Cross-modal terrain matching positioning method

Abstract

The invention provides a cross-modal terrain matching and positioning method. The method comprises the following steps: acquiring a target digital elevation model and a target two-dimensional delay Doppler map of a flight platform in a current terrain; performing cross-modal fusion on the target digital elevation model and the target two-dimensional delay Doppler map to obtain fusion matching data; performing feature extraction on the fusion matching data through different feature extraction modules to obtain a first feature vector of the target digital elevation model and a second feature vector of the target two-dimensional delay Doppler graph; and determining positioning data of the flight platform based on the first feature vector and the second feature vector. According to the method, the cross-modal digital elevation model and the two-dimensional delay Doppler map features are fused, the positioning result is obtained based on the features extracted after fusion, and the problems that aircraft positioning depends on a global positioning system, the scene generalization ability is insufficient, heterogeneous gaps exist in cross-modal terrain matching and the like are effectively solved.

Description

Technical Field

[0001] The present application relates to the technical field of aircraft positioning, and in particular to a cross-modal terrain matching positioning method. Background Art

[0002] Aircraft positioning is critical to flight safety. The onboard synthetic aperture radar altimeter (SARAL) illuminates the ground in a near-vertical manner. When the platform flies directly above a ground scattering point, the radar transmits a broadband signal, receives the echo after a certain delay, and processes the echo to obtain a two-dimensional delay Doppler map (DDM). The two-dimensional delay Doppler map (DDM) feature reflects the unexpected fluctuations of the aircraft's lowest point and can be used to detect terrain changes.

[0003] Since the two-dimensional delay-Doppler map (DDM) is a projection of the three-dimensional terrain onto a two-dimensional plane by the synthetic aperture radar altimeter (SARAL), people cannot directly and intuitively interpret the terrain changes from the two-dimensional delay-Doppler map (DDM). When the two-dimensional delay-Doppler map (DDM) formed by radar echo data is matched with terrain data, since the two belong to two heterogeneous modes and have different quantization scales, there is a semantic gap between the modes, which easily leads to the problem of limited matching accuracy; this problem is more serious in the case of non-flat terrain, complex changes in scattering characteristics, and platform errors. Summary of the invention

[0004] In view of this, the embodiment of the present application provides a cross-modal terrain matching positioning method to solve the problem that the existing digital elevation model (DEM) and two-dimensional delay Doppler map (DDM) belong to different modes and have a heterogeneous gap and are difficult to directly match. The following technical solutions are specifically adopted.

[0005] A cross-modal terrain matching positioning method is used for a flying platform, wherein the flying platform is equipped with an airborne synthetic aperture radar altimeter, and the method comprises: Acquire a target digital elevation model and a target two-dimensional delay Doppler map of the flight platform in the current terrain, wherein the two-dimensional delay Doppler map is obtained by performing a radar scan of the current terrain by the airborne synthetic aperture radar altimeter; The target digital elevation model and the target two-dimensional delay-Doppler map are respectively encoded into a first coding space by a first encoder and a second encoder to obtain a first coding feature corresponding to the target digital elevation model and a second coding feature corresponding to the target two-dimensional delay-Doppler map; In the first coding space, first feature data corresponding to the first coding feature is calculated, and second feature data corresponding to the second coding feature is calculated; Based on the second feature data, a first cross-modal fusion vector corresponding to the first coding feature is calculated; based on the first feature data, a second cross-modal fusion vector corresponding to the second coding feature is calculated; Decoding the first cross-modal fusion vector and the second cross-modal fusion vector respectively through a first decoder and a second decoder to obtain a first decoding feature corresponding to the first cross-modal fusion vector and a second decoding feature corresponding to the second cross-modal fusion vector; The first decoding feature and the second decoding feature are respectively extracted by different feature extraction modules to obtain a first feature vector of the target digital elevation model and a second feature vector of the target two-dimensional delay-Doppler map; Based on the first feature vector and the second feature vector, positioning data of the flying platform is determined.

[0006] In the above scheme, the cross-modal terrain matching and aircraft positioning network integrates the cross-modal digital elevation model (DEM) and two-dimensional delay Doppler map (DDM) features, and then performs feature extraction. Based on the extracted features, the positioning results are obtained, which effectively solves the problems of aircraft positioning relying on the global positioning system (GPS), insufficient scene generalization capabilities, and heterogeneous gaps in cross-modal terrain matching.

[0007] In some embodiments, the step of encoding the target digital elevation model and the target two-dimensional delay-Doppler map into a first coding space by a first encoder and a second encoder respectively to obtain a first coding feature corresponding to the target digital elevation model and a second coding feature corresponding to the target two-dimensional delay-Doppler map specifically includes: The target digital elevation model is encoded by a first encoder to obtain a first encoding feature map; And, encoding the target two-dimensional delay-Doppler map by a second encoder to obtain a second encoding feature map; The first coding feature map and the second coding feature map are respectively subjected to matrix multiplication calculations using unit matrices of different dimensions to obtain a first coding feature corresponding to the target digital elevation model and a second coding feature corresponding to the target two-dimensional delay-Doppler map, wherein the first coding feature and the second coding feature have the same dimension.

[0008] In some embodiments, the step of calculating, in the first coding space, first feature data corresponding to the first coding feature and the step of calculating second feature data corresponding to the second coding feature specifically include: For the first coding feature, linear mapping is performed through the first weight matrix to obtain a first mapping matrix corresponding to the first coding feature; for the second coding feature, linear mapping is performed through the second weight matrix and the third weight matrix to obtain a second mapping matrix and a third mapping matrix corresponding to the second coding feature; For the second coding feature, linear mapping is performed through the first weight matrix to obtain a fourth mapping matrix corresponding to the second coding feature. For the first coding feature, linear mapping is performed through the second weight matrix and the third weight matrix to obtain a sixth mapping matrix and a third mapping matrix corresponding to the fifth coding feature.

[0009] In some embodiments, the step of calculating a first cross-modal fusion vector corresponding to the first coding feature based on the second feature data, and calculating a second cross-modal fusion weight matrix corresponding to the second coding feature based on the first feature data specifically includes: Based on the first mapping matrix and the second mapping matrix, obtaining a second cross-modal fusion weight matrix corresponding to the first encoding feature; Based on the fourth mapping matrix and the fifth mapping matrix, obtaining a first cross-modal fusion weight matrix corresponding to the second encoding feature; Performing matrix multiplication on the second cross-modal fusion weight matrix and the third mapping matrix, and normalizing the matrix multiplication result to obtain a first cross-modal fusion vector; And, matrix multiplication is performed on the first cross-modal fusion weight matrix and the sixth mapping matrix, and the matrix multiplication result is standardized to calculate and obtain a second cross-modal fusion vector.

