Near-field sparse imaging method based on spatial convolution

Abstract

The invention relates to the technical field of millimeter wave imaging, in particular to a near-field sparse imaging method based on spatial convolution, and the method comprises the steps: obtaining the three-dimensional echo data of a to-be-measured target, and carrying out the sparse sampling of the three-dimensional echo data, and obtaining the sparse echo data; respectively imaging the three-dimensional echo data and the sparse echo data to obtain a corresponding three-dimensional imaging result and a sparse imaging result; predicting a corresponding predicted three-dimensional imaging result by using a pre-constructed spatial convolutional network; and obtaining a corresponding loss value by using a pre-constructed loss function, and obtaining a final spatial convolutional network under the condition that the loss value is smaller than a certain threshold value so as to predict a final three-dimensional imaging result. Therefore, the problems that in the related technology, design complexity and operation burden are high depending on echo data, and operation efficiency is reduced are solved; the matched filtering technology is limited in actual use scene and difficult to realize; the deep learning imaging effect is unstable, and even target information cannot be accurately reconstructed.

Description

Technical Field

[0001] The present application relates to the field of millimeter wave imaging technology, and in particular to a near-field sparse imaging method based on spatial convolution. Background Art

[0002] Millimeter-wave imaging technology, an emerging non-contact human inspection method, is widely used in security checks at key public safety sites such as airports and subways. It can effectively identify dangerous items hidden under clothing. However, traditional millimeter-wave imaging methods often face problems such as low imaging efficiency and complex data processing. In near-field imaging, balancing high resolution with low data volume is a major challenge.

[0003] In related technologies, high-resolution results can be obtained by combining large amounts of fully sampled echo data with matched filtering technology; or end-to-end deep learning can be adopted to automatically learn features from the data to form high-quality images.

[0004] However, in related technologies, over-reliance on echo data increases the complexity of system design and the computational burden, reducing the system's operating efficiency; matched filtering technology needs to meet strict spatial sampling and bandwidth requirements, making system implementation difficult and limiting practical scenarios; deep learning relies solely on training data for model training when imaging data is difficult to obtain, resulting in unstable imaging effects and even the inability to accurately reconstruct target information in some cases, which urgently needs improvement. Summary of the Invention

[0005] This application provides a near-field sparse imaging method based on spatial convolution to solve the problems in related technologies, such as reliance on echo data, high design complexity and computational burden, which reduces operating efficiency; matched filtering technology has limited practical application scenarios and is difficult to implement; deep learning imaging effects are unstable and may even be unable to accurately reconstruct target information.

[0006] The first aspect of the present application provides a near-field sparse imaging method based on spatial convolution, which is applied to the network training stage, wherein the method includes the following steps: obtaining three-dimensional echo data of the target to be measured, and sparsely sampling the three-dimensional echo data to obtain corresponding sparse echo data; imaging the three-dimensional echo data and the sparse echo data respectively to obtain corresponding three-dimensional imaging results and sparse imaging results; inputting the sparse imaging results into a pre-constructed spatial convolution network to predict the predicted three-dimensional imaging results corresponding to the sparse imaging results; substituting the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results into a pre-constructed loss function to obtain a corresponding loss value, and when the loss value is less than a preset threshold, obtaining a final spatial convolution network based on the loss value, the three-dimensional echo data, the three-dimensional imaging result, the sparse imaging result and the predicted three-dimensional imaging result, so as to use the spatial convolution network to predict the final three-dimensional imaging result of the three-dimensional echo data.

[0007] Optionally, in one embodiment of the present application, before inputting the sparse imaging result into a pre-constructed spatial convolutional network, it also includes: based on the real data of the sparse imaging result, in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the real data, constructing a real spatial convolution network of the spatial convolution network; based on the imaginary data of the sparse imaging result, in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the imaginary data, constructing an imaginary spatial convolution network of the spatial convolution network; based on the real spatial convolutional network and the imaginary spatial convolutional network, constructing the spatial convolutional network.

[0008] Optionally, in one embodiment of the present application, before substituting the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result into a pre-constructed loss function, it also includes: constructing an imaging loss function and an image loss function based on the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result, respectively; determining the imaging loss coefficient corresponding to the imaging loss function and the image loss coefficient corresponding to the image loss function based on the imaging loss function and the image loss function; constructing a loss function based on the imaging loss function, the image loss function, the imaging loss coefficient and the image loss coefficient.

[0009] Optionally, in one embodiment of the present application, an imaging loss function and an image loss function are respectively constructed based on the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result, including: projecting the three-dimensional imaging result to obtain a corresponding two-dimensional projection image; projecting the predicted three-dimensional imaging result to obtain a corresponding predicted two-dimensional projection image; and constructing the image loss function based on the two-dimensional projection image and the predicted two-dimensional projection image.

[0010] Optionally, in one embodiment of the present application, an imaging loss function and an image loss function are respectively constructed based on the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result, including: restoring the predicted three-dimensional imaging result to obtain corresponding imaging echo data; and constructing the imaging loss function based on the three-dimensional echo data and the imaging echo data.

[0011] Optionally, in one embodiment of the present application, before sparse sampling the three-dimensional echo data, the method further includes: correcting the three-dimensional echo data using preset data to obtain corrected three-dimensional echo data.

[0012] The second aspect of the present application provides a near-field sparse imaging method based on spatial convolution, which is applied to the network application stage, wherein the method includes the following steps: obtaining actual three-dimensional echo data of the actual target to be measured, and sparsely sampling the actual three-dimensional echo data to obtain corresponding actual sparse echo data; imaging the actual three-dimensional echo data and the actual sparse echo data respectively to obtain corresponding actual three-dimensional imaging results and actual sparse imaging results; inputting the actual sparse imaging results into a pre-trained spatial convolution network to predict the final three-dimensional imaging result corresponding to the actual three-dimensional echo data, wherein the pre-trained spatial convolution network is trained by the loss function corresponding to the actual three-dimensional echo data, the actual three-dimensional imaging result and the actual sparse imaging result.

