Navigation interference identification method based on satellite-borne GNSS-R data, storage medium and equipment

Abstract

The invention discloses a navigation interference identification method based on satellite-borne GNSS-R data, a storage medium and equipment. The method comprises the following steps: firstly, acquiring the satellite-borne GNSS-R data, generating a time delay Doppler diagram of a channel, and dividing a training set, a verification set and a test set of semi-supervised learning according to a proportion; applying two different random enhancements to each DDM image to generate two views, inputting the views into a neural network model, designing a contrast loss function, performing contrast learning on the two preprocessed views, loading a pre-training weight, executing a training process, optimizing model parameters, and storing a model with minimum contrast loss; a binary classification network head is designed as classification output, training and verification are carried out through DDM image data with an RFI tag, an early stop mechanism is introduced, test set data are input into a trained RFI recognition model for RFI recognition, and development of satellite-borne GNSS-R data in the navigation interference recognition direction is promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of navigation interference source identification and positioning, and mainly relates to a navigation interference identification method, a storage medium and a device based on spaceborne GNSS-R data. Background Art

[0002] With the rapid development of electromagnetic technology, the problem of radio frequency interference (RFI) has become increasingly prominent. Especially in the fields of radar, satellite communication and radio astronomy, RFI will seriously affect the signal quality and lead to the degradation of system performance. Among them, for the RFI identification of the global navigation satellite system (GNSS), GNSS / low-orbit navigation observation data, ephemeris data, ADS-B and AIS data can be used, and special interference monitoring equipment can be deployed for detection (Patent CN109359523B). Navigation interference identification based on GNSS-R data has great application value due to its wide coverage, fast data update and multi-satellite networking.

[0003] GNSS-R is a remote sensing means that uses the reflected signals of navigation systems such as GNSS to invert physical parameters such as the ocean, the earth's surface, and vegetation. Spaceborne GNSS-R can obtain delay-Doppler maps (DDMs) for calculating geophysical parameters such as soil moisture and sea surface wind speed. When the DDM data is affected by RFI, it will present different signal characteristics, which can effectively characterize navigation interference. However, the DDM data has characteristics such as sparsity and irregularity, which pose great challenges to RFI interference identification. At present, the evaluation of RFI in spaceborne GNSS-R data mainly relies on parameters such as the kurtosis value of the noise floor power of the DDM data, with poor accuracy and weak robustness, resulting in limited inversion accuracy of geophysical parameters.

[0004] Semi-supervised learning, as a learning method between supervised learning and unsupervised learning, can learn by using a small amount of labeled data and a large amount of unlabeled data, and is particularly suitable for application scenarios where RFI labeled data is scarce in GNSS-R. However, RFI identification based on semi-supervised learning of delay-Doppler map data still faces many challenges. First, there are many types of RFI with complex characteristics, which are difficult to describe with a simple mathematical model. Second, in actual application scenarios, RFI often overlaps with target signals and is difficult to separate. In addition, the performance of semi-supervised learning algorithms highly depends on data quality and feature extraction methods. How to design effective feature extraction methods and semi-supervised learning algorithms is the key to achieving effective RFI identification. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention proposes a navigation interference recognition method, a storage medium and a device based on spaceborne GNSS-R data. First, spaceborne GNSS-R data is acquired to generate a delay-Doppler map of a channel, and a training set, a validation set and a test set for semi-supervised learning are divided in proportion. Apply two different random augmentations to each DDM image to generate two views, and input them into the RFI recognition model. Design a contrastive loss function to perform contrastive learning on the two pre-processed views. Load the pre-trained weights, execute the training process, optimize the model parameters, and save the model with the minimum contrastive loss. Design a binary classification network head as the classification output, train and validate it through the DDM image data with RFI labels, and introduce an early stopping mechanism. Input the test set data into the trained RFI recognition model for RFI recognition. The method of the present invention combines a semi-supervised learning method to construct a DDM feature extraction model, which can further improve the ability to accurately identify RFI globally using spaceborne GNSS-R technology, and promotes the development of machine learning in spaceborne GNSS-R interference recognition.