[0010] In some embodiments, further, the step of extracting the first decoding feature and the second decoding feature respectively through different feature extraction modules to obtain a first feature vector of the target digital elevation model and a second feature vector of the target two-dimensional delay-Doppler map specifically includes: Performing feature extraction processing on the first decoded data by a first feature extraction module to obtain a first feature vector of the target digital elevation model, wherein the first feature vector corresponds to the first decoded data; The second decoded data is subjected to feature extraction processing by a second feature extraction module to obtain a second feature vector of the target two-dimensional delay-Doppler image, wherein the second feature vector corresponds to the second decoded data.

[0011] In some embodiments, the step of determining the positioning data of the flying platform based on the first feature vector and the second feature vector specifically includes: Calculating the cosine similarity between the first feature vector and the second feature vector; Based on the cosine similarity, the positioning data of the flying platform is determined.

[0012] In some embodiments, the step of calculating the cosine similarity between the first feature vector and the second feature vector specifically includes: The cosine similarity satisfies the following formula:

[0013] in, represents the cosine similarity between the first feature vector and the second feature vector, represents the first eigenvector, represents the second eigenvector, represents the coordinates of the target in the two-dimensional delay Doppler map, Represents the coordinates of the target digital elevation model.

[0014] In some embodiments, the step of determining the positioning data of the flying platform based on the cosine similarity specifically includes: For each position coordinate in the target two-dimensional delay-Doppler image, based on the second feature vector, selecting a plurality of the second feature vectors with the greatest similarity; Based on the plurality of second feature vectors with the greatest similarity, reversely retrieve candidate coordinates mapped in the target digital elevation model, each candidate coordinate corresponding to one second feature vector; Based on the multiple candidate coordinates, the positioning data of the flying platform is determined.

[0015] In some embodiments, the step of determining the positioning data of the flying platform based on the plurality of candidate coordinates specifically includes: Performing weighted calculation on the plurality of candidate coordinates to obtain weighted coordinate values; Based on the weighted coordinate values, the positioning data of the flying platform is determined.

[0016] The present application discloses a cross-modal terrain matching positioning method, which obtains the target digital elevation model and the target two-dimensional delay Doppler map of the flight platform in the current terrain; cross-modally fuses the target digital elevation model and the target two-dimensional delay Doppler map to obtain fused matching data; extracts features from the fused matching data through different feature extraction modules to obtain the first feature vector of the target digital elevation model and the second feature vector of the target two-dimensional delay Doppler map; based on the first feature vector and the second feature vector, determines the positioning data of the flight platform. The present application fuses the features of the cross-modal digital elevation model and the two-dimensional delay Doppler map, and obtains the positioning result based on the features extracted after feature fusion. This method effectively solves the problems of aircraft positioning relying on the global positioning system (GPS), insufficient scene generalization capability, and heterogeneous gap in cross-modal terrain matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a flowchart of a cross-modal terrain matching positioning method provided in an embodiment of the present application; Figure 2 It is a schematic diagram of a model architecture for a cross-modal terrain matching positioning method provided in an embodiment of the present application; Figure 3 is a schematic diagram of a simulated two-dimensional delay Doppler map provided in an embodiment of the present application; Figure 4 is a schematic diagram of a measured two-dimensional delay Doppler map provided in an embodiment of the present application; Figure 5 It is a regional digital elevation model (DEM) and a schematic diagram of a certain track provided in an embodiment of the present application; Figure 6 This is a result diagram of terrain matching and real-time aircraft positioning using this embodiment provided by an embodiment of the present application. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those generally understood by ordinary technicians in the technical field to which the present application belongs. In addition, it should also be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof. In the description of the present application, it is necessary to understand that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", etc. is based on the orientation or positional relationship shown in the accompanying drawings, which is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present application. The terms "first", "second", "third", and "fourth" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. In the description of this application, unless otherwise specified, "plurality" means two or more, unless otherwise clearly defined.

[0021] The Synthetic Aperture Radar Altimeter (SARAL) is not limited by lighting and weather conditions and is capable of generating a two-dimensional delay Doppler map (DDM). The airborne Synthetic Aperture Radar Altimeter (SARAL) illuminates the ground in a near-vertical manner. When the platform flies directly over a ground scattering point, the radar transmits a broadband signal, receives the echo after a certain delay, and processes the echo to generate a two-dimensional delay Doppler map (DDM). The two-dimensional delay Doppler map (DDM) feature reflects the unexpected fluctuations of the aircraft's lowest point and can be used to detect terrain changes.

[0022] A digital elevation model (DEM) is a digital representation of the elevation of the earth's surface, which records elevation information at different locations in a discrete form. At the same time, digital elevation model (DEM) data is available worldwide. Since the digital elevation model (DEM) reflects the ground elevation, this application uses the digital elevation model (DEM) as a reference to match the real-time two-dimensional delay Doppler map (DDM) to achieve aircraft positioning without relying on the global positioning system (GPS).

[0023] Deep learning has strong nonlinear fitting and massive data processing capabilities, which attracts researchers to use deep learning to solve image matching problems in different scenarios. These methods can be broadly divided into two categories: single-modal matching and positioning methods and cross-modal matching and positioning methods. Specifically, single-modal matching and positioning methods make predictions by analyzing data from a single information source or form, while cross-modal matching and positioning methods require combining information from two or more modalities for prediction. In addition, existing image matching methods all rely on visual images that are recognizable to the human eye and understandable to the human brain. However, the two-dimensional delay Doppler map (DDM) is a synthetic aperture radar altimeter (SARAL) that projects the three-dimensional terrain onto a two-dimensional plane, and people cannot directly and intuitively interpret terrain changes from the two-dimensional delay Doppler map (DDM).

[0024] It is difficult to match two-dimensional delay-Doppler maps (DDM) and digital elevation models (DEM) with existing solutions for the following two reasons: first, two-dimensional delay-Doppler maps (DDM) and digital elevation models (DEM) belong to different modes and cannot be directly matched; second, deep learning lacks formula guidance and has limited interpretation capabilities, so it is difficult to achieve mapping between digital elevation models (DEM) and two-dimensional delay-Doppler maps (DDM); third, there are no public data sets and data set generation methods on the market, and it is difficult to obtain sufficient training data sets.