[0013] In a third aspect, an embodiment of the present application provides a near-field sparse imaging device based on spatial convolution, which is applied to a network training stage, wherein the device includes: a first acquisition module for acquiring three-dimensional echo data of a target to be measured and sparsely sampling the three-dimensional echo data to obtain corresponding sparse echo data; a first imaging module for imaging the three-dimensional echo data and the sparse echo data respectively to obtain corresponding three-dimensional imaging results and sparse imaging results; a first prediction module for inputting the sparse imaging result into a pre-constructed spatial convolution network to predict a predicted three-dimensional imaging result corresponding to the sparse imaging result; a generation module for substituting the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result into a pre-constructed loss function to obtain a corresponding loss value, and when the loss value is less than a preset threshold, a final spatial convolution network is obtained based on the loss value, the three-dimensional echo data, the three-dimensional imaging result, the sparse imaging result and the predicted three-dimensional imaging result, so as to use the spatial convolution network to predict the final three-dimensional imaging result of the three-dimensional echo data.

[0014] Optionally, in one embodiment of the present application, it also includes: a first construction module for constructing a real spatial convolution network of the spatial convolution network based on the real data of the sparse imaging result, in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the real data, before inputting the sparse imaging result into a pre-constructed spatial convolution network; a second construction module for constructing an imaginary spatial convolution network of the spatial convolution network based on the imaginary data of the sparse imaging result, in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the imaginary data; a third construction module for constructing the spatial convolution network based on the real spatial convolution network and the imaginary spatial convolution network.

[0015] Optionally, in one embodiment of the present application, it also includes: a fourth construction module, which is used to construct an imaging loss function and an image loss function based on the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result before substituting the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result into a pre-constructed loss function; a determination module, which is used to determine the imaging loss coefficient corresponding to the imaging loss function and the image loss coefficient corresponding to the image loss function based on the imaging loss function and the image loss function; and a fifth construction module, which is used to construct a loss function based on the imaging loss function, the image loss function, the imaging loss coefficient and the image loss coefficient.

[0016] Optionally, in one embodiment of the present application, the fourth construction module includes: a first projection unit, used to project the three-dimensional imaging result to obtain a corresponding two-dimensional projection image; a second projection unit, used to project the predicted three-dimensional imaging result to obtain a corresponding predicted two-dimensional projection image; and a first construction unit, used to construct the projection image loss function based on the two-dimensional projection image and the predicted two-dimensional projection image.

[0017] Optionally, in one embodiment of the present application, the fourth construction module includes: a restoration unit for restoring the predicted three-dimensional imaging result to obtain corresponding imaging echo data; and a second construction unit for constructing the imaging loss function based on the three-dimensional echo data and the imaging echo data.

[0018] Optionally, in one embodiment of the present application, the method further includes: a correction module configured to correct the three-dimensional echo data using preset data before sparse sampling the three-dimensional echo data to obtain corrected three-dimensional echo data.

[0019] The fourth aspect of the present application provides a near-field sparse imaging device based on spatial convolution, which is applied to the network application stage, wherein the device includes: a second acquisition module, used to acquire actual three-dimensional echo data of the actual target to be measured, and sparsely sample the actual three-dimensional echo data to obtain corresponding actual sparse echo data; a second imaging module, used to image the actual three-dimensional echo data and the actual sparse echo data respectively to obtain corresponding actual three-dimensional imaging results and actual sparse imaging results; a second prediction module, used to input the actual sparse imaging result into a pre-trained spatial convolution network to predict the final three-dimensional imaging result corresponding to the actual three-dimensional echo data, wherein the pre-trained spatial convolution network is trained by the loss function corresponding to the actual three-dimensional echo data, the actual three-dimensional imaging result and the actual sparse imaging result.

[0020] The fifth aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the near-field sparse imaging method based on spatial convolution as described in the above embodiment.

[0021] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the program is executed by a processor, the near-field sparse imaging method based on spatial convolution is implemented.

[0022] The seventh aspect of the present application provides a computer program product, including a computer program, which, when executed, implements the above-mentioned near-field sparse imaging method based on spatial convolution.

[0023] The embodiment of the present application can obtain the three-dimensional echo data of the target to be measured, and perform sparse sampling on the three-dimensional echo data to obtain the corresponding sparse echo data, and respectively image the three-dimensional echo data and the sparse echo data to obtain the corresponding three-dimensional imaging results and sparse imaging results, and then use the pre-constructed spatial convolution network to predict the corresponding predicted three-dimensional imaging results, and substitute the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results into the pre-constructed loss function to obtain the corresponding loss value, and when the loss value is less than a certain threshold, obtain the final spatial convolution network, and then predict the final three-dimensional imaging result. While reducing the amount of data and computational complexity, it can fully utilize the spatial features in millimeter wave imaging to improve the imaging effect, and when processing targets with complex structures, it can more efficiently and accurately restore the target information and achieve high-quality imaging results. Thus, the problems in the related technology that the design complexity and computational burden of relying on echo data are high, which reduces the operating efficiency; the actual use scenarios of matched filtering technology are limited and difficult to implement; the deep learning imaging effect is unstable, and it may even be impossible to accurately reconstruct the target information.

