the two mesoscale vortex detection results, and improves the accuracy of the mesoscale vortex detection.

[0073] The specific implementation of each of the above steps will be introduced below.

[0074] Before the step S102 of extracting the local extreme value of the sea surface abnormal data from the sea surface height data, the method further includes the following step S201.

[0075] Step S201, correcting the sea surface height data according to the meteorological environment data of the preset area to obtain the corrected sea surface height data.

[0076] In this embodiment, satellite remote sensing data is obtained by satellite remote sensing technology, and the satellite remote sensing data includes sea surface height data, meteorological and environmental data; wherein the meteorological and environmental data includes but is not limited to atmospheric pressure data, wind speed data, temperature data; the influence of factors such as swells and tidal currents on the sea surface height is corrected by ocean models such as HYCOM (Hybrid Coordinate Ocean Model) and NEMO (Nucleus for European Modelling of the Ocean), and then the meteorological and environmental data in the satellite data and other ground observation data (such as buoy data and tide station data) are assimilated to enhance the accuracy of the sea surface height data.

[0077] Wherein, the step S102 of extracting the local extreme value of the sea surface abnormal data from the sea surface height data includes the following steps S202 to S203.

[0078] Step S202, extracting sea level anomaly data from the sea level height data.

[0079] Step S203, determining local extreme values in the sea level anomaly data.

[0080] The sea level anomaly data is calculated by the following formula:

[0081] SLA(x, y) = H obs (x, y) - H mean (x, y)

[0082] Wherein, SLA(x, y) is sea level anomaly data, H obs (x, y) is the corrected sea level height data, H mean (x, y) is the historical average sea level height data, x, y represent longitude and latitude.

[0083] In this embodiment, the sea level anomaly data of the preset area is calculated by the corrected sea level height data and the historical average sea level height data.

[0084] Step S203 in the sea level anomaly data includes the following steps S301 to S302.

[0085] Step S301, dividing the sea level anomaly data into multiple local windows by the sliding window method.

[0086] Step S302, determining the local extreme values of the sea level anomaly data in each local window, including local minimum and local maximum.

[0087] In this embodiment, the cyclonic vortex appears as a negative extreme value region of sea level anomaly value, and the anticyclonic vortex appears as a positive extreme value region of sea level anomaly value; the extreme value characteristics of the sea level anomaly value can effectively identify these vortex regions in the SLA data, which is conducive to improving the accuracy of the preliminary determination of the mesoscale vortex region in the ocean.

[0088] In this embodiment, the sea level anomaly data obtained by the above calculation is divided into multiple local windows by the sliding window method, and the local minimum and local maximum of the sea level anomaly data are found in each local window. According to the characteristics of cyclonic vortex and anticyclonic vortex, the cyclonic vortex will appear in the local minimum value region, and the anticyclonic vortex will appear in the local maximum value region. The local minimum and local maximum points are compared with the values in the surrounding field, and the vortex properties of the preset area are preliminarily judged, so as to eliminate the influence of noise data in the sea level anomaly data and improve the accuracy of the subsequent mesoscale vortex detection results in the preset area.

[0089] In the step S202, the local extreme value in the sea surface height anomaly data is extracted from the sea surface height data, including the following steps S401 to S403.

[0090] In the step S401, the contour map of the local minimum value and the local maximum value is drawn by the contour method.

[0091] In the step S402, it is determined whether it is a mesoscale vortex area according to the contour map.

[0092] In the step S403, if it is a mesoscale vortex area, the specific type of the mesoscale vortex area is determined according to the sea surface height anomaly data in the contour map.

[0093] In the embodiment, for each detected local minimum value and local maximum value, the vortex area is further defined by the contour method, that is, the contour map is drawn around each local minimum value and local maximum value by the contour method, the closed area of the contour is regarded as a possible mesoscale vortex area, the sea surface height anomaly data in the mesoscale vortex area is extracted, and the nature of the mesoscale vortex area is further confirmed in combination with the characteristics (positive or negative extreme value) of the mesoscale vortex to obtain the first mesoscale vortex detection result.

[0094] In some embodiments, the minimum value point is taken as the vortex center, the maximum value point is taken as the antivortex center, after detecting the vortex center position, the SLA contour method is used to detect the vortex boundary. With a step of 0.5 cm, the SLA contour is searched from the vortex center position outward; since the measurement error of the altimeter is generally 2-3 cm, the SLA contour with a difference greater than 3 cm from the vortex center position and closed is determined as the boundary of the vortex. The height anomaly value in the SLA data is used to detect the potential mesoscale vortex area.