[0025] In response to the practical needs of cross-modal terrain matching and real-time aircraft positioning, this application uses the two-dimensional delay Doppler map (DDM) formed by the synthetic aperture radar altimeter (SARAL) echo and the digital elevation model (DEM) to directly perform cross-modal matching, conducts in-depth research on neural network representation learning methods, studies elevation features that improve data representation capabilities, establishes a neural network model for explicit matching and positioning, establishes a loss function that introduces quantitative features, realizes the explicit derivation between the best matching position coordinates and terrain data, ensures that the objects of terrain matching have the same physical meaning, and the parameters can be quantified and evaluated, effectively improves the cross-modal two-dimensional delay Doppler map (DDM) terrain data matching accuracy, complex scene generalization ability and network interpretability, and provides reliable theoretical and technical support for practical scene matching and positioning.

[0026] like Figure 1 As shown, Figure 1 It is a flow chart of a cross-modal terrain matching and positioning method provided in an embodiment of the present application. The cross-modal terrain matching and positioning method is used for a flying platform, and the flying platform is equipped with an airborne synthetic aperture radar altimeter. The method includes the following steps 101 to 107.

[0027] Step 101, obtaining a target digital elevation model and a target two-dimensional delay Doppler map of the flight platform on the current terrain.

[0028] In the embodiment of the present application, the digital elevation model (DEM) is a digital representation of the elevation of the earth's surface, which records the elevation information at different locations in a discrete form. At the same time, digital elevation model (DEM) data is available worldwide. Therefore, the target digital elevation model of the flight platform in the current terrain can be directly obtained.

[0029] The two-dimensional delay Doppler map is obtained by performing a radar scan of the current terrain using an airborne synthetic aperture radar altimeter. The synthetic aperture radar altimeter transmits linear frequency modulated wide-beam pulses to the current terrain range in a near-vertical observation mode, and receives the echo signal after a certain delay, and obtains the corresponding target two-dimensional delay Doppler map based on the echo signal. The target two-dimensional delay Doppler map is the measured two-dimensional delay Doppler map, and its specific calculation method is the same as the calculation method of the measured two-dimensional delay Doppler map.

[0030] Step 102, respectively encode the target digital elevation model and the target two-dimensional delay Doppler map into a first coding space through a first encoder and a second encoder to obtain a first coding feature corresponding to the target digital elevation model and a second coding feature corresponding to the target two-dimensional delay Doppler map.

[0031] In an embodiment of the present application, encoder 1 (first encoder) is used to encode the target digital elevation model (DEM), which is gradually compressed to a 50-dimensional low-dimensional representation through multi-layer linear transformations and activation functions in encoder 1 to obtain a first encoding feature of the target digital elevation model, and encoder 2 (second encoder) is used to encode the target two-dimensional delay-Doppler map (DDM), and a second encoding feature of the target two-dimensional delay-Doppler map is obtained through multi-layer linear transformations and activation functions in encoder 2.

[0032] A linear transformation is performed on the output encoded data of encoder 1 and encoder 2 to obtain first encoded features and second encoded features of the same dimension. The first encoded features and second encoded features of the same dimension can be used for cross-modal fusion.

[0033] Step 103: In the first coding space, first feature data corresponding to the first coding feature is calculated, and second feature data corresponding to the second coding feature is calculated.

[0034] In an embodiment of the present application, the first feature data includes feature data of a target digital elevation model (DEM), and the second feature data includes feature data of a target two-dimensional delay Doppler map (DDM). Specifically, the first feature data may include a first mapping matrix, a second mapping matrix, and a third mapping matrix. The second feature data may include a fourth mapping matrix, a fifth mapping matrix, and a sixth mapping matrix.

[0035] Step 104: Based on the second feature data, a first cross-modal fusion vector corresponding to the first coding feature is calculated, and based on the first feature data, a second cross-modal fusion vector corresponding to the second coding feature is calculated.

[0036] In the embodiment of the present application, based on the second feature data, a first cross-modal fusion vector corresponding to the first coding feature is calculated, so that the first cross-modal fusion vector is fused with the feature data of the target two-dimensional delay Doppler map (DDM) on the basis of the first coding feature. Based on the first feature data, a second cross-modal fusion weight matrix corresponding to the second coding feature is calculated, so that the second cross-modal fusion vector is fused with the feature data of the target digital elevation model (DEM) on the basis of the second coding feature.

[0037] In this embodiment, since the two-dimensional delay Doppler map (DDM) has relatively rich low-level features, the feature fusion of the digital elevation model (DEM) and the two-dimensional delay Doppler map (DDM) is adopted as follows Figure 2 The feature fusion algorithm shown. Figure 2 It is a schematic diagram of a model architecture for a cross-modal terrain matching and positioning method provided in an embodiment of the present application, which illustrates a schematic diagram of feature fusion. This feature fusion utilizes the connection and interaction between the low-level features of each modality, has a low algorithm complexity, and only requires single model training. In an embodiment of the present application, an encoder-decoder structure can be used to establish a connection between the features of a two-dimensional delay Doppler map (DDM) and a digital elevation model (DEM) to obtain two cross-modal fusion vectors that fuse the feature information of the two modalities. Specifically, a first encoder structure is used to fuse the features of a digital elevation model (DEM) and a two-dimensional delay Doppler map (DDM) to obtain a first cross-modal fusion vector containing the features of both, and a second encoder structure is used to fuse the features of a digital elevation model (DEM) and a two-dimensional delay Doppler map (DDM) to obtain a second cross-modal fusion vector containing the features of both, and then a decoder structure is used to decode the fused cross-modal fusion vector, and finally a fused new feature vector containing the information of the other party is obtained. Therefore, the terrain matching and aircraft positioning algorithm proposed in this embodiment is suitable for non-visual and different modality data matching, and can specifically realize the matching of digital elevation model (DEM) and two-dimensional delay Doppler map (DDM).

[0038] Then, inspired by the dot product operation of the attention mechanism, the cross-modal fusion weight matrix is ​​calculated by the following formula:

[0039] in, Represents the dimension of the weight matrix. And use the calculated cross-modal fusion weight matrix to and Perform weighted summation to obtain the first cross-modal fusion vector after fusion and the second cross-modal fusion vector .

[0040] Step 105 , decoding the first cross-modal fusion vector and the second cross-modal fusion vector respectively by a first decoder and a second decoder to obtain a first decoding feature corresponding to the first cross-modal fusion vector and a second decoding feature corresponding to the second cross-modal fusion vector.