[0006] To achieve the above object, the technical solution adopted by the present invention is: a navigation interference recognition method based on spaceborne GNSS-R data, including the following steps:

[0007] S1: Acquire spaceborne GNSS-R reflection observation data, generate a delay-Doppler map of a channel according to the data, and divide a training set, a validation set and a test set for semi-supervised learning in proportion; the channels include but are not limited to GPS, Beidou, Galileo and GLONASS satellite navigation systems, and the spaceborne GNSS-R reflection observation data includes at least DDM waveform data;

[0008] S2: Perform two different random augmentations on each DDM image to generate two augmented views; the data augmentation methods include at least random color jitter, random grayscale conversion and random horizontal flipping, wherein the random color jitter is divided into random brightness, random contrast, random saturation and random hue adjustment;

[0009] S3: Input the two augmented views generated in step S2 into the RFI recognition model, set a contrastive loss function for contrastive learning training; load the pre-trained model weights, execute the training process, calculate the loss of two image pairs after forward propagation, and optimize the model parameters through backpropagation to obtain the model with the minimum contrastive loss;

[0010] S4: Use a binary classification network head as the classification output, load the weights of the model with the minimum contrastive loss obtained in step S3, train and validate it through the DDM image data with RFI labels, and introduce an early stopping mechanism; input the test set data into the trained RFI recognition model for RFI recognition.

[0011] As an improvement of the present invention, in the step S1, the on-board GNSS-R reflection observation data includes DDM waveform data, and the data set is randomly divided into an unsupervised learning data set and a supervised learning data set according to the ratio of 80% and 20%. Among them, the supervised learning data set is further divided into a training set, a validation set and a test set according to the ratio of 80%, 10% and 10%.

[0012] As an improvement of the present invention, in the step S2, the random brightness enhancement in the random color jitter is specifically:

[0013] I bright (x, y, c) = β · I(x, y, c), c ∈ {R, G, B}

[0014] where β is a random brightness factor, and I(x, y, c) is the pixel value of the c-th channel of the image at the position (x, y);

[0015] The calculation process of the random contrast enhancement is specifically:

[0016] I contrast (x, y, c) = μ + α · (I(x, y, c) - μ)

[0017] where α is a random contrast factor and μ is the average gray value of the image.

[0018] As another improvement of the present invention, in the step S2, the random saturation enhancement in the random color jitter is specifically:

[0019] Convert the image to a grayscale image I gray :

[0020] I gray (x, y) = 0.2989 · I R (x, y) + 0.5870 · I G (x, y) + 0.1140 · I B (x, y)

[0021] where I R (x, y), I G (x, y) and I B (x, y) are the pixel values of the R, G, and B channels of the image at the position (x, y), respectively.

[0022] Mix the RGB channels:

[0023] I saturation = (1 - γ) · I gray (x, y) + γ · I(x, y, c)

[0024] where γ is a random saturation factor.

[0025] As another improvement of the present invention, in the step S2, the random hue adjustment in the random color jittering is specifically as follows:

[0026] Convert the image from the RGB space to the HSV space, and perform an addition operation on the hue channel H:

[0027] H hue =(H + δ) mod 1

[0028] Then convert the image back to the RGB space;

[0029] The calculation process of the random grayscale is as follows:

[0030] I gray I(x,y) = 0.2989·I R (x,y) + 0.5870·I G (x,y) + 0.1140·I B (x,y)

[0031]

[0032] where p is the probability of grayscale conversion, I raw (x,y) is the pixel value of the original image, and I gray (x,y) is the pixel value of the image after grayscale processing.

[0033] As yet another improvement of the present invention, the RFI recognition model in the step S3 includes but is not limited to using SwinTransformer, MobilenetV2, and Resnet models as the backbone networks, and using the NTXentLoss function as the contrast loss function, specifically:

[0034]

[0035] where z i , z j are the feature embedding vectors of two enhanced views of the same DDM image respectively, sim(·) is the cosine similarity, and τ is the temperature coefficient.