[0095] In the training process of the mesoscale vortex detection model in the step S104, the following steps S501 to S504 are included.

[0096] In the step S501, the historical sea surface height anomaly data and the historical remote sensing data of a preset area are obtained.

[0097] In the step S502, the labeled mesoscale vortex instances are extracted from the historical sea surface height anomaly data and the historical remote sensing data.

[0098] In the step S503, the detection model based on the neural network is constructed.

[0099] In the step S504, the detection model is trained according to the labeled mesoscale vortex instances to obtain the trained mesoscale vortex detection model.

[0100] In the embodiment, the historical remote sensing image data comes from optical or multispectral sensors that complement the sea surface height data, which record information such as the temperature, color, wind field, etc. of the ocean surface. In this embodiment, after obtaining the high-resolution historical remote sensing image of the preset area, the historical remote sensing image is standardized to adapt to the input requirements of the neural network.

[0101] In some embodiments, the standardization of the historical remote sensing image specifically refers to scaling the pixel values of the historical remote sensing image to between 0 and 1 through normalization processing, or adjusting it to a zero-mean unit variance form (i.e. subtracting the mean and then dividing by the standard deviation) to eliminate the amplitude, brightness and contrast differences between different historical remote sensing images, so as to improve the training effect of the neural network. And in order to adapt to the fixed size of the neural network input, the historical remote sensing image is cropped or resampled into a preset image size, and the preset image size is set to 224x224 pixels. It should be noted that the preset image size can also be set according to specific circumstances.

[0102] In the embodiment, the labeled mesoscale vortex instances are extracted from the historical sea surface height anomaly data and the historical remote sensing data, and the labeled mesoscale vortex instances are used to construct a mesoscale vortex labeled sample set. Seventy percent of the mesoscale vortex labeled sample set is used as the training data set of the mesoscale vortex model, and seventy percent of the mesoscale vortex labeled sample set is used as the test data set of the mesoscale vortex model.

[0103] In the embodiment, the detection model is a ResNet neural network model. ResNet (Residual Network) is a kind of deep convolutional neural network (CNN). The biggest feature of the ResNet network model is to introduce the concept of residual learning (Residual Learning). Through the use of skip connection (or called identity mapping), the gradient vanishing problem in the training process of the deep neural network is effectively solved, so that the network can reach an unprecedented depth.

[0104] In the embodiment, the ResNet network model mainly includes a convolutional layer (which extracts features from the image and generates a low-dimensional space representation), a residual layer (which adds the input directly to the output through a skip connection, so that the network can learn the residual between the input and the output), and a fully connected layer. The ResNet network model can be represented by the following formula:

[0105]

[0106] where y represents the output of the ResNet network model, represents the residual function, x is the input, and {W i} is the trainable parameter in the ResNet network model.

[0107] In the embodiment, the ResNet model is trained by a training data set, the deviation of the prediction result from the true result is measured by a cross-entropy loss function, and the model parameters of the ResNet network model are adjusted by an Adam algorithm combined with momentum and adaptive learning rate adjustment to accelerate the convergence of the model and prevent gradient shock.

[0108] In the embodiment, the ResNet network model is trained by a training data set for multiple rounds, and the model parameters of the ResNet network model are adjusted by a back propagation algorithm in each round of training to minimize the loss value, so as to obtain a trained mesoscale vortex detection model; the performance of the mesoscale vortex detection model is evaluated by a test data set, the accuracy of the mesoscale detection model is calculated, and if the accuracy is greater than or equal to a preset accuracy, the trained mesoscale vortex detection model is applied to mesoscale vortex detection, and if the accuracy is less than the preset accuracy, the ResNet network model is retrained until the accuracy meets the requirements. The preset accuracy can be set according to the actual situation, or can be set to 90%.

[0109]

[0110] wherein L is the minimized loss function value, N is the sample quantity, y i is the true label, is the prediction probability.