[0024] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 A flowchart of a near-field sparse imaging method based on spatial convolution according to an embodiment of the present application; Figure 2 A schematic diagram of the structure of a spatial convolutional network provided according to one embodiment of the present application; Figure 3 A flowchart of calculating a loss value according to an embodiment of the present application is provided; Figure 4 A flowchart of the working principle of a near-field sparse imaging method based on spatial convolution according to an embodiment of the present application; Figure 5 Schematic diagram of a block diagram of a near-field sparse imaging device based on spatial convolution according to an embodiment of the present application; Figure 6 A flowchart of a near-field sparse imaging method based on spatial convolution according to another embodiment of the present application; Figure 7 A schematic diagram showing a comparison of imaging effects provided according to an embodiment of the present application; Figure 8 Schematic block diagram of a near-field sparse imaging device based on spatial convolution according to another embodiment of the present application; Figure 9 A schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] The following describes in detail embodiments of the present application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0027] The following describes a near-field sparse imaging method based on spatial convolution according to an embodiment of the present application with reference to the accompanying drawings. In response to the problems mentioned in the above background technology, such as the high design complexity and computational burden of relying on echo data, which reduces operating efficiency; the limited actual use scenarios of matched filtering technology and the difficulty in implementation; the unstable deep learning imaging effect, and even the inability to accurately reconstruct the target information, the present application provides a near-field sparse imaging method based on spatial convolution. In this method, three-dimensional echo data of the target to be measured can be obtained, and the three-dimensional echo data can be sparsely sampled to obtain corresponding sparse echo data, and the three-dimensional echo data and the sparse echo data can be imaged respectively to obtain corresponding three-dimensional imaging results and sparse imaging results, and then the corresponding predicted three-dimensional imaging results are predicted using a pre-constructed spatial convolution network, and the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results are substituted into the pre-constructed loss function to obtain the corresponding loss value, and when the loss value is less than a certain threshold, the final spatial convolution network is obtained, and then the final three-dimensional imaging result is predicted. While reducing the amount of data and computational complexity, it can fully utilize the spatial features in millimeter wave imaging to improve the imaging effect, and when processing targets with complex structures, it can more efficiently and accurately restore the target information to achieve high-quality imaging results. This solves the problems in related technologies, such as the high design complexity and computational burden of relying on echo data, which reduces operating efficiency; the limited actual usage scenarios of matched filtering technology, which makes it difficult to implement; and the unstable deep learning imaging effect, which may even be unable to accurately reconstruct target information.

[0028] Before introducing the near-field sparse imaging method based on spatial convolution proposed in the embodiment of the present application, the symbols involved in the embodiment of the present application are explained first. Specifically, Table 1 is a schematic table of symbol parameters provided according to an embodiment of the present application.

[0029] Table 1

[0030] Specifically, Figure 1 This is a flowchart of a near-field sparse imaging method based on spatial convolution provided according to an embodiment of the present application.

[0031] like Figure 1 As shown, the near-field sparse imaging method based on spatial convolution is applied to the network training stage, wherein the method includes the following steps: In step S101 , three-dimensional echo data of a target to be measured is acquired, and sparse sampling is performed on the three-dimensional echo data to obtain corresponding sparse echo data.

[0032] It is understandable that in the embodiments of the present application, the target to be measured may include but is not limited to a human body to be measured, an object to be measured, etc., and the present application does not impose any specific restrictions.

[0033] In some embodiments, the embodiments of the present application can collect three-dimensional echo data of the human body and the object to be tested from a laboratory environment, or can use an actual millimeter-wave human body security detector to collect three-dimensional echo data of the human body and the object to be tested. The specific settings can be made by technicians in this field according to actual conditions, and this application does not impose any specific restrictions.

[0034] Furthermore, the embodiment of the present application can perform sparse sampling on the three-dimensional echo data to obtain corresponding sparse echo data. Perform array or frequency point sparse sampling to obtain sparse echo data , its sampling process can be expressed as but not limited to: , , in, is the sampling matrix; Represents matrix dot product; is a set of spatial sampling point indexes; Frequency sampling point index set ； The index value of the element in the three-dimensional sampling matrix.

[0035] Optionally, in one embodiment of the present application, before sparse sampling is performed on the three-dimensional echo data, the method further includes: correcting the three-dimensional echo data using preset data to obtain corrected three-dimensional echo data.

[0036] Those skilled in the art will appreciate that, in the embodiments of the present application, prior to sparse sampling of the 3D echo data, certain data may be used to correct the 3D echo data, thereby obtaining corrected 3D echo data. The certain conditions and data may be configured by those skilled in the art based on actual circumstances and are not specifically limited in this application.

[0037] For example, the embodiment of the present application uses the difference in radar system equipment to correct the three-dimensional echo data, that is, the correction data based on the difference in radar system equipment obtained under air measurement conditions is used to perform matrix multiplication correction on the three-dimensional echo data, thereby obtaining three-dimensional echo data that meets certain conditions. .

[0038] In step S102 , the three-dimensional echo data and the sparse echo data are imaged respectively to obtain corresponding three-dimensional imaging results and sparse imaging results.

[0039] It is understandable that, in the imaging process of the embodiment of the present application, the PSM (Phase Shift Migration) algorithm in the millimeter wave near-field imaging method may be selected for imaging.

[0040] Furthermore, the embodiment of the present application can set the range resolution of the radar system in the imaging space. Reference distance, using the PSM algorithm implicit imaging operator Imaging of three-dimensional echo data and sparse echo data to obtain corresponding three-dimensional imaging results and sparse imaging results The expressions of the three-dimensional imaging results and the sparse imaging results can be, but are not limited to,: , , in, is the implicit imaging operator, representing the imaging process of the PSM algorithm. The inverse symbol is used to more clearly indicate that imaging is the inverse process of echo data acquisition. and They represent two-dimensional Fourier forward transform and two-dimensional inverse Fourier transform respectively; Represents the spatial wave number domain and Directional wave number and The spherical function of , whose expression can be but not limited to: .

[0041] During actual execution, the embodiment of the present application can image the three-dimensional echo data and the sparse echo data respectively, and thus obtain corresponding three-dimensional imaging results and sparse imaging results.

[0042] Optionally, in one embodiment of the present application, before the sparse imaging result is input into a pre-constructed spatial convolutional network, it also includes: based on the real data of the sparse imaging result, in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the real data, constructing a real spatial convolutional network of the spatial convolutional network; based on the imaginary data of the sparse imaging result, in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the imaginary data, constructing an imaginary spatial convolutional network of the spatial convolutional network; based on the real spatial convolutional network and the imaginary spatial convolutional network, constructing a spatial convolutional network.

[0043] It can be understood that in the embodiment of the present application, the sparse imaging results and the sparse echo data are both in complex form, and the real data and the imaginary data have their own unique spatial structures. In addition, the complex form of data is not convenient for propagation in the convolutional network. Therefore, the embodiment of the present application constructs two real space convolutional networks with the same structure but different parameters. and imaginary space convolutional networks , respectively, predictions are made from the real and imaginary data of the sparse imaging results.