[0041] In an embodiment of the present application, the first decoder 1 and the second decoder 2 are respectively used to decode the first cross-modal fusion vector and the second cross-modal fusion vector to obtain the first decoded data fused with the target two-dimensional delay Doppler map and the target digital elevation model, and the second decoded data fused with the target digital elevation model and the target two-dimensional delay Doppler map. Specifically, the first decoder 1 and the second decoder 2 are both composed of a linear layer and an activation layer, and each linear layer is followed by an activation layer. The first decoder 1 obtains a first decoding feature of a size of 256*512 through 2 linear layers and 2 activation layers. The second decoder 2 obtains a second decoding feature of a size of 512*2048 through 2 linear layers and 2 activation layers.

[0042] Step 106 , extracting features from the first decoding feature and the second decoding feature respectively through different feature extraction modules to obtain a first feature vector of the target digital elevation model and a second feature vector of the target two-dimensional delay-Doppler map.

[0043] In an embodiment of the present application, network architecture 1 can be used to extract features from a first decoded feature that is a fusion of a target two-dimensional delay-Doppler map and a target digital elevation model to obtain a first feature vector, and network architecture 2 can be used to extract features from a second decoded feature that is a fusion of a target digital elevation model and a target two-dimensional delay-Doppler map to obtain a second feature vector.

[0044] Step 107: Determine the positioning data of the flight platform based on the first eigenvector and the second eigenvector.

[0045] In the embodiment of the present application, after feature extraction, a feature vector (first feature vector) of a reference digital elevation model (DEM) and a feature vector (second feature vector) of a real-time two-dimensional delay Doppler map (DDM) can be obtained respectively.

[0046] In order to measure the similarity between the feature vectors of the reference digital elevation model (DEM) and the real-time two-dimensional delay-Doppler map (DDM) belonging to different modes after feature extraction, it is used to obtain the terrain matching results, such as Figure 2As shown in the figure, cosine similarity is used to measure the similarity between the feature vectors of the digital elevation model (DEM) and the two-dimensional delay Doppler map (DDM). Different from the similarity measurement algorithm that directly solves the linear distance between two data in n-dimensional space, cosine similarity measures the difference between the two by solving the cosine value of the angle between two vectors in the vector space. It reflects the difference in direction in the vector space without considering the size of the vector. This feature is compatible with the low degree of cross-modality attention to the size of heterogeneous data.

[0047] In order to measure the similarity between the first eigenvector and the second eigenvector, cosine similarity is used for similarity measurement. Cosine similarity satisfies the following formula:

[0048] in, represents the cosine similarity between the first eigenvector and the second eigenvector, represents the first eigenvector, represents the second eigenvector, represents the coordinates of the target in the two-dimensional delay-Doppler map, Represents the coordinates of the target digital elevation model.

[0049] During the training process, the mean square error (MSE) loss and the contrast loss function are weighted as the global loss function. MSE loss is a commonly used regression loss measurement method in machine learning. It can represent the square of the absolute difference between the predicted value and the true value. The smaller the loss value, the more accurate the model prediction result. MSE loss satisfies the following formula:

[0050] Where N is the total number of eigenvectors of the first eigenvector, The second eigenvector, is the first eigenvector. MSE loss can also be used to evaluate the deviation between the positioning coordinates and the reference coordinates, expressed as follows:

[0051] Where M is the number of terrain coordinates, Indicates that based on the multiple second eigenvectors with the largest similarity, the candidate coordinates mapped in the target digital elevation model are retrieved in reverse. Indicates positioning coordinates.

[0052] The contrast loss function is used to evaluate the matching degree of the two-modal mapping matrices. The contrast loss is expressed as follows:

[0053] The label indicates are similar samples, that is, the first and second eigenvectors have the same labels. express are dissimilar samples. represents the Euclidean distance between two samples, is a hyperparameter that defines the minimum distance between sample pairs.

[0054] According to the established MSE loss and contrast loss, the global loss function can be modeled as:

[0055] in, is the MSE loss coefficient of the first eigenvector and the second eigenvector, is the contrast loss coefficient between the first eigenvector and the second eigenvector, is the MSE loss coefficient of the candidate coordinates and the positioning coordinates.

[0056] After the above steps are performed to forward propagate and obtain the first eigenvector and the second eigenvector as well as the candidate coordinates to calculate the global loss, the back propagation algorithm is used to calculate the gradient of the parameters layer by layer, and the optimizer is used to adjust the parameters of each layer according to the gradient. The global loss is back-propagated to the weights of each layer of the encoder, decoder, feature fusion module and feature extraction module to obtain the optimized first eigenvector , the second eigenvector , and the candidate coordinates mapped in the target digital elevation model after reverse retrieval of cosine similarity , used for a new round of global loss function calculation, and the above steps are repeated until the loss function converges.

[0057] In an embodiment of the present application, a target digital elevation model and a target two-dimensional delay Doppler map of the flight platform in the current terrain are obtained, and the two-dimensional delay Doppler map is obtained by performing a radar scan of the current terrain by an airborne synthetic aperture radar altimeter; the target digital elevation model and the target two-dimensional delay Doppler map are cross-modally fused to obtain fused matching data; different feature extraction modules are used to extract features from the fused matching data to obtain a first feature vector of the target digital elevation model and a second feature vector of the target two-dimensional delay Doppler map; based on the first feature vector and the second feature vector, the positioning data of the flight platform is determined. The present application fuses the features of a cross-modal digital elevation model (DEM) and a two-dimensional delay Doppler map (DDM), then performs feature extraction, and performs similarity measurement on the feature vectors after feature extraction, and obtains the positioning result according to the similarity score, which effectively solves the problems of aircraft positioning relying on the global positioning system (GPS), insufficient scene generalization capability, and heterogeneous gaps in cross-modal terrain matching.

[0058] It is understandable that in the specific implementation of this application, related data such as terrain data, flight data, digital elevation model data, etc. are involved. When the embodiments in this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data, as well as the training and use of various models need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0059] Optionally, the steps of respectively encoding the target digital elevation model and the target two-dimensional delay Doppler map into the first coding space through the first encoder and the second encoder to obtain a first coding feature corresponding to the target digital elevation model and a second coding feature corresponding to the target two-dimensional delay Doppler map specifically include: encoding the target digital elevation model through the first encoder to obtain a first coding feature map; and encoding the target two-dimensional delay Doppler map through the second encoder to obtain a second coding feature map; performing matrix multiplication calculations on the first coding feature map and the second coding feature map through unit matrices of different dimensions to obtain the first coding feature corresponding to the target digital elevation model and the second coding feature corresponding to the target two-dimensional delay Doppler map, the first coding feature and the second coding feature having the same dimension.