[0111] In the embodiment, the convolutional neural network (ResNet network) is combined to make up for the shortcomings of traditional algorithms in feature extraction, so that the mesoscale vortex detection model can automatically learn deep features, thereby improving the detection accuracy of mesoscale vortexes. The robustness to complex data and noise is improved by the deep learning method, the dependence on manual feature design of traditional methods is reduced, the mesoscale vortex detection model can still maintain good detection ability when facing various environmental changes, the structure and algorithm of the ResNet neural network are optimized to improve the calculation efficiency, so that large-scale SLA data can be processed, and real-time or near-real-time mesoscale vortex detection can be realized.

[0112] In step S105, the mesoscale vortex detection result of the preset area is obtained according to the first mesoscale vortex detection result and the second mesoscale vortex detection result, including the following step S601.

[0113] In step S601, the first mesoscale vortex detection result and the second mesoscale vortex detection result are weighted and fused by the following formula to obtain the mesoscale vortex detection result of the preset area:

[0114] P final (x,y)=α·PCNN (x,y)+(1-α)·P SLA (x,y)

[0115] Among them, P final is the mesoscale eddy detection result in the prediction area, α is the fusion weight parameter, P CNN is the second mesoscale vortex detection result, P SLA is the first mesoscale vortex detection result, and (x, y) is the longitude and latitude of the preset area.

[0116] In this embodiment, α can be set according to specific circumstances. This embodiment achieves more accurate and automated mesoscale vortex identification by combining the detection results based on sea surface height anomaly data and the detection results based on the mesoscale vortex detection model.

[0117] In this embodiment, Figure 2 is the vortex map of the sea surface height anomaly data, Figure 3 Based on Figure 2 The mesoscale vortex detection result diagram is obtained by using the technical solution provided by the embodiment of the present application to obtain the abnormal sea surface height data. It can be seen that the technical solution of the embodiment of the present application can accurately detect the mesoscale vortex.

[0118] In this embodiment, local extreme values ​​of sea surface anomaly data are extracted from sea surface height data for a preset region. Based on these local extreme values, mesoscale vortex detection is performed using the contour line method to obtain a first mesoscale vortex detection result. A pre-trained mesoscale vortex detection model is then used to perform a second mesoscale vortex detection result. Finally, based on the first and second mesoscale vortex detection results, a mesoscale vortex detection result for the preset region is obtained. By combining these two mesoscale vortex detection results, the accuracy of mesoscale vortex detection is improved.

[0119] like Figure 3 As shown, some embodiments of the present application provide a mesoscale vortex detection device based on a neural network and sea level anomaly data, the device comprising a data acquisition module 1100, a data extraction module 1200, a first detection module 1300, a second detection module 1400, and a vortex detection module 1500. Specifically:

[0120] The data acquisition module 1100 is used to obtain sea level height data of a preset area;

[0121] A data extraction module 1200 is used to extract local extreme values ​​in the sea level height anomaly data from the sea level height data;

[0122] The first detection module 1300 is configured to detect mesoscale eddies by using the contour method to obtain a first mesoscale eddy detection result.

[0123] The second detection module 1400 is configured to input the local extreme value into a mesoscale eddy detection model to obtain a second mesoscale eddy detection result.

[0124] The eddy detection module 1500 is configured to obtain a mesoscale eddy detection result of the preset area according to the first mesoscale eddy detection result and the second mesoscale eddy detection result.

[0125] It should be noted that the mesoscale eddy detection device based on the neural network and the sea surface height anomaly data provided in the embodiment and the mesoscale eddy detection method based on the neural network and the sea surface height anomaly data described above are based on the same inventive concept, and therefore the related content of the mesoscale eddy detection method based on the neural network and the sea surface height anomaly data described above is also applicable to the content of the mesoscale eddy detection device based on the neural network and the sea surface height anomaly data, and therefore, the content will not be described here.

[0126] As Figure 4 The embodiment of the present application further provides an electronic device, and the electronic device comprises:

[0127] At least one hydrogen fuel cell;

[0128] At least one memory;

[0129] At least one processor;

[0130] At least one program;

[0131] The program is stored in the memory, and the processor executes the at least one program to implement the mesoscale eddy detection method based on the neural network and the sea surface height anomaly data described above.

[0132] The electronic device can be any intelligent terminal, such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a vehicle-mounted computer, etc.

[0133] The electronic device of the embodiment of the present application will be described in detail below.

[0134] The processor 1600 can be implemented by a general-purpose central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), or one or more integrated circuits, and is configured to execute related programs to implement the technical solutions provided by the embodiments of the present disclosure.