[0044] Furthermore, the spatial convolutional network of the embodiment of the present application The structure of Figure 2 As shown in the figure, the main structure of the network is U-Net (Convolutional Networks for Biomedical Image Segmentation), which includes several spatial convolution layers, downsampling modules, activation functions, residual modules, upsampling modules and convolution modules. There are jump connections between the layers with the same feature shapes in the downsampling modules and the upsampling modules. The operations for retaining the real and imaginary data of the sparse imaging results are respectively and , then the mathematical expression of the complete prediction process of the spatial convolutional network can be but is not limited to: , In step S103, the sparse imaging result is input into a pre-built spatial convolutional network to predict a predicted three-dimensional imaging result corresponding to the sparse imaging result.

[0045] In some embodiments, the embodiments of the present application can input the sparse imaging results into a pre-built spatial convolutional network to obtain the corresponding predicted three-dimensional imaging results.

[0046] For example, the embodiments of the present application can be combined with Figure 2 As shown, the output of the pre-built spatial convolutional network is calculated, and the sparse imaging results are input into a network with parameters Spatial Convolutional Neural Network In the forward reasoning, we can get the predicted 3D imaging results predicted by the spatial convolutional neural network. .

[0047] Optionally, in one embodiment of the present application, before substituting the three-dimensional echo data, three-dimensional imaging results and predicted three-dimensional imaging results into a pre-constructed loss function, it also includes: constructing an imaging loss function and an image loss function based on the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results, respectively; determining the imaging loss coefficient corresponding to the imaging loss function and the image loss coefficient corresponding to the image loss function based on the imaging loss function and the image loss function; constructing a loss function based on the imaging loss function, the image loss function, the imaging loss coefficient and the image loss coefficient.

[0048] It is understandable that the embodiment of the present application can calculate the total loss of the spatial convolutional network, and its corresponding loss function includes imaging loss function, image loss function, imaging loss coefficient and image loss coefficient, wherein the loss function The mathematical expression can be but not limited to: , in, represents the image loss coefficient, represents the image loss function, represents the imaging loss coefficient, represents the imaging loss function.

[0049] Furthermore, the embodiment of the present application can update the parameters of the spatial convolutional network through the loss function defined by the above formula, calculate the total loss used in the training of the spatial convolutional network, and obtain the final spatial convolutional network after multiple iterative training.

[0050] Optionally, in one embodiment of the present application, an imaging loss function and an image loss function are constructed based on three-dimensional echo data, three-dimensional imaging results and predicted three-dimensional imaging results, respectively, including: projecting the three-dimensional imaging results to obtain a corresponding two-dimensional projection image; projecting the predicted three-dimensional imaging results to obtain a corresponding predicted two-dimensional projection image; and constructing an image loss function based on the two-dimensional projection image and the predicted two-dimensional projection image.

[0051] It is understandable that when constructing the image loss function, the embodiment of the present application can calculate the layered maximum projection result, and the modulus of the three-dimensional imaging result and the predicted three-dimensional imaging result can be calculated by the layered maximum projection method. Projected onto a two-dimensional plane to form a high-quality two-dimensional projection image and predict the 2D projection image .

[0052] Among them, in the embodiment of this application, maximum projection is a simple and feasible three-dimensional data visualization method, which is widely used in millimeter wave imaging. Maximum projection compresses three-dimensional data into one dimension by retaining the maximum modulus value in the distance dimension, thereby generating a single-channel two-dimensional image. For example, the mathematical expression of the maximum projection can be, but is not limited to: , , in, Indicates that in the distance direction, the index value belongs to the set Among all the tangent planes, select the maximum value among the pixels with the same coordinates; To calculate the modulus of each element of a matrix.

[0053] The layered maximum projection is an improvement on the maximum projection. It projects each layer of the distance dimension of the three-dimensional data onto the three color channels in turn to generate a three-channel two-dimensional image. Compared with the maximum projection, the layered maximum projection can retain more spatial information while taking advantage of the convenience of the image format. For example, the mathematical expression of hierarchical maximum projection can be, but is not limited to: , , in, Represents matrices concatenated in the channel dimension; Represents the set of natural numbers.

[0054] Furthermore, the embodiment of the present application calculates the three-dimensional imaging result according to the layered maximum projection formula The two-dimensional compression result can be expressed as, but not limited to,: , Furthermore, the embodiment of the present application can calculate the projection loss between the two-dimensional projection image and the predicted two-dimensional projection image, and then construct an image loss function .

[0055] It should be noted that in the embodiment of the present application, the two-dimensional projection image and the predicted two-dimensional projection image are more concerned than the imaging result. Therefore, a high-quality two-dimensional projection image can be used. and predict the 2D projection image The Structural Similarity Index (SSIM) between images is used as the image loss function, and its mathematical expression can be but is not limited to: , As a possible implementation method, the embodiment of the present application can project the three-dimensional imaging result and the predicted three-dimensional imaging result respectively, and then obtain the corresponding two-dimensional projection image and the predicted two-dimensional projection image, thereby constructing an image loss function.

[0056] Optionally, in one embodiment of the present application, an imaging loss function and an image loss function are constructed based on the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results, respectively, including: restoring the predicted three-dimensional imaging results to obtain the corresponding imaging echo data; and constructing an imaging loss function based on the three-dimensional echo data and the imaging echo data.

[0057] It is understandable that when constructing the imaging loss function, the embodiment of the present application can predict the three-dimensional imaging result through the implicit operator of the PSM imaging algorithm. Restore to corresponding imaging echo data, wherein the mathematical expression of the restoration process can be but is not limited to: , In addition, the embodiment of the present application takes into account the randomness of the sparse sampling matrix, and therefore uses the mean square error (MSE) to measure the similarity between the real and imaginary parts of the three-dimensional echo data and the imaging echo data. The mathematical expression of the imaging loss function can be, but is not limited to,: , In some embodiments, the embodiments of the present application can restore the predicted three-dimensional imaging results to obtain corresponding imaging echo data, and construct an imaging loss function based on the three-dimensional echo data and the imaging echo data.

[0058] In step S104, the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results are substituted into a pre-constructed loss function to obtain a corresponding loss value. When the loss value is less than a preset threshold, the final spatial convolutional network is obtained based on the loss value, the three-dimensional echo data, the three-dimensional imaging results, the sparse imaging results and the predicted three-dimensional imaging results, so as to use the spatial convolutional network to predict the final three-dimensional imaging result of the three-dimensional echo data.