[0060] In the embodiments of the present application, Figure 2 As shown in the figure, encoder 1 (first encoder) is used to encode the target digital elevation model (DEM), which is gradually compressed to a low-dimensional representation of 50 dimensions through multi-layer linear transformation and activation functions in encoder 1 to obtain the first encoding feature of the target digital elevation model, and encoder 2 (second encoder) is used to encode the target two-dimensional delay Doppler map (DDM), and the second encoding feature of the target two-dimensional delay Doppler map is obtained through multi-layer linear transformation and activation functions in encoder 2. Specifically, encoder 1 and encoder 2 are both composed of linear layers and activation layers, and each linear layer is followed by an activation layer. Encoder 1 reduces the first encoded data from 512*256 to 256*50 through 2 linear layers and 2 activation layers. Encoder 2 reduces the second encoded feature from 2048*512 to 512*50 through 2 linear layers and 2 activation layers.

[0061] For the output encoded data of encoder 1 and encoder 2, assuming that the feature map output by encoder 1 is , the feature map output by encoder 2 is The output of encoder 1 is right-multiplied by a size of The identity matrix of The first coded feature The output of encoder 2 is multiplied by a value of The identity matrix is ​​multiplied by a matrix of size The identity matrix of The second coded feature Among them, the parameters It is manually adjustable.

[0062] Optionally, in the first coding space, the steps of calculating the first feature data corresponding to the first coding feature and calculating the second feature data corresponding to the second coding feature specifically include: for the first coding feature, linear mapping is performed through the first weight matrix to obtain a first mapping matrix corresponding to the first coding feature; for the second coding feature, linear mapping is performed through the second weight matrix and the third weight matrix to obtain a second mapping matrix and a third mapping matrix corresponding to the second coding feature; for the second coding feature, linear mapping is performed through the first weight matrix to obtain a fourth mapping matrix corresponding to the second coding feature; for the first coding feature, linear mapping is performed through the second weight matrix and the third weight matrix to obtain a sixth mapping matrix and a third mapping matrix corresponding to the fifth coding feature.

[0063] In the embodiment of the present application, the first characteristic data may include a first mapping matrix, a second mapping matrix, and a third mapping matrix. The second characteristic data may include a fourth mapping matrix, a fifth mapping matrix, and a sixth mapping matrix.

[0064] For the first encoding feature , through linear mapping Generate the first mapping matrix , for the second encoding feature , through linear mapping Generate the second mapping matrix And the third mapping matrix ; On the contrary, for the second coding feature , through linear mapping Generate the fourth mapping matrix , for the first coded feature , through linear mapping Generate the second mapping matrix And the sixth mapping matrix ,in, is the weight matrix.

[0065] Optionally, based on the second feature data, the step of calculating the first cross-modal fusion vector corresponding to the first coding feature, and based on the first feature data, the step of calculating the second cross-modal fusion weight matrix corresponding to the second coding feature specifically includes: based on the first mapping matrix and the second mapping matrix, obtaining the second cross-modal fusion weight matrix corresponding to the first coding feature; based on the fourth mapping matrix and the fifth mapping matrix, obtaining the first cross-modal fusion weight matrix corresponding to the second coding feature; performing matrix multiplication on the second cross-modal fusion weight matrix and the third mapping matrix, and normalizing the result of the matrix multiplication to obtain the first cross-modal fusion vector; and, performing matrix multiplication on the first cross-modal fusion weight matrix and the sixth mapping matrix, and normalizing the result of the matrix multiplication to calculate the second cross-modal fusion vector.

[0066] In the embodiment of the present application, after obtaining the first mapping matrix, the second mapping matrix, the third mapping matrix, the fourth mapping matrix, the fifth mapping matrix and the sixth mapping matrix, the cross-modal fusion weight matrix is ​​calculated by the following formula:

[0067] in, Represents the dimension of the weight matrix. And use the calculated cross-modal fusion weight matrix to and Perform weighted summation to obtain the first cross-modal fusion vector after fusion and the second cross-modal fusion vector .

[0068] The fused matrix and After normalization, it has zero mean and unit standard deviation, and the second standardized coding feature of the target two-dimensional delay-Doppler map is obtained, which is expressed as follows:

[0069] in, is the second cross-modal fusion vector obtained by fusing encoder 1 and encoder 2, yes The mean of yes The variance of .

[0070] The second standardized coding feature of the target two-dimensional delay-Doppler image is mean-adjusted to have the same distribution characteristics as the first coding feature, and a cross-modal fusion vector is obtained, which is expressed as follows:

[0071] in, is the second coded data The second standardized coding feature of the target two-dimensional delay-Doppler map represented by is the final cross-modal fusion vector, It is the first coded data obtained by combining encoder 1 and encoder 2 The mean of yes After standardization and mean adjustment, the first coding feature of the target digital elevation model obtained by encoder 1 and the second coding feature of the target two-dimensional delay-Doppler map obtained by encoder 2 are mapped to the same feature space, aligned in the feature space, and the cross-modal fusion vector of the first coding feature and the second standardized coding feature is obtained.

[0072] Optionally, the steps of respectively performing feature extraction on the first decoding feature and the second decoding feature through different feature extraction modules to obtain a first feature vector of the target digital elevation model and a second feature vector of the target two-dimensional delay Doppler map specifically include: performing feature extraction processing on the first decoding data through the first feature extraction module to obtain a first feature vector of the target digital elevation model, and the first feature vector corresponds to the first decoding data; performing feature extraction processing on the second decoding data through the second feature extraction module to obtain a second feature vector of the target two-dimensional delay Doppler map, and the second feature vector corresponds to the second decoding data.

[0073] In the embodiment of the present application, the near-vertical observation mode of the synthetic aperture radar altimeter causes the nadir point echo to be coupled with the surrounding scattering point echo, that is, near-vertical coupling, which makes the two-dimensional delay Doppler maps (DDM) of adjacent apertures highly similar and the image quality low, making it difficult to effectively distinguish the two-dimensional delay Doppler maps (DDM) of adjacent apertures, which brings challenges to correct terrain matching and high-precision aircraft positioning. Therefore, a suitable network architecture and loss function are selected to effectively extract features from the fused data.

[0074] like Figure 2 The middle part shown, from Figure 2 It can be seen that different network architectures are used to extract features from the first cross-modal fusion vector of the digital elevation model (DEM) and the second cross-modal fusion vector of the two-dimensional delay Doppler map (DDM) that fuse each other's information, so that the global, local and fine features of the digital elevation model (DEM) and the two-dimensional delay Doppler map (DDM) can be efficiently extracted, and the digital elevation models (DEM) at different locations can be effectively distinguished, and the two-dimensional delay Doppler maps (DDM) with different apertures can be effectively distinguished, laying a solid foundation for subsequent terrain matching.

