Ionized layer super-resolution refining reconstruction method and device based on improved ViT model, equipment and storage medium

Abstract

The invention discloses an ionosphere super-resolution refining reconstruction method, device and equipment based on an improved ViT model and a storage medium, and relates to the technical field of ionosphere modeling, and the method comprises the steps: obtaining the TEC data and VTEC data of an ionosphere in a to-be-detected time period; taking the ionized layer TEC data as a low-resolution image, taking the VTEC data as a pixel true value, and generating a mask layer according to whether the pixel true value comprises an observation value or not; combining the low-resolution image, the pixel true value and the mask layer to obtain a multi-channel tensor; and taking the multi-channel tensor as input, and obtaining a global ionosphere map based on output of the improved ViT model. ViT is applied to super-resolution reconstruction of the ionosphere, dependence on the number of ground sites can be effectively reduced, the defect that the image is excessively smooth is overcome, meanwhile, the spatial resolution of the ionosphere image can be improved, and then the precision of an ionosphere model is improved.

Description

Technical Field

[0001] The present invention relates to a method, device, equipment and storage medium for ionospheric super-resolution refinement reconstruction based on an improved ViT model, and belongs to the technical field of ionospheric modeling. Background Art

[0002] Traditional modeling of ionospheric TEC usually relies on satellite observation data of ground GNSS stations and is modeled based on mathematical models (such as spherical harmonic functions). This often depends on a large number of ground stations, and modeling based on mathematical functions (such as spherical harmonic functions) also faces the drawback of over-smoothing.

[0003] With the development of artificial intelligence, in computer image processing, super-resolution is a technology that converts low-resolution blurred images into high-resolution clear images. Introducing super-resolution technology into ionospheric model construction is beneficial to overcoming the drawbacks of traditional models that rely on a large number of ground stations and over-smoothing of the models. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method for ionospheric super-resolution refinement reconstruction using an improved ViT model. Applying the ViT to the super-resolution reconstruction of the ionosphere can effectively reduce the dependence on the number of ground stations, overcome the drawback of over-smoothing of images, and at the same time improve the spatial resolution of ionospheric images, thereby improving the accuracy of ionospheric models.

[0005] To achieve the above object, the present invention is implemented by the following technical solutions: In the first aspect, the present invention provides a method for ionospheric super-resolution refinement reconstruction based on an improved ViT model, including: Obtaining ionospheric TEC data and VTEC data within a period to be measured; Taking the ionospheric TEC data as a low-resolution image, the VTEC data as the true value of pixels, and generating a mask layer according to whether the true value of pixels includes observed values; Combining the low-resolution image, the true value of pixels, and the mask layer to obtain a multi-channel tensor; Taking the multi-channel tensor as an input and outputting a global ionospheric map based on the improved ViT model; The improved ViT model is obtained by improving the ViT model. The ViT model includes a single-channel input layer, an image feature embedding module, a Transformer encoder, an MLP classification module, and an output layer connected in sequence. Replacing the single-channel input layer with a multi-channel input layer and replacing the MLP classification module with an upsampling module to obtain the improved ViT model.

[0006] Further, the ionospheric TEC data is provided by the IRI2020 model, and the VTEC data is provided by ground GNSS stations.

[0007] Further, generating a mask layer according to whether the pixel true value includes an observation value includes: Mark the pixel true value containing the observation value as valid, and mark the pixel true value without the observation value as invalid. The valid is represented by 1, and the invalid is represented by 0; The marking results of the pixel true values form a binary mask layer.

[0008] Further, the Transformer encoder includes multiple Transformer encoding layers. Each Transformer encoding layer is used to take the pixel true value and the mask layer as additional channel information, and use the additional channel information as weights to dynamically adjust the global information capture through the multi-head self-attention mechanism to obtain low-resolution features.

[0009] Further, the upsampling module includes an upsampling unit and a convolutional layer connected in sequence. The upsampling unit uses a transposed convolutional layer or pixel recombination. The upsampling module is used to map the low-resolution features to a high-resolution image.

[0010] Further, it also includes pre-training the improved ViT model. The pre-training method includes: Obtain historical ionospheric TEC data and VTEC data; Divide the ionospheric TEC data and VTEC data corresponding in time into a training set and a test set; Use the training set data as input to train the improved ViT model, determine the optimal parameters of the improved ViT model, and obtain a preliminary improved ViT model; Use the test set data as input to test the improved ViT model to obtain test results; Calculate evaluation metrics according to the test results, and obtain the pre-trained improved ViT model based on the evaluation metrics.