[0059] As a possible implementation method, embodiments of the present application can substitute 3D echo data, 3D imaging results, and predicted 3D imaging results into a pre-constructed loss function to obtain a corresponding loss value. When the loss value is less than a certain threshold, a final spatial convolutional network is obtained, and the spatial convolutional network is then used to predict the final 3D imaging result of the 3D echo data. The threshold can be set by those skilled in the art based on actual circumstances and is not specifically limited by this application.

[0060] For example, combined Figure 3 As shown, the content of calculating the loss value in the embodiment of the present application can be: Step S301: Acquire three-dimensional echo data of the target to be measured.

[0061] Step S302: Acquire a three-dimensional imaging result of the target to be measured.

[0062] Among them, the embodiment of the present application can use the PSM algorithm implicit imaging operator The three-dimensional echo data is imaged to obtain the corresponding three-dimensional imaging results.

[0063] Step S303: Obtain the predicted three-dimensional imaging result of the pre-built spatial convolutional network.

[0064] Among them, the embodiment of the present application can perform sparse sampling on the three-dimensional echo data to obtain the corresponding sparse echo data, and use the PSM algorithm implicit imaging operator The sparse echo data is imaged to obtain the corresponding sparse imaging results, which are then input into a pre-built spatial convolutional network to obtain the corresponding predicted three-dimensional imaging results.

[0065] Step S304: Obtain a corresponding two-dimensional projection image and a predicted two-dimensional projection image using layered maximum projection.

[0066] Step S305: Calculate image loss according to the image loss function.

[0067] Step S306: Calculate imaging loss according to the imaging loss function.

[0068] Step S307: Calculate the corresponding loss value according to the image loss and the imaging loss.

[0069] The working principle of the near-field sparse imaging method based on spatial convolution proposed in the embodiment of the present application is introduced below with reference to a specific embodiment.

[0070] in, Figure 4 The present invention is a flowchart illustrating the working principle of a near-field sparse imaging method based on spatial convolution according to an embodiment of the present application.

[0071] In the spatial convolutional network reasoning process, the embodiment of the present application can obtain the three-dimensional echo data of the target to be measured, and sample and process the three-dimensional echo data to obtain the corresponding sparse echo data, and use the PSM algorithm implicit imaging operator The sparse echo data is imaged to obtain a sparse imaging result, and the real data and imaginary data of the sparse imaging result are respectively input into the corresponding real space convolution network and imaginary space convolution network to predict the predicted three-dimensional imaging result of the three-dimensional echo data.

[0072] During the spatial convolutional network training process, the present application embodiment can use the PSM algorithm implicit imaging operator The three-dimensional echo data is imaged to obtain the corresponding three-dimensional imaging results, and the three-dimensional imaging results and the predicted three-dimensional imaging results are projected using layered maximum projection to obtain the corresponding two-dimensional projection image and the predicted two-dimensional projection image, and the image loss function is used to calculate the image loss.

[0073] In addition, the embodiment of the present application can predict the three-dimensional imaging result through the implicit operator of the PSM imaging algorithm. The corresponding imaging echo data is restored, and then the imaging loss is calculated based on the three-dimensional echo data and the imaging echo data in combination with the imaging loss function.

[0074] Furthermore, the embodiment of the present application can obtain corresponding loss values based on image loss and imaging loss, and obtain the final spatial convolutional network when the loss value is less than a certain threshold.

[0075] According to the near-field sparse imaging method based on spatial convolution proposed in the embodiment of the present application, three-dimensional echo data of the target to be measured can be obtained, and the three-dimensional echo data can be sparsely sampled to obtain corresponding sparse echo data, and the three-dimensional echo data and the sparse echo data can be imaged respectively to obtain corresponding three-dimensional imaging results and sparse imaging results, and then the corresponding predicted three-dimensional imaging results can be predicted using a pre-constructed spatial convolution network, and the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results are substituted into the pre-constructed loss function to obtain the corresponding loss value, and when the loss value is less than a certain threshold, the final spatial convolution network is obtained, and then the final three-dimensional imaging result is predicted. While reducing the amount of data and computational complexity, the spatial features in millimeter wave imaging can be fully utilized to improve the imaging effect, and when processing targets with complex structures, the target information can be restored more efficiently and accurately to achieve high-quality imaging results. Thus, the problems in the related art that the design complexity and computational burden of relying on echo data are high, which reduces the operating efficiency; the actual application scenarios of matched filtering technology are limited and difficult to implement; the deep learning imaging effect is unstable and may even be unable to accurately reconstruct the target information are solved.

[0076] Next, a near-field sparse imaging device based on spatial convolution proposed in accordance with an embodiment of the present application is described with reference to the accompanying drawings.

[0077] Figure 5 A schematic block diagram of a near-field sparse imaging device based on spatial convolution according to an embodiment of the present application.

[0078] like Figure 5 As shown, the near-field sparse imaging device 10 based on spatial convolution is applied to the network training stage, wherein the near-field sparse imaging device 10 based on spatial convolution includes: a first acquisition module 100, a first imaging module 200, a first prediction module 300 and a generation module 400.

[0079] The first acquisition module 100 is used to acquire three-dimensional echo data of the target to be measured, and perform sparse sampling on the three-dimensional echo data to obtain corresponding sparse echo data.

[0080] The first imaging module 200 is used to image the three-dimensional echo data and the sparse echo data respectively to obtain corresponding three-dimensional imaging results and sparse imaging results.

[0081] The first prediction module 300 is used to input the sparse imaging result into a pre-built spatial convolutional network to predict the predicted three-dimensional imaging result corresponding to the sparse imaging result.

[0082] A generation module 400 is used to substitute the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results into a pre-constructed loss function to obtain a corresponding loss value, and when the loss value is less than a preset threshold, a final spatial convolutional network is obtained based on the loss value, the three-dimensional echo data, the three-dimensional imaging results, the sparse imaging results and the predicted three-dimensional imaging results, so as to use the spatial convolutional network to predict the final three-dimensional imaging result of the three-dimensional echo data.

[0083] Optionally, in one embodiment of the present application, it further includes: a first building module, a second building module and a third building module.

[0084] Among them, the first construction module is used to construct the real part spatial convolution network of the spatial convolution network based on the real data of the sparse imaging results, combined with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the real data before inputting the sparse imaging results into the pre-constructed spatial convolution network.