[0036] As a further improvement of the present invention, the step S4 specifically includes the following steps:

[0037] S41: Use BCEWithLogitsLoss and dynamically set the weight coefficient of the positive samples, calculate the loss according to the positive and negative sample ratio of the training data, so as to alleviate the class imbalance problem, use the AdamW optimizer, set the learning rate to 0.00005, and use the OneCycleLR scheduler to dynamically adjust the learning rate, taking into account both the training speed and stability;

[0038] S42: Use Autocast + GradScaler for mixed-precision training, dynamically manage the numerical precision of different operators, convert some operators to FP16 (half-precision floating-point numbers), and keep the rest as FP32 (single-precision floating-point numbers) to accelerate the calculation; GradScaler ensures the effectiveness of backpropagation by dynamically scaling the gradients;

[0039] S43: Introduce an early stopping mechanism. The main metrics are Loss and Accuracy. Tolerate no improvement for 8 rounds, and real-time plot the trends of Loss and Accuracy for visual diagnosis to avoid overfitting;

[0040] S44: After each round of training, the model is evaluated on the validation set, the validation loss and accuracy are calculated, and the weights of the fine-tuned RFI recognition model are saved;

[0041] S45: Input the test set divided in step S1 into the RFI recognition model, and calculate relevant metrics to evaluate the model performance.

[0042] As a further improvement of the present invention, in step S41, the calculation process of the BCEWithLogitsLoss function is as follows:

[0043] loss(x,y) = -[y·log(σ(x)) + (1 - y)·log(1 - σ(x))]

[0044] where x is the weight output by the model, σ(x) is the Sigmoid function, and y is the true label;

[0045] In step S43, the calculation process of Accuracy is as follows:

[0046]

[0047] To achieve the above object, the technical solution adopted by the present invention is also: a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the navigation interference recognition method based on spaceborne GNSS-R data as described in any one of claims 1-8.

[0048] To achieve the above object, the technical solution adopted by the present invention is also: a computer device, comprising:

[0049] A memory for storing instructions;

[0050] A processor for executing the instructions, so that the computer device performs the operations of the navigation interference recognition method based on spaceborne GNSS-R data as described in any one of claims 1-8.

[0051] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention discloses a navigation interference recognition method, storage medium and device based on spaceborne GNSS-R data, which uses a learning method combining unsupervised learning and supervised learning, and uses the delay-Doppler map in GNSS-R data for RFI recognition; The difference from the traditional RFI recognition method is that; using the complete delay-Doppler map information combined with the lightweight and high-precision network model Swin Transformer enhances the generalization ability of the RFI recognition method and improves the robustness; in the application of spaceborne GNSS-R data, this method can effectively identify the RFI suffered by GNSS signals. Description of the Drawings

[0052] Figure 1 It is the flowchart of the steps of the method of the present invention;

[0053] Figure 2 They are delay-Doppler maps in different scenarios, where

[0054] Figure 2 (a) is a classic delay-Doppler map affected by RFI;

[0055] Figure 2 (b) is a classic delay-Doppler map of sea surface reflection;

[0056] Figure 2 (c) is a classic delay-Doppler map of ice surface reflection;

[0057] Figure 2 (d) is a delay-Doppler map suspected of being damaged by RFI;

[0058] Figure 3 It is the neural network structure diagram of swin_base_patch4_window7_224 in Embodiment 1 of the present invention. Detailed Embodiments

[0059] The following further clarifies the present invention in conjunction with the drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and not to limit the scope of the present invention.

[0060] Embodiment 1

[0061] The navigation interference recognition based on spaceborne GNSS-R data is a research direction with important research significance and application value. By deeply studying the feature extraction method of RFI interference signals and semi-supervised learning algorithms, it is expected to achieve high-accuracy and robust navigation interference recognition and provide guarantee for global navigation security. As Figure 1 shown, a navigation interference recognition method based on spaceborne GNSS-R data, the method includes the following steps:

[0062] Step S1: Obtain spaceborne GNSS-R data, generate a delay-Doppler map for a certain channel (such as GPS, Beidou, Galileo, and GLONASS satellite navigation systems) using this data, and then divide it into a training set, a validation set, and a test set for semi-supervised learning according to a certain ratio.

[0063] The spaceborne GNSS-R reflection observation data includes DDM waveform data (Ddm_raw_data); among them, 80% of the data in the dataset is used for unsupervised contrast learning, 80% of the remaining data is used for supervised training, 10% is used for validation, and the remaining 10% is used for testing.