[0011] In a second aspect, the present invention also provides an ionospheric super-resolution refinement reconstruction device based on an improved ViT model, including: A data acquisition module configured to acquire ionospheric TEC data and VTEC data during a period to be measured; A data processing module configured to use the ionospheric TEC data as a low-resolution image, the VTEC data as a pixel true value, and generate a mask layer according to whether the pixel true value includes an observation value, and merge the low-resolution image, the pixel true value, and the mask layer to obtain a multi-channel tensor; The ionosphere reconstruction module is configured to take a multi-channel tensor as input and output a global ionosphere map based on the improved ViT model.

[0012] In a third aspect, the present invention also 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 ionosphere super-resolution refinement reconstruction method based on the improved ViT model as described in any one of the above.

[0013] In a fourth aspect, the present invention also provides a device, including: A memory for storing instructions; A processor for executing the instructions, so that the device executes the ionosphere super-resolution refinement reconstruction method based on the improved ViT model as described in any one of the above.

[0014] Compared with the prior art, the beneficial effects achieved by the present invention are: The present invention uses the global self-attention mechanism of the ViT model to effectively capture the spatial dependence relationship of the ionosphere image, and introduces the pixel true value and the mask layer, which enhances the model's reconstruction ability for different data regions. The use of multi-channel fusion, upsampling operation and convolutional layer processing ensures the high resolution and detail accuracy of the reconstructed image; The present invention applies the ViT model to the super-resolution reconstruction of the ionosphere, which can effectively reduce the dependence on the number of ground stations, overcome the shortcoming of over-smoothing of the image, and at the same time improve the spatial resolution of the ionosphere image, thereby improving the accuracy of the ionosphere model. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic flowchart of the ionosphere super-resolution refinement reconstruction method based on the improved ViT model in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] The present invention will be further described below with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention, and cannot be used to limit the protection scope of the present invention. Embodiment 1

[0017] As Figure 1 shown, the embodiment of the present invention provides an ionosphere super-resolution refinement reconstruction method based on the improved ViT model, including the following steps: First, obtain the ionosphere TEC data provided by the IRI2020 model and the VTEC data of ground GNSS stations during the period to be measured.

[0018] Take the ionospheric TEC data as a low-resolution image, the VTEC data as the true pixel value, and then generate a mask layer according to whether the true pixel value includes the observed value. Mark the true pixel value containing the observed value as valid, represented by 1, and the true pixel value not containing the observed value as invalid, represented by 0. The marking results of all true pixel values form a binary mask layer.

[0019] Merge the low-resolution image, the true pixel value, and the mask layer to form a multi-channel tensor. The shape of the multi-channel tensor is [B, C, H, W], where B represents the batch size, C represents the number of channels, H represents the image height, and W represents the image width.

[0020] Construct an improved ViT model, which is improved based on the ViT model. The ViT model includes a single-channel input layer, an image feature embedding module, a Transformer encoder, an MLP classification module, and an output layer connected in sequence. In this embodiment, the single-channel input layer is replaced with a multi-channel input layer, and the MLP classification module is replaced with an upsampling module to obtain the improved ViT model. The specific data processing process includes: The multi-channel tensor is input by the multi-channel input layer. The multi-channel tensor is segmented by the image embedding module to obtain several small blocks (patches). The size of each patch is P×P, and each flattened patch is embedded into a high-dimensional space to form an input sequence.

[0021] The Transformer encoder includes multiple Transformer encoding layers. Each Transformer encoding layer takes the true pixel value and the mask layer as additional channel information, and uses the additional channel information as weights to dynamically adjust the global information capture of the input sequence through the multi-head self-attention mechanism to obtain low-resolution features.

[0022] The upsampling module consists of an upsampling unit and a convolutional layer connected in sequence. The upsampling unit uses a transposed convolutional layer or pixel recombination. The low-resolution features are mapped to a high-resolution image through the upsampling module, and the local detail information is retained for further refined reconstruction. Finally, the high-resolution global ionospheric map is output through the output layer.

[0023] Pre-train the constructed improved ViT model. The pre-training method includes: Download the ionospheric TEC data provided by the IRI2020 model from 2010 to 2020 for a total of 11 years (one solar activity cycle) as the low-resolution image, and the VTEC data obtained from ground GNSS stations during the same period as the true pixel value. At the same time, generate a mask layer according to whether the true pixel value includes the observed value. The low-resolution image, the true pixel value, and the mask layer during the same period constitute the dataset.

[0024] The dataset is divided into a training set and a test set according to a certain ratio.

[0025] Using the training set data as input, train the improved ViT model to determine the optimal parameters of the improved ViT model, and obtain a preliminary improved ViT model.

[0026] Using the test set data as input, test the improved ViT model to obtain test results.

[0027] Calculate evaluation metrics based on the test results, and obtain a pre-trained improved ViT model based on the evaluation metrics.