[0085] The second construction module is used to construct an imaginary part spatial convolution network of the spatial convolution network based on the imaginary part data of the sparse imaging results, combined with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the imaginary part data.

[0086] The third building module is used to build a spatial convolutional network based on the real space convolutional network and the imaginary space convolutional network.

[0087] Optionally, in one embodiment of the present application, it further includes: a fourth building module, a determination module and a fifth building module.

[0088] Among them, the fourth construction module is used to construct an imaging loss function and an image loss function based on the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results before substituting the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results into the pre-constructed loss function.

[0089] A determination module is used to determine an imaging loss coefficient corresponding to the imaging loss function and an image loss coefficient corresponding to the image loss function based on the imaging loss function and the image loss function.

[0090] The fifth construction module is used to construct a loss function based on the imaging loss function, the image loss function, the imaging loss coefficient and the image loss coefficient.

[0091] Optionally, in one embodiment of the present application, the fourth building module includes: a first projection unit, a second projection unit and a first building unit.

[0092] The first projection unit is used to project the three-dimensional imaging result to obtain a corresponding two-dimensional projection image.

[0093] The second projection unit is used to project the predicted three-dimensional imaging result to obtain a corresponding predicted two-dimensional projection image.

[0094] The first construction unit is used to construct a projection image loss function based on the two-dimensional projection image and the predicted two-dimensional projection image.

[0095] Optionally, in one embodiment of the present application, the fourth building block includes: a reduction unit and a second building unit.

[0096] The restoration unit is used to restore the predicted three-dimensional imaging result to obtain corresponding imaging echo data.

[0097] The second construction unit is configured to construct an imaging loss function based on the three-dimensional echo data and the imaging echo data.

[0098] Optionally, in one embodiment of the present application, it further includes: a correction module.

[0099] The correction module is used to correct the three-dimensional echo data using preset data before sparse sampling of the three-dimensional echo data to obtain corrected three-dimensional echo data.

[0100] It should be noted that the above explanation of the embodiment of the near-field sparse imaging method based on spatial convolution is also applicable to the near-field sparse imaging device based on spatial convolution in this embodiment, and will not be repeated here.

[0101] According to the near-field sparse imaging device based on spatial convolution proposed in the embodiment of the present application, the three-dimensional echo data of the target to be measured can be obtained, and the three-dimensional echo data can be sparsely sampled to obtain corresponding sparse echo data, and the three-dimensional echo data and the sparse echo data can be imaged respectively to obtain corresponding three-dimensional imaging results and sparse imaging results, and then the corresponding predicted three-dimensional imaging results can be predicted using a pre-constructed spatial convolution network, and the three-dimensional echo data, the three-dimensional imaging results and the predicted three-dimensional imaging results can be substituted into the pre-constructed loss function to obtain the corresponding loss value, and when the loss value is less than a certain threshold, the final spatial convolution network is obtained, and then the final three-dimensional imaging result is predicted. While reducing the amount of data and computational complexity, the spatial features in millimeter wave imaging can be fully utilized to improve the imaging effect, and when processing targets with complex structures, the target information can be restored more efficiently and accurately to achieve high-quality imaging results. Thus, the problems in the related art that the design complexity and computational burden of relying on echo data are high, which reduces the operating efficiency; the actual application scenarios of matched filtering technology are limited and difficult to implement; the deep learning imaging effect is unstable and may even be unable to accurately reconstruct the target information are solved.

[0102] The above embodiment describes the network training phase. The following describes an embodiment of the network application phase.

[0103] Figure 6 The present invention provides a flowchart of a near-field sparse imaging method based on spatial convolution according to another embodiment of the present application.

[0104] like Figure 6 As shown, the near-field sparse imaging method based on spatial convolution is applied to the network application stage, wherein the method includes the following steps: In step S601 , actual three-dimensional echo data of an actual target to be measured is acquired, and sparse sampling is performed on the actual three-dimensional echo data to obtain corresponding actual sparse echo data.

[0105] In step S602 , the actual three-dimensional echo data and the actual sparse echo data are imaged respectively to obtain corresponding actual three-dimensional imaging results and actual sparse imaging results.

[0106] In step S603, the actual sparse imaging result is input into a pre-trained spatial convolutional network to predict the final three-dimensional imaging result corresponding to the actual three-dimensional echo data, wherein the pre-trained spatial convolutional network is trained by the actual three-dimensional echo data, the actual three-dimensional imaging result, and the loss function corresponding to the actual sparse imaging result.

[0107] During the actual execution process, the embodiment of the present application can obtain the actual three-dimensional echo data of the actual target to be measured, and perform sparse sampling on the actual three-dimensional echo data to obtain the corresponding actual sparse echo data, and respectively image the actual three-dimensional echo data and the actual sparse echo data to obtain the corresponding actual three-dimensional imaging results and the actual sparse imaging results, and then use the pre-trained spatial convolutional network to predict the final three-dimensional imaging result corresponding to the actual three-dimensional echo data.

[0108] For example, the embodiments of the present application can be combined with Figure 7 As shown, the imaging effect is demonstrated.

[0109] According to the near-field sparse imaging method based on spatial convolution proposed in the embodiment of the present application, the actual three-dimensional echo data of the actual target to be measured can be obtained, and the actual three-dimensional echo data can be sparsely sampled to obtain the corresponding actual sparse echo data, and the actual three-dimensional echo data and the actual sparse echo data can be imaged respectively to obtain the corresponding actual three-dimensional imaging results and the actual sparse imaging results. The final three-dimensional imaging results can then be predicted using a pre-trained spatial convolutional network. While reducing the amount of data and computational complexity, the spatial features in millimeter wave imaging can be fully utilized to improve the imaging effect. When processing targets with complex structures, the target information can be restored more efficiently and accurately to achieve high-quality imaging results. This solves the problems in related technologies such as the high design complexity and computational burden of relying on echo data, which reduces operating efficiency; the limited practical application scenarios of matched filtering technology, which makes implementation difficult; and the unstable imaging effect of deep learning, which may even be unable to accurately reconstruct target information.