[0075] The network architecture 1 can be used to extract the first decoding features that are fused with the target two-dimensional delay-Doppler map and the target digital elevation model to obtain a first feature vector, and the network architecture 2 can be used to extract the second decoding features that are fused with the target digital elevation model and the target two-dimensional delay-Doppler map to obtain a second feature vector.

[0076] Specifically, the network architecture 1 has 7 convolutional layers, followed by a maximum pooling layer for feature map downsampling, and finally the feature map is converted into a one-dimensional feature vector by flattening. Among them, for the 7 convolutional layers, the convolution kernel size of the 1st-2nd layers is 3×3, the number of output channels is 32, and the step size is 1; the convolution kernel size of the 3rd layer is 3×3, the number of output channels is 64, and the step size is 2; the convolution kernel size of the 4th layer is 3×3, the number of output channels is 64, and the step size is 1; the convolution kernel size of the 5th layer is 3×3, the number of output channels is 128, and the step size is 2; the convolution kernel size of the 6th layer is 3×3, the number of output channels is 128, and the step size is 1; the convolution kernel size of the 7th layer is 8×8, the number of output channels is 128, and the step size is 1. The network structure 2 is ResNet18, which consists of 18 weighted layers including convolutional layers and fully connected layers, including an initial convolutional layer, followed by 16 convolutional layers, and then a fully connected layer. The initial convolutional layer has a convolution kernel size of 7×7, 64 output channels, and a stride of 2. For 16 convolutional layers, the convolution kernel size of the 1st to 4th layers is 3×3, the number of output channels is 64, and the stride is 1; the convolution kernel size of the 5th layer is 3×3, the number of output channels is 128, and the stride is 2; the convolution kernel size of the 6th to 8th layers is 3×3, the number of output channels is 128, and the stride is 1; the convolution kernel size of the 9th layer is 3×3, the number of output channels is 256, and the stride is 2; the convolution kernel size of the 10th to 12th layers is 3×3, the number of output channels is 256, and the stride is 1; the convolution kernel size of the 13th layer is 3×3, the number of output channels is 512, and the stride is 2; the convolution kernel size of the 14th to 16th layers is 3×3, the number of output channels is 128, and the stride is 1.

[0077] Optionally, the step of determining the positioning data of the flight platform based on the first eigenvector and the second eigenvector specifically includes: calculating the cosine similarity between the first eigenvector and the second eigenvector; and determining the positioning data of the flight platform based on the cosine similarity.

[0078] In the embodiment of the present application, the feature vector of the reference digital elevation model (DEM) and the feature vector of the real-time two-dimensional delay Doppler map (DDM) can be obtained respectively. In order to measure the similarity between the feature vectors of the reference digital elevation model (DEM) and the real-time two-dimensional delay Doppler map (DDM) belonging to different modes after feature extraction, the terrain matching result is obtained, such as Figure 4The figure shows a schematic diagram of similarity measurement, using cosine similarity to measure the similarity between the feature vectors of the digital elevation model (DEM) and the two-dimensional delay Doppler map (DDM). Different from the similarity measurement algorithm that directly solves the linear distance between two data in n-dimensional space, cosine similarity measures the difference between the two by solving the cosine value of the angle between the two vectors in the vector space. It reflects the difference in direction in the vector space without considering the size of the vector. This feature is compatible with the low degree of cross-modality attention to the size of heterogeneous data.

[0079] Optionally, the step of calculating the cosine similarity between the first eigenvector and the second eigenvector specifically includes: Assumptions is the real-time two-dimensional delay-Doppler map (DDM) feature vector, is the reference digital elevation model (DEM) feature vector, the specific calculation process of the cosine similarity between the two is shown in the following expression (1): (1) in, represents the cosine similarity between the first eigenvector and the second eigenvector, represents the first eigenvector, represents the second eigenvector, represents the coordinates of the target in the two-dimensional delay-Doppler map, represents the coordinates of the target digital elevation model. The × operator is the inner product operator. From expression (1), we can see that the smaller the angle between the two vectors, the larger the cosine value, and the higher the vector similarity. According to the cosine similarity, the similarity score between each pair of digital elevation model (DEM) and two-dimensional delay Doppler map (DDM) is obtained for subsequent aircraft positioning.

[0080] Optionally, the step of determining the positioning data of the flight platform based on cosine similarity specifically includes: for each position coordinate in the target two-dimensional delay Doppler map, based on the second eigenvector, selecting multiple second eigenvectors with the greatest similarity; based on the multiple second eigenvectors with the greatest similarity, reversely retrieving candidate coordinates mapped in the target digital elevation model, each candidate coordinate corresponds to a second eigenvector; based on the multiple candidate coordinates, determining the positioning data of the flight platform.

[0081] In the embodiment of the present application, after obtaining the similarity score between each pair of digital elevation models (DEM) and two-dimensional delay Doppler maps (DDM), the similarity scores are sorted in descending order to obtain a descending index. The top three reference digital elevation models (DEMs) with the similarity scores to the real-time two-dimensional delay Doppler map (DDM) are selected according to the similarity scores, and the coordinates mapped by these digital elevation models (DEMs) are obtained by reverse retrieval.

[0082] Optionally, the step of determining the positioning data of the flight platform based on multiple candidate coordinates specifically includes: performing weighted calculation on the multiple candidate coordinates to obtain weighted coordinate values; and determining the positioning data of the flight platform based on the weighted coordinate values.

[0083] In an embodiment of the present application, the coordinates mapped by these digital elevation models (DEMs) are obtained by reverse retrieval. To improve the credibility of the coordinate positioning results, the three positioning coordinates with the highest similarity scores are weighted averaged in engineering applications to determine the aircraft positioning results.

[0084] Optionally, the method also includes: obtaining a simulation data set and a measured data set, the simulation data set including a simulated two-dimensional delay-Doppler map obtained based on a digital elevation model calculation, and the measured data set including a measured two-dimensional delay-Doppler map obtained in actual measurement based on an airborne synthetic aperture radar altimeter; based on the simulation data set and the measured data set, performing a simulation test on the positioning data of the flight platform.

[0085] In the embodiment of the present application, the simulation conditions are set as follows:

[0086] Figure 3 The simulated two-dimensional delay Doppler map (DDM) results were obtained using matlab software.