[0028] Taking the multi-channel tensor composed of the low-resolution image, pixel ground truth, and mask layer as input, the global ionospheric map (GIM) of the Center for Orbit Determination in Europe (CODE) can be output based on the pre-trained improved ViT model. Embodiment 2

[0029] Based on Embodiment 1, this embodiment also provides an ionospheric super-resolution refinement reconstruction device based on the improved ViT model, including: A data acquisition module configured to acquire ionospheric TEC data and VTEC data during the period to be measured.

[0030] A data processing module configured to use the ionospheric TEC data as a low-resolution image, the VTEC data as pixel ground truth, and generate a mask layer according to whether the pixel ground truth includes observed values, and merge the low-resolution image, pixel ground truth, and mask layer to obtain a multi-channel tensor.

[0031] An ionospheric reconstruction module configured to take the multi-channel tensor as input and output a global ionospheric map based on the improved ViT model. Embodiment 3

[0032] Based on Embodiment 1, this embodiment also 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 ionospheric super-resolution refinement reconstruction method based on the improved ViT model as described in Embodiment 1. Embodiment 4

[0033] Based on Embodiment 1, this embodiment also provides a device, characterized by including: A memory for storing instructions; A processor for executing the instructions, such that the device executes the ionospheric super-resolution refinement reconstruction method based on the improved ViT model as described in Embodiment 1.

[0034] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for ionospheric super-resolution refinement reconstruction based on an improved ViT model, characterized in that: include: Obtain ionospheric TEC data and VTEC data during the test period; The ionospheric TEC data is used as a low-resolution image, the VTEC data is used as a pixel true value, and a mask layer is generated according to whether the pixel true value includes an observed value; Combine the low-resolution image, pixel true value, and mask layer to obtain a multi-channel tensor; Taking multi-channel tensors as input, the global ionosphere map is obtained based on the output of the improved ViT model; The improved ViT model is obtained based on the improvement of the ViT model. The ViT model includes a single-channel input layer, an image feature embedding module, a Transformer encoder, an MLP classification module and an output layer connected in sequence. The single-channel input layer is replaced by a multi-channel input layer, and the MLP classification module is replaced by an upsampling module to obtain the improved ViT model.

2. The ionospheric super-resolution refined reconstruction method based on the improved ViT model according to claim 1, characterized in that: The ionospheric TEC data is provided by the IRI2020 model, and the VTEC data is provided by the ground GNSS station.

3. The ionospheric super-resolution refined reconstruction method based on the improved ViT model according to claim 1, characterized in that: The step of generating a mask layer according to whether the true value of the pixel includes the observed value comprises: Mark the true value of the pixel containing the observed value as valid, and mark the true value of the pixel not containing the observed value as invalid, where the valid value is represented by 1 and the invalid value is represented by 0; The marking results of the true values ​​of the pixels form a binary mask layer.

4. The ionospheric super-resolution refined reconstruction method based on the improved ViT model according to claim 1, characterized in that: The Transformer encoder includes multiple Transformer encoding layers, each of which is used to take the true value of the pixel and the mask layer as additional channel information, and use the additional channel information as a weight to dynamically adjust the global information through a multi-head self-attention mechanism to capture low-resolution features.

5. The ionospheric super-resolution refined reconstruction method based on the improved ViT model according to claim 1, characterized in that: The upsampling module includes an upsampling unit and a convolution layer connected in sequence, the upsampling unit adopts a deconvolution layer or pixel reorganization, and the upsampling module is used to map low-resolution features to a high-resolution image.

6. The ionospheric super-resolution refined reconstruction method based on the improved ViT model according to claim 1, characterized in that: The improved ViT model is also pre-trained, and the pre-training method includes: Obtain historical ionospheric TEC data and VTEC data; The ionospheric TEC data and VTEC data corresponding to the time are divided into training set and test set; The training set data is used as input to train the improved ViT model, determine the optimal parameters of the improved ViT model, and obtain a preliminary improved ViT model; The test set data is used as input to test the improved ViT model and obtain the test results; The evaluation indicators are calculated according to the test results, and the pre-trained improved ViT model is obtained based on the evaluation indicators.

7. An ionospheric super-resolution refined reconstruction device based on an improved ViT model, characterized in that: include: A data acquisition module is configured to acquire ionospheric TEC data and VTEC data within a time period to be measured; A data processing module is configured to use the ionospheric TEC data as a low-resolution image and the VTEC data as a pixel true value, generate a mask layer according to whether the pixel true value includes an observed value, and merge the low-resolution image, the pixel true value and the mask layer to obtain a multi-channel tensor; The ionospheric reconstruction module is configured to take multi-channel tensors as input and obtain the global ionospheric map based on the improved ViT model output.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the ionospheric super-resolution refined reconstruction method based on the improved ViT model as described in any one of claims 1 to 6 is implemented.

9. A device, characterized in that: include: A memory for storing instructions; A processor is used to execute the instruction so that the device implements the ionospheric super-resolution refined reconstruction method based on the improved ViT model as described in any one of claims 1 to 6.

Images