[0110] Next, a near-field sparse imaging device based on spatial convolution proposed in accordance with an embodiment of the present application is described with reference to the accompanying drawings.

[0111] Figure 8 The present invention is a block diagram of a near-field sparse imaging device based on spatial convolution according to another embodiment of the present application.

[0112] like Figure 8 As shown, the near-field sparse imaging device 20 based on spatial convolution is applied in the network application stage, wherein the near-field sparse imaging device 20 based on spatial convolution includes: a second acquisition module 500, a second imaging module 600 and a second prediction module 700.

[0113] The second acquisition module 500 is used to acquire actual three-dimensional echo data of the actual target to be measured, and perform sparse sampling on the actual three-dimensional echo data to obtain corresponding actual sparse echo data.

[0114] The second imaging module 600 is used to image the actual three-dimensional echo data and the actual sparse echo data respectively to obtain corresponding actual three-dimensional imaging results and actual sparse imaging results.

[0115] The second prediction module 700 is used to input the actual sparse imaging result into a pre-trained spatial convolutional network to predict the final three-dimensional imaging result corresponding to the actual three-dimensional echo data, wherein the pre-trained spatial convolutional network is trained by the actual three-dimensional echo data, the actual three-dimensional imaging result, and the loss function corresponding to the actual sparse imaging result.

[0116] It should be noted that the above explanation of the embodiment of the near-field sparse imaging method based on spatial convolution is also applicable to the near-field sparse imaging device based on spatial convolution in this embodiment, and will not be repeated here.

[0117] According to the near-field sparse imaging device based on spatial convolution proposed in the embodiment of the present application, the actual three-dimensional echo data of the actual target to be measured can be obtained, and the actual three-dimensional echo data can be sparsely sampled to obtain the corresponding actual sparse echo data, and the actual three-dimensional echo data and the actual sparse echo data can be imaged respectively to obtain the corresponding actual three-dimensional imaging results and the actual sparse imaging results. The final three-dimensional imaging results can then be predicted using a pre-trained spatial convolutional network. While reducing the amount of data and computational complexity, the spatial features in millimeter wave imaging can be fully utilized to improve the imaging effect. When processing targets with complex structures, the target information can be restored more efficiently and accurately to achieve high-quality imaging results. This solves the problems in related technologies such as the high design complexity and computational burden of relying on echo data, which reduces operating efficiency; the limited practical application scenarios of matched filtering technology, which makes implementation difficult; and the unstable imaging effect of deep learning, which may even be unable to accurately reconstruct target information.

[0118] Figure 9 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. The electronic device may include: A memory 901 , a processor 902 , and a computer program stored in the memory 901 and executable on the processor 902 .

[0119] When the processor 902 executes the program, the near-field sparse imaging method based on spatial convolution provided in the above embodiment is implemented.

[0120] Furthermore, the electronic device further includes: The communication interface 903 is used for communication between the memory 901 and the processor 902 .

[0121] The memory 901 is used to store computer programs that can be run on the processor 902 .

[0122] The memory 901 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0123] If the memory 901, processor 902, and communication interface 903 are implemented independently, the communication interface 903, memory 901, and processor 902 can be connected to each other via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 9 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0124] Optionally, in a specific implementation, if the memory 901, the processor 902 and the communication interface 903 are integrated on a chip, the memory 901, the processor 902 and the communication interface 903 can communicate with each other through an internal interface.

[0125] The processor 902 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0126] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned near-field sparse imaging method based on spatial convolution.

[0127] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed, implements the above-mentioned near-field sparse imaging method based on spatial convolution.

[0128] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0129] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of this application, "N" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0130] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing a custom logical function or process step, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed in a different order than shown or discussed, including performing functions in a substantially simultaneous manner or in a reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application pertain.

[0131] The logic and / or steps represented in a flowchart or otherwise described herein, for example, can be considered a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" is any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (not exhaustive) of computer-readable media include: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically by optically scanning the paper or other medium and then editing, interpreting, or otherwise processing in a suitable manner as necessary, and then storing it in a computer memory.

[0132] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiment, the N steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented using hardware, as in another embodiment, it can be implemented using any one or a combination of the following technologies known in the art: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0133] Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above-mentioned embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0134] In addition, the functional units in the various embodiments of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into a module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0135] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A near-field sparse imaging method based on spatial convolution, characterized in that: Applied to the network training stage, wherein the method comprises the following steps: Acquiring three-dimensional echo data of the target to be measured, and performing sparse sampling on the three-dimensional echo data to obtain corresponding sparse echo data; Imaging the three-dimensional echo data and the sparse echo data respectively to obtain corresponding three-dimensional imaging results and sparse imaging results; Inputting the sparse imaging result into a pre-built spatial convolutional network to predict a predicted three-dimensional imaging result corresponding to the sparse imaging result; The three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result are substituted into a pre-constructed loss function to obtain a corresponding loss value, and when the loss value is less than a preset threshold, a final spatial convolutional network is obtained based on the loss value, the three-dimensional echo data, the three-dimensional imaging result, the sparse imaging result and the predicted three-dimensional imaging result, so as to use the spatial convolutional network to predict the final three-dimensional imaging result of the three-dimensional echo data.

2. The method according to claim 1, characterized in that Before inputting the sparse imaging result into the pre-built spatial convolutional network, the method further includes: Based on the real part data of the sparse imaging result, in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the real part data, a real part spatial convolution network of the spatial convolution network is constructed; Based on the imaginary part data of the sparse imaging result, in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the imaginary part data, an imaginary part spatial convolution network of the spatial convolution network is constructed; Based on the real part spatial convolutional network and the imaginary part spatial convolutional network, the spatial convolutional network is constructed.

3. The method according to claim 1, characterized in that Before substituting the three-dimensional echo data, the three-dimensional imaging result, and the predicted three-dimensional imaging result into a pre-constructed loss function, the method further includes: constructing an imaging loss function and an image loss function based on the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result; Based on the imaging loss function and the image loss function, determining an imaging loss coefficient corresponding to the imaging loss function and an image loss coefficient corresponding to the image loss function; A loss function is constructed based on the imaging loss function, the image loss function, the imaging loss coefficient, and the image loss coefficient.