[0135] The memory 1700 can be implemented by a read only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1700 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present disclosure are implemented by software or firmware, the related program codes are stored in the memory 1700 and are called and executed by the processor 1600 to implement a neural network-based mesoscale vortex detection method based on sea surface height anomaly data.

[0136] The input / output interface 1800 is configured to implement information input and output.

[0137] The communication interface 1900 is configured to implement the communication interaction between the device and other devices. The communication can be implemented by a wired manner (for example, a USB, a network cable, etc.) or a wireless manner (for example, a mobile network, WIFI, Bluetooth, etc.).

[0138] The bus 2000 is configured to transmit information between various components (for example, the processor 1600, the memory 1700, the input / output interface 1800, and the communication interface 1900) of the device.

[0139] The processor 1600, the memory 1700, the input / output interface 1800, and the communication interface 1900 are connected to each other by the bus 2000 to realize the communication connection between them in the device.

[0140] The embodiments of the present disclosure further provide a storage medium, which is a computer readable storage medium and stores computer executable instructions for causing a computer to execute the neural network-based mesoscale vortex detection method based on sea surface height anomaly data.

[0141] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include a high-speed random access memory and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory that is remotely disposed relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0142] The embodiments described in the embodiments of the present disclosure are used to more clearly illustrate the technical solutions of the embodiments of the present disclosure, and do not constitute a limitation on the technical solutions provided by the embodiments of the present disclosure. Those skilled in the art can know that, as technology evolves and new application scenarios appear, the technical solutions provided by the embodiments of the present disclosure are also applicable to similar technical problems.

[0143] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present disclosure, and can include more or fewer steps than the figures shown, or combine certain steps, or different steps.

[0144] The device embodiments described above are only schematic, and units described as separate components can or can not be physically separate, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments.

[0145] Those skilled in the art can understand that all or some of the steps in the above disclosed method, the functional modules / units in the system and the device can be implemented as software, firmware, hardware and their appropriate combinations.

[0146] The terms "first", "second", "third", "fourth" and the like used in the specification of the present application and the above-described drawings (if any) are used to distinguish similar objects, and do not necessarily have to describe a particular order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0147] It should be understood that, in the application, "at least one" refers to one or more, and "multiple" refers to two or more. "And / or" is used to describe the association relationship of the associated objects, which means that there can be three relationships, for example, "A and / or B" can represent three cases of only A, only B and A and B existing at the same time, wherein A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after it. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can represent a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b and c can be single or multiple.

[0148] In several embodiments provided in the application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed mutual units can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0149] The units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on multiple network units. According to actual needs, some or all of the units can be selected to achieve the purpose of the embodiment scheme.

[0150] In addition, the functional units in each embodiment of the application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0151] If the integrated unit is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes multiple instructions for causing an electronic device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, and various program storage media.

[0152] The above is a specific description of the preferred implementation of the embodiments of the present application, but the embodiments of the present application are not limited to the above-mentioned implementation. Those skilled in the art can make various equivalent modifications or replacements without departing from the spirit of the embodiments of the present application, and these equivalent modifications or replacements are all included in the scope defined by the claims of the embodiments of the present application.

[0153] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited to the above-mentioned embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the purpose of the present application.

Claims

1. A mesoscale vortex detection method based on neural networks and sea surface height anomaly data, characterized in that, The method comprises: acquiring sea surface height data of a preset area; extracting local extreme values in sea surface height anomaly data from the sea surface height data; detecting mesoscale eddies from the local extreme values by means of contour lines to obtain a first mesoscale eddy detection result; inputting the local extreme values into a mesoscale eddy detection model to obtain a second mesoscale eddy detection result; the mesoscale eddy detection model is obtained by training based on a neural network; obtaining a mesoscale eddy detection result of the preset area according to the first mesoscale eddy detection result and the second mesoscale eddy detection result; the mesoscale eddy detection from the local extreme values by means of contour lines to obtain the first mesoscale eddy detection result comprises: drawing contour lines of local minimum values and local maximum values by means of contour lines; determining whether it is a mesoscale eddy area according to the contour lines; if it is the mesoscale eddy area, determining a specific type of the mesoscale eddy area according to sea surface height anomaly data in the contour lines; the mesoscale eddy detection result of the preset area is obtained according to the first mesoscale eddy detection result and the second mesoscale eddy detection result, comprising: performing weighted fusion on the first mesoscale eddy detection result and the second mesoscale eddy detection result by means of the following formula to obtain the mesoscale eddy detection result of the preset area: P final (x,y) = a - P CNN (x,y) + (1 - a) - P SLA (x,y) wherein P final is a mesoscale vortex detection result of the prediction region, a is a fusion weight parameter, P CNN is a second mesoscale vortex detection result, P SLA is a first mesoscale vortex detection result, and (x, y) is the latitude and longitude of the preset region.