[0087] Optionally, the step of obtaining a simulation data set specifically includes: obtaining simulation data, the simulation data including the current position of the flying platform, the flight speed of the platform, and the coordinates of each scattering point in the digital elevation model; obtaining a position vector of each scattering point based on the current position of the platform and the coordinates of the scattering point; calculating the Doppler frequency of each scattering point based on the flight speed of the platform, the position vector of each scattering point, and the radar wavelength of an airborne synthetic aperture radar altimeter; determining a relative distance between the coordinates of each scattering point and the current position of the platform based on the current position of the platform and the coordinates of the scattering point; determining a reflection of each scattering point based on the Doppler frequency of each scattering point, the relative distance between the coordinates of each scattering point and the current position of the platform, and a backscattering coefficient; obtaining a simulated two-dimensional delay Doppler map based on the Doppler frequency of each scattering point and the reflection of each scattering point; and constructing a simulation data set based on the simulated two-dimensional delay Doppler map.

[0088] In the embodiment of the present application, the simulation data set is specifically constructed as follows: High-precision two-dimensional delay Doppler map (DDM) training data is the basis for terrain matching and aircraft positioning. The coordinates of each scattering point can be obtained according to the digital elevation model (DEM). , the current position of the platform is , the flight speed is Although the synthetic aperture radar altimeter (SARAL) observes the scattering point at the nadir point, its energy is mainly based on backscattering, and the position vector can be obtained as follows (2): (2) in, is the current position of the platform, are the coordinates of each scattering point, and A collection of scattered points.

[0089] The Doppler frequency of each scattering point is expressed as follows (3): (3) in, is the real-time speed of the platform, is the radar wavelength. The relative distance of each scattering point is given by expression (3), and the backscatter coefficient is given by expression (4):

[0090] (4) (5) in, ,and Depends on the type of land cover medium. For expressions (4) and (5), the reflectance of a single scattering point can be obtained as follows (6): (6) Each element of the two-dimensional delay-Doppler map (DDM) matrix is ​​accumulated according to the distance and Doppler index set to the following expression (7): (7) in, It is A set of range gate indices, It is After the above processing, the two-dimensional delay-Doppler map (DDM) power matrix can be obtained as ,in is the number of pulses, is the number of range gates. Therefore, sufficient reference two-dimensional delay-Doppler maps (DDMs) based on digital elevation models (DEMs) can be obtained, and the simulated two-dimensional delay-Doppler map (DDM) imaging results are as follows: Figure 3 shown.

[0091] Optionally, the steps of measuring the data set specifically include: controlling the flight platform equipped with an airborne synthetic aperture radar altimeter to fly along the X-axis direction at a preset altitude and fixed speed, and obtaining the coordinates of the ground scattering point; when the airborne synthetic aperture radar altimeter passes the target scattering point, emitting a linear frequency modulated wide-beam pulse, and receiving the echo signal after a certain delay; after downsampling and distance compressing the echo signal, a baseband echo signal is obtained, and the baseband echo signal includes an azimuth slow time; based on the azimuth frequency modulation rate, a linear relationship between the Doppler frequency and the azimuth slow time is given, and a Fourier transform is performed in the azimuth direction to transform the baseband echo signal into a delay Doppler domain; Fourier transform is performed on the delay Doppler domain again to obtain an echo signal in a two-dimensional frequency domain; a phase function for realizing distance migration correction is multiplied with the echo signal in the two-dimensional frequency domain to obtain a measured two-dimensional delay Doppler map.

[0092] In the embodiment of the present application, the measured two-dimensional delay Doppler map (DDM) data set is constructed as follows: an airborne synthetic aperture radar altimeter (SARAL) is used to measure the altitude of the and fixed speed Fly along the X-axis. The coordinates of the ground scattering point are ,in The coordinates of the airborne radar position B are , which is located above point L. The synthetic aperture radar altimeter (SARAL) can obtain a high-resolution two-dimensional delay Doppler map (DDM) at the L position. When the airborne synthetic aperture radar altimeter (SARAL) passes through point L, it transmits a linear frequency modulated wide-beam pulse and receives the echo after a certain delay. The received echo can be expressed as the following expression (8): (8) in, It is the distance envelope after distance compression, often expressed as a Sinc function. is the center frequency. It is the distance to fast time, It is the azimuthally slow time. It is the frequency modulation. is the speed of light, is a constant representing the back-propagation coefficient, represents the phase change of the radar signal caused by the surface scattering process, is the instantaneous distance between the synthetic aperture radar altimeter (SARAL) and the ground scattering point. After downsampling and distance compression, the baseband echo signal is expressed as follows (9): (9) in is a complex domain constant and is ignored in subsequent derivations; It is the distance envelope after distance compression, often expressed as a Sinc function. The instantaneous distance of the scattering point The following expression (10): (10) in, is the reference distance. Substituting expression (9) into expression (8), the baseband signal of the target can be obtained as follows (11): (11) in, is an imaginary unit, satisfying The azimuth phase modulation can be seen from the second exponential term in expression (10). Since the phase is The azimuth frequency modulation is a function of , and the azimuth frequency modulation is expressed as follows (12): (12) Where the azimuth modulation frequency gives the Doppler frequency and slow time The linear relationship between them is as follows (13): (13) Next, Fourier transform is performed in the azimuth direction to transform expression (10) into the delay Doppler domain as follows (14): (14) in, It is distance migration, is the azimuth antenna pattern frequency domain form.

[0093] By performing Fourier transform on expression (14), we can obtain the echo in the two-dimensional frequency domain as follows (15): (15) in, , is the range-to-frequency domain, yes frequency domain form.

[0094] To achieve range migration correction, a phase function can be obtained as follows (16): (16) By multiplying expression (16) with expression (14), the range compression signal can be obtained as follows (17): (17) According to the processing of expressions (8) to (17), the two-dimensional delay Doppler map (DDM) can be obtained from the synthetic aperture radar altimeter (SARAL) echo to construct a measured data set to meet the training, verification and testing requirements. The measured two-dimensional delay Doppler map (DDM) imaging results are as follows: Figure 4 shown.

[0095] The method of this embodiment can be used to perform cross-modal terrain matching and real-time aircraft positioning. In order to verify the feasibility of the proposed cross-modal network in terrain matching and the effect of real-time aircraft positioning, a simulation and measured data set was constructed; then feature fusion, feature extraction, and similarity measurement were performed using the method of this embodiment, and finally the aircraft positioning result was obtained. Figure 5-Figure 6 shown. Figure 5 It is a regional digital elevation model (DEM) and a schematic diagram of a certain track, where the red line represents the flight direction of the aircraft and the letter v represents the velocity vector. Figure 6 This is the result of terrain matching and real-time aircraft positioning using this application, where the black dotted line represents the preset track, the blue circle represents the point to be matched, the red triangle represents the real coordinates of the aircraft, and the red cross represents the positioning result obtained by this application. It can be seen that the method of this application can achieve cross-modal digital elevation model (DEM) and two-dimensional delay Doppler map (DDM) terrain matching, and then can achieve real-time aircraft positioning, and the positioning result is reliable.