[0064] For example, one can download one-day reflection L1-level data by accessing the official website of the Tianmu-1 remote sensing meteorological detection constellation, obtain the delay-Doppler map of the C1 channel from 0 to 24 hours. In this embodiment, delay-Doppler maps in different scenarios are selected, specifically as Figure 2 shown, Figure 2 (a) is a classic delay-Doppler map affected by RFI, Figure 2 (b) is a classic delay-Doppler map of sea surface reflection, Figure 2 (c) is a classic delay-Doppler map of ice surface reflection, Figure 2 (d) is a suspected delay-Doppler map damaged by RFI, where the horizontal axis is the frequency shift axis and the vertical axis is the delay axis, and it is randomly divided into an unsupervised learning dataset and a supervised learning dataset according to the ratio of 80% and 20%. Among them, the supervised learning dataset is divided into a training set, a validation set, and a test set according to the ratio of 80%, 10%, and 10%.

[0065] Step S2: Apply two different random augmentations to each DDM image to generate two views for contrast learning;

[0066] Perform different random augmentations on the images in the unsupervised contrast learning data, including random color jitter, random grayscale conversion, and random horizontal flipping, to obtain a set of image pairs for contrast learning; among them, random color jitter is divided into random brightness, contrast, saturation, and hue adjustments. The calculation process of random brightness adjustment is as follows:

[0067] I bright (x, y, c) = β · I(x, y, c), c ∈ {R, G, B}

[0068] where β is the random brightness factor and I is the input image.

[0069] The calculation process of random contrast adjustment is as follows:

[0070] I contrast (x, y, c) = μ + α · (I(x, y, c) - μ)

[0071] Where α is a random contrast factor and μ is the average gray value of the image.

[0072] The calculation process of random saturation adjustment is as follows:

[0073] First, convert the image to a grayscale image I gray :

[0074] I gray (x,y) = 0.2989·I R + 0.5870·I G + 0.1140·I B

[0075] Then, perform mixing on the RGB channels:

[0076] I saturation = (1 - γ)·I gray (x,y) + γ·I(x,y,c)

[0077] Where γ is a random saturation factor.

[0078] For random hue adjustment, first convert the image from the RGB space to the HSV space, and perform an addition operation on the hue channel H. The calculation process is as follows:

[0079] H hue = (H + δ) mod 1

[0080] Then, convert the image back to the RGB space.

[0081] The calculation process of random grayscale conversion is as follows:

[0082] gray(x,y) = 0.2989·R(x,y) + 0.5870·G(x,y) + 0.1140·B(x,y)

[0083]

[0084] Where p is the probability of grayscale conversion.

[0085] In this embodiment, the random brightness adjustment factor, random contrast adjustment factor, random saturation adjustment factor, and random hue adjustment factor are respectively set to 0.8, 0.8, 0.8, and 0.2, and the random grayscale conversion factor is set to 0.2. Two different DDMs are randomly enhanced to obtain a pair of images, and they are normalized to the range of [0,1].

[0086] Step S3: Select a deep neural network model such as Swin Transformer, MobilenetV2, Resnet, etc. as the backbone network; design a contrastive loss function to perform contrastive learning on the two preprocessed views; load the pre-trained weights, execute the training process, optimize the model parameters, and save the model with the minimum contrastive loss.

[0087] In an embodiment of the present invention, the pre-trained model of the Swin Transformer network is the swin_base_patch4_window7_224 model provided by the HuggingFace website, which has a high image classification accuracy.

[0088] The contrastive learning loss function uses the NTXentLoss function. By calculating the similarity between the features of the two views, it maximizes the similarity of positive sample pairs (different views of the same image) while minimizing the similarity of negative sample pairs (views of different images); the calculation process is as follows:

[0089]

[0090] where z i , z j are the feature embedding vectors of two augmented views of the same DDM image respectively, sim(·) is the cosine similarity, and τ is the temperature coefficient, which is set to 0.5 in this embodiment. Adam is introduced as the optimizer, and the initial learning rate is set to 0.0001.

[0091] Load the pre-trained model weights, set the number of self-supervised learning epochs to 500, execute the training process, input the image pairs into the network in each epoch, calculate the losses of the two image pairs after forward propagation, optimize the parameters through backpropagation, calculate the output losses and optimize the parameters, record the optimal loss, and dynamically save the model parameters corresponding to the optimal loss.