4. The method according to claim 3, characterized in that The constructing of an imaging loss function and an image loss function based on the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result respectively includes: Projecting the three-dimensional imaging result to obtain a corresponding two-dimensional projection image; Projecting the predicted three-dimensional imaging result to obtain a corresponding predicted two-dimensional projection image; The image loss function is constructed based on the two-dimensional projection image and the predicted two-dimensional projection image.

5. The method according to claim 3, characterized in that The constructing of an imaging loss function and an image loss function based on the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result respectively includes: Restoring the predicted three-dimensional imaging result to obtain corresponding imaging echo data; The imaging loss function is constructed based on the three-dimensional echo data and the imaging echo data.

6. The method according to claim 1, characterized in that Before sparse sampling is performed on the three-dimensional echo data, the method further includes: The three-dimensional echo data is corrected using the preset data to obtain corrected three-dimensional echo data.

7. A near-field sparse imaging method based on spatial convolution, characterized in that: Applied to the network application stage, wherein the method includes the following steps: Acquiring actual three-dimensional echo data of an actual target to be measured, and performing sparse sampling on the actual three-dimensional echo data to obtain corresponding actual sparse echo data; Imaging the actual three-dimensional echo data and the actual sparse echo data respectively to obtain corresponding actual three-dimensional imaging results and actual sparse imaging results; The actual sparse imaging result is input into a pre-trained spatial convolutional network to predict a final three-dimensional imaging result corresponding to the actual three-dimensional echo data, wherein the pre-trained spatial convolutional network is trained by a loss function corresponding to the actual three-dimensional echo data, the actual three-dimensional imaging result, and the actual sparse imaging result.

8. A near-field sparse imaging device based on spatial convolution, characterized in that: Applied to the network training stage, wherein the device includes: A first acquisition module is used to acquire three-dimensional echo data of the target to be measured, and perform sparse sampling on the three-dimensional echo data to obtain corresponding sparse echo data; a first imaging module, configured to image the three-dimensional echo data and the sparse echo data respectively to obtain corresponding three-dimensional imaging results and sparse imaging results; A first prediction module is configured to input the sparse imaging result into a pre-built spatial convolutional network to predict a predicted three-dimensional imaging result corresponding to the sparse imaging result; A generation module is used to substitute the three-dimensional echo data, the three-dimensional imaging result and the predicted three-dimensional imaging result into a pre-constructed loss function to obtain a corresponding loss value, and when the loss value is less than a preset threshold, obtain a final spatial convolutional network based on the loss value, the three-dimensional echo data, the three-dimensional imaging result, the sparse imaging result and the predicted three-dimensional imaging result, so as to use the spatial convolutional network to predict the final three-dimensional imaging result of the three-dimensional echo data.

9. The device according to claim 8, characterized in that Also includes: A first construction module is configured to construct a real part spatial convolution network of the spatial convolution network based on the real part data of the sparse imaging result and in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the real part data before inputting the sparse imaging result into the pre-constructed spatial convolution network; A second construction module is used to construct an imaginary part spatial convolution network of the spatial convolution network based on the imaginary part data of the sparse imaging result, in combination with the spatial convolution layer, downsampling module, activation function, residual module, upsampling module and convolution module corresponding to the imaginary part data; The third construction module is used to construct the spatial convolutional network based on the real spatial convolutional network and the imaginary spatial convolutional network.

10. The device according to claim 8, characterized in that Also includes: a fourth construction module, configured to construct an imaging loss function and an image loss function based on the three-dimensional echo data, the three-dimensional imaging result, and the predicted three-dimensional imaging result, respectively, before substituting the three-dimensional echo data, the three-dimensional imaging result, and the predicted three-dimensional imaging result into a pre-constructed loss function; a determination module, configured to determine, based on the imaging loss function and the image loss function, an imaging loss coefficient corresponding to the imaging loss function and an image loss coefficient corresponding to the image loss function; A fifth construction module is used to construct a loss function based on the imaging loss function, the image loss function, the imaging loss coefficient and the image loss coefficient.

11. The device according to claim 10, characterized in that The fourth building block includes: a first projection unit, configured to project the three-dimensional imaging result to obtain a corresponding two-dimensional projection image; a second projection unit, configured to project the predicted three-dimensional imaging result to obtain a corresponding predicted two-dimensional projection image; A first construction unit is configured to construct the projection image loss function based on the two-dimensional projection image and the predicted two-dimensional projection image.

12. The device according to claim 10, characterized in that The fourth building block includes: a restoration unit, configured to restore the predicted three-dimensional imaging result to obtain corresponding imaging echo data; The second constructing unit is configured to construct the imaging loss function based on the three-dimensional echo data and the imaging echo data.

13. The device according to claim 8, characterized in that Also includes: The correction module is used to correct the three-dimensional echo data using preset data before sparse sampling the three-dimensional echo data to obtain corrected three-dimensional echo data.

14. A near-field sparse imaging device based on spatial convolution, characterized in that: Applied to the network application stage, wherein the device includes: A second acquisition module is used to acquire actual three-dimensional echo data of the actual target to be measured, and perform sparse sampling on the actual three-dimensional echo data to obtain corresponding actual sparse echo data; a second imaging module, configured to image the actual three-dimensional echo data and the actual sparse echo data respectively, to obtain corresponding actual three-dimensional imaging results and actual sparse imaging results; A second prediction module is used to input the actual sparse imaging result into a pre-trained spatial convolutional network to predict a final three-dimensional imaging result corresponding to the actual three-dimensional echo data, wherein the pre-trained spatial convolutional network is trained by a loss function corresponding to the actual three-dimensional echo data, the actual three-dimensional imaging result, and the actual sparse imaging result.

15. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the near-field sparse imaging method based on spatial convolution according to any one of claims 1 to 6 or the near-field sparse imaging method based on spatial convolution according to claim 7.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the near-field sparse imaging method based on spatial convolution according to any one of claims 1 to 6 or the near-field sparse imaging method based on spatial convolution according to claim 7.

17. A computer program product, characterized in that The invention comprises a computer program, which, when executed, is used to implement the near-field sparse imaging method based on spatial convolution according to any one of claims 1 to 6 or the near-field sparse imaging method based on spatial convolution according to claim 7.

Images