2. The method of claim 1, wherein the method is based on neural networks and sea surface height anomaly data. before the local extreme values in the sea surface height anomaly data are extracted from the sea surface height data, the method further comprises: correcting the sea surface height data according to meteorological environment data of the preset area to obtain corrected sea surface height data; the local extreme values in the sea surface height anomaly data are extracted from the sea surface height data, comprising: extracting sea surface height anomaly data from the sea surface height data; determining local extreme values in the sea surface height anomaly data; wherein the sea surface height anomaly data is calculated by means of the following formula: SLA(x, y) = H obs (x, y) - H mean (x, y) wherein SLA(x, y) is the sea level anomaly data, H obs (x, y) is the corrected sea level data, H mean (x, y) is the historical average sea level data, x, y represent longitude and latitude.

3. The method of claim 2, wherein the method is based on neural networks and sea surface height anomaly data. the local extreme values in the sea surface height anomaly data are determined, comprising: dividing the sea surface height anomaly data into a plurality of local windows by means of a sliding window method; determining local extreme values of the sea surface height anomaly data in each local window, the local extreme values comprising local minimum values and local maximum values.

4. The method of claim 1, wherein the method is based on a neural network and sea surface height anomaly data. the training process of the mesoscale eddy detection model comprises: acquiring historical sea surface height anomaly data and historical remote sensing data of the preset area; extracting labeled mesoscale eddy instances from the historical sea surface height anomaly data and the historical remote sensing data; constructing a detection model based on a neural network; training the detection model according to the labeled mesoscale eddy instances to obtain the trained mesoscale eddy detection model.

5. The method of claim 4, wherein the neural network is trained using a dataset of mesoscale eddies and a dataset of sea level anomaly data. The detection model is a ResNet network model.

6. A mesoscale eddy detection apparatus based on neural networks and sea surface height anomaly data, characterized by, The device comprises: a data acquisition module for acquiring sea surface height data of a preset area; a data extraction module for extracting local extreme values in sea surface height anomaly data from the sea surface height data; a first detection module for detecting mesoscale eddies from the local extreme values by means of contour lines to obtain a first mesoscale eddy detection result; The first detection module is further configured to draw an isogram of the local minimum value and the local maximum value by using an isogram method; determine whether the area is a mesoscale vortex area according to the isogram; if the area is the mesoscale vortex area, determine a specific type of the mesoscale vortex area according to sea surface height anomaly data in the isogram; The second detection module is configured to input the local extreme value into a mesoscale vortex detection model to obtain a second mesoscale vortex detection result, wherein the mesoscale vortex detection model is obtained based on neural network training. The vortex detection module is configured to obtain the mesoscale vortex detection result of the preset area according to the first mesoscale vortex detection result and the second mesoscale vortex detection result. The vortex detection module is further configured to perform weighted fusion on the first mesoscale vortex detection result and the second mesoscale vortex detection result according to the following formula to obtain the mesoscale vortex detection result of the preset area: P final (x,y) = a - P CNN (x,y) + (1 - a) - P SLA (x,y) wherein P final is a mesoscale vortex detection result of the prediction region, a is a fusion weight parameter, P CNN is a second mesoscale vortex detection result, P SLA is a first mesoscale vortex detection result, and (x, y) is the latitude and longitude of the preset region.

7. An electronic device, comprising: The at least one control processor and a memory connected in communication with the at least one control processor; the memory stores instructions executable by the at least one control processor, and the instructions are executed by the at least one control processor to enable the at least one control processor to perform the mesoscale vortex detection method based on neural network and sea surface height anomaly data according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer readable storage medium stores computer executable instructions for causing a computer to perform the mesoscale vortex detection method based on neural network and sea surface height anomaly data according to any one of claims 1 to 5.

Images