[0092] Step S4: Design a binary classification network head as the classification output, load the weights of the model with the minimum contrastive loss saved in the previous step, train and validate through the DDM image data with RFI labels, and introduce an early stopping mechanism; input the test set data into the trained RFI recognition model for RFI recognition.

[0093] Load the model weights trained in Step S3, optimize the existing network according to the neural network structure diagram as Figure 3 shown, use multi-layer perception for progressive dimensionality reduction, equip each layer with BatchNorm, GELU, and Dropout, and modify the classification head to adapt to the binary classification task.

[0094] Introduce the class balance loss function BCEWithLogitsLoss, which automatically adjusts the loss weight according to the positive and negative sample ratio to alleviate the class imbalance problem; use a lower learning rate (lr / 10) for the Backbone parameters and a normal learning rate (5e-5) for the classification head; use the AdamW optimizer with a learning rate set to 0.00005; use OneCycleLR to achieve periodic changes in the learning rate, taking into account both training speed and stability, and combine warm start to improve the convergence speed; use mixed-precision training Autocast+GradScaler to accelerate the training process; the calculation process of the BCEWithLogitsLoss function is as follows:

[0095] loss(x,y) = -[y·log(σ(x))+(1-y)·log(1-σ(x))]

[0096] Where x is the weight output by the model, σ(x) is the Sigmoid function, and y is the true label.

[0097] Take 100 training epochs, execute the training process, and record and plot the epoch change trends of Loss and Accuracy in real time; introduce an early stopping mechanism, and terminate the training process if the Accuracy does not improve after eight rounds of training, and save the optimal model parameters;

[0098] Input the test set into the RFI recognition model obtained in the previous step for RFI recognition testing, and use accuracy (Accuracy), precision (Precision), and recall (Recall) as evaluation criteria.

[0099] In a specific embodiment of the present invention, an RFI recognition model is trained with the Swin Transformer network, and the comparison of evaluation indicators and the Tianmu-1 reflected L1-level data quality flag bit9 (RFI flag) is shown in Table 1:

[0100] Table 1 Comparison index statistics of two methods for identifying RFI

[0101] RFI Detection Method Accuracy Precision Recall Swin Transformer 0.9981 1 0.9959 bit9 0.6630 0.9697 0.3937

[0102] As can be seen from Table 1, compared with the interference recognition method of Tianmu-1 itself, the Accuracy and Recall of the RFI recognition model trained with the Swin Transformer network model have increased significantly, indicating that the former has obvious advantages over the latter in detecting RFI.

[0103] In summary, the implementation scheme of the present invention not only proposes a new scheme for satellite-borne GNSS-R to identify RFI, effectively improving the accuracy of RFI identification, but also combines supervised learning with unsupervised learning, which not only enhances the generalization ability of the model but also improves the identification performance of the model.

[0104] It should be noted that the above content only illustrates the technical idea of the present invention and cannot be used to limit the protection scope of the present invention. For those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements all fall within the protection scope of the claims of the present invention.

Claims

1. A navigation interference recognition method based on spaceborne GNSS-R data, characterized in that, It includes the following steps: S1: Obtain on-orbit GNSS-R reflection observation data, generate the delay-Doppler map of the channel according to the data, and divide the training set, validation set, and test set of semi-supervised learning proportionally; the channels include but are not limited to satellite navigation systems such as GPS, Beidou, Galileo, and GLONASS, and the on-orbit GNSS-R reflection observation data includes at least DDM waveform data; S2: Perform two different random augmentations on each DDM image to generate two data-augmented views; the data augmentation methods include at least random color jitter, random grayscale conversion, and random horizontal flipping. Among them, random color jitter is divided into random brightness, random contrast, random saturation, and random hue adjustment; S3: Input the two data-augmented views generated in step S2 into the RFI recognition model, set the contrast loss function for contrast learning training; load the pre-trained model weights, execute the training process, calculate the loss of the two image pairs after forward propagation, and optimize the model parameters through backpropagation to obtain the model with the minimum contrast loss; S4: Use the binary classification network head as the classification output, load the model weights with the minimum contrast loss obtained in step S3, train and validate through the DDM image data with RFI labels, and introduce an early stopping mechanism; input the test set data into the trained RFI recognition model for RFI recognition.