[0096] Finally, it should be noted that the above specific implementation methods are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred implementation methods, a person of ordinary skill in the art should understand that the technical solution of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present application, which should be included in the scope of the claims of the present application.

Claims

1. A cross-modal terrain matching positioning method, characterized in that: Used for a flying platform, the flying platform being equipped with an airborne synthetic aperture radar altimeter, the method comprising: Acquire a target digital elevation model and a target two-dimensional delay Doppler map of the flight platform in the current terrain, wherein the two-dimensional delay Doppler map is obtained by performing a radar scan of the current terrain by the airborne synthetic aperture radar altimeter; The target digital elevation model and the target two-dimensional delay-Doppler map are respectively encoded into a first coding space by a first encoder and a second encoder to obtain a first coding feature corresponding to the target digital elevation model and a second coding feature corresponding to the target two-dimensional delay-Doppler map; In the first coding space, first feature data corresponding to the first coding feature is calculated, and second feature data corresponding to the second coding feature is calculated; Based on the second feature data, a first cross-modal fusion vector corresponding to the first coding feature is calculated; based on the first feature data, a second cross-modal fusion vector corresponding to the second coding feature is calculated; Decoding the first cross-modal fusion vector and the second cross-modal fusion vector respectively through a first decoder and a second decoder to obtain a first decoding feature corresponding to the first cross-modal fusion vector and a second decoding feature corresponding to the second cross-modal fusion vector; The first decoding feature and the second decoding feature are respectively extracted by different feature extraction modules to obtain a first feature vector of the target digital elevation model and a second feature vector of the target two-dimensional delay-Doppler map; Based on the first feature vector and the second feature vector, positioning data of the flying platform is determined.

2. The cross-modal terrain matching positioning method according to claim 1, characterized in that: The step of respectively encoding the target digital elevation model and the target two-dimensional delay-Doppler map into a first coding space by using a first encoder and a second encoder to obtain a first coding feature corresponding to the target digital elevation model and a second coding feature corresponding to the target two-dimensional delay-Doppler map specifically includes: The target digital elevation model is encoded by a first encoder to obtain a first encoding feature map; And, encoding the target two-dimensional delay-Doppler map by a second encoder to obtain a second encoding feature map; The first coding feature map and the second coding feature map are respectively subjected to matrix multiplication calculations using unit matrices of different dimensions to obtain a first coding feature corresponding to the target digital elevation model and a second coding feature corresponding to the target two-dimensional delay-Doppler map, wherein the first coding feature and the second coding feature have the same dimension.

3. The cross-modal terrain matching positioning method according to claim 1, characterized in that: The step of calculating the first feature data corresponding to the first coding feature and the step of calculating the second feature data corresponding to the second coding feature in the first coding space specifically include: For the first coding feature, linear mapping is performed through the first weight matrix to obtain a first mapping matrix corresponding to the first coding feature; for the second coding feature, linear mapping is performed through the second weight matrix and the third weight matrix to obtain a second mapping matrix and a third mapping matrix corresponding to the second coding feature; For the second coding feature, linear mapping is performed through the first weight matrix to obtain a fourth mapping matrix corresponding to the second coding feature. For the first coding feature, linear mapping is performed through the second weight matrix and the third weight matrix to obtain a sixth mapping matrix and a third mapping matrix corresponding to the fifth coding feature.

4. The cross-modal terrain matching positioning method according to claim 3, characterized in that: The step of calculating a first cross-modal fusion vector corresponding to the first coding feature based on the second feature data, and calculating a second cross-modal fusion weight matrix corresponding to the second coding feature based on the first feature data specifically includes: Based on the first mapping matrix and the second mapping matrix, obtaining a second cross-modal fusion weight matrix corresponding to the first encoding feature; Based on the fourth mapping matrix and the fifth mapping matrix, obtaining a first cross-modal fusion weight matrix corresponding to the second encoding feature; Performing matrix multiplication on the second cross-modal fusion weight matrix and the third mapping matrix, and normalizing the matrix multiplication result to obtain a first cross-modal fusion vector; And, matrix multiplication is performed on the first cross-modal fusion weight matrix and the sixth mapping matrix, and the matrix multiplication result is normalized to calculate and obtain a second cross-modal fusion vector.

5. The cross-modal terrain matching positioning method according to claim 1, characterized in that: The step of extracting features from the first decoding feature and the second decoding feature respectively through different feature extraction modules to obtain a first feature vector of the target digital elevation model and a second feature vector of the target two-dimensional delay-Doppler map specifically includes: Performing feature extraction processing on the first decoded data by a first feature extraction module to obtain a first feature vector of the target digital elevation model, wherein the first feature vector corresponds to the first decoded data; The second decoded data is subjected to feature extraction processing by a second feature extraction module to obtain a second feature vector of the target two-dimensional delay-Doppler image, wherein the second feature vector corresponds to the second decoded data.

6. The cross-modal terrain matching positioning method according to claim 1, characterized in that: The step of determining the positioning data of the flight platform based on the first feature vector and the second feature vector specifically includes: Calculating the cosine similarity between the first feature vector and the second feature vector; Based on the cosine similarity, the positioning data of the flying platform is determined.

7. The cross-modal terrain matching positioning method according to claim 6, characterized in that: The step of calculating the cosine similarity between the first feature vector and the second feature vector specifically includes: The cosine similarity satisfies the following formula: in, represents the cosine similarity between the first eigenvector and the second eigenvector, represents the first eigenvector, represents the second eigenvector, represents the coordinates of the target in the two-dimensional delay Doppler map, Represents the coordinates of the target digital elevation model.

8. The cross-modal terrain matching positioning method according to claim 7, characterized in that: Based on the cosine similarity, the step of determining the positioning data of the flight platform specifically includes: For each position coordinate in the target two-dimensional delay-Doppler image, based on the second feature vector, selecting a plurality of the second feature vectors with the greatest similarity; Based on the plurality of second feature vectors with the greatest similarity, reversely retrieve candidate coordinates mapped in the target digital elevation model, each candidate coordinate corresponding to one second feature vector; Based on the multiple candidate coordinates, the positioning data of the flying platform is determined.

9. The cross-modal terrain matching positioning method according to claim 8, characterized in that: The step of determining the positioning data of the flight platform based on the plurality of candidate coordinates specifically includes: Performing weighted calculation on the plurality of candidate coordinates to obtain weighted coordinate values; Based on the weighted coordinate values, the positioning data of the flying platform is determined.

Images