2. The navigation interference recognition method based on spaceborne GNSS-R data according to claim 1, characterized in that: In step S1, the on-orbit GNSS-R reflection observation data includes DDM waveform data. The dataset is randomly divided into an unsupervised learning dataset and a supervised learning dataset according to a ratio of 80% and 20%. Among them, the supervised learning dataset is further divided into a training set, a validation set, and a test set according to a ratio of 80%, 10%, and 10%.

3. The navigation interference recognition method based on spaceborne GNSS-R data according to claim 1, characterized in that: In step S2, the random brightness enhancement in random color jitter is specifically: I bright (x, y, c) = β · I(x, y, c), c ∈ {R, G, B} where β is the random brightness factor, and I(x,y,c) is the pixel value of the c-th channel at the (x,y) position of the image; The calculation process of random contrast enhancement is specifically: I contrast (x, y, c) = μ + α · (I(x, y, c) - μ) where α is the random contrast factor and μ is the average image grayscale value.

4. The navigation interference recognition method based on spaceborne GNSS-R data according to claim 3, wherein: In step S2, the random saturation enhancement in random color jitter is specifically: Convert the image to a grayscale image I gray : I gray (x,y) = 0.2989·I R (x,y) + 0.5870·I G (x,y) + 0.1140·I B (x,y) where I R (x, y), I G (x, y) and I B (x, y) are the pixel values of the R, G, and B channels of the image at the position (x, y) respectively; the RGB channels are mixed as follows: I saturation (x,y,c) = (1 - γ)·I gray (x,y) + γ·I(x,y,c) where γ is the random saturation factor.

5. The navigation interference recognition method based on spaceborne GNSS-R data according to claim 4, wherein: In step S2, the random hue adjustment in random color jitter is specifically: Convert the image from the RGB space to the HSV space, and perform an addition operation on the hue channel H: H hue = (H + δ) mod 1 Then convert the image back to the RGB space.

6. The navigation interference recognition method based on spaceborne GNSS-R data according to claim 1, wherein: The RFI recognition model in step S3 includes but is not limited to using Swin Transformer, MobileNetV2, and ResNet models as the backbone networks, and adopting the NTXentLoss function as the contrast loss function, specifically: where z i , z j are the feature embedding vectors of two enhanced views of the same DDM image, sim(·) is the cosine similarity, and τ is the temperature coefficient, respectively.

7. The navigation interference recognition method based on spaceborne GNSS-R data according to claim 1, characterized in that: Step S4 specifically includes the following steps: S41: Use BCEWithLogitsLoss and dynamically set the weight coefficient of the positive samples, and calculate the loss according to the positive and negative sample ratio of the training data; S42: Use Autocast + GradScaler for mixed-precision training to dynamically manage the numerical precision of different operators, convert some operators to half-precision floating-point numbers, and keep the rest as single-precision floating-point numbers; GradScaler scales the gradients dynamically; S43: Introduce an early stopping mechanism. The main metrics are Loss and accuracy. If they do not improve for 8 rounds, and the trend graphs of Loss and Accuracy are plotted in real time for visual diagnosis; S44: After each round of training, the model is evaluated on the validation set, the validation loss and accuracy are calculated, and the weights of the fine-tuned RFI recognition model are saved; S45: Input the test set divided in step S1 into the RFI recognition model, and calculate evaluation metrics to evaluate the model performance.

8. The navigation interference identification method based on spaceborne GNSS-R data according to claim 7, characterized in that: In the said step S41, the calculation process of the BCEWithLogitsLoss function is as follows: loss(x,y) = -[y·log(σ(x))+(1 - y)·log(1 - σ(x))] where x is the weight output by the model, σ(x) is the Sigmoid function, and y is the true label; In the said step S43, the calculation process of Accuracy is as follows:

9. A computer-readable storage medium, characterized in that: Stored thereon is a computer program which, when executed by a processor, implements the method for identifying navigation interference based on spaceborne GNSS-R data as described in any one of claims 1-8.

10. A computer device, characterized in that: Comprising a memory for storing instructions; a processor for executing the instructions, such that the computer device performs the operations of the method for identifying navigation interference based on spaceborne GNSS-R data as described in any one of claims 1-8.

Images