Method, device and medium for generating medium image of non-cooperative target

Abstract

The present application relates to the technical field of machine learning, and particularly relates to a non-cooperative target medium image generation method, device, equipment and medium. The method comprises the following steps: acquiring a non-cooperative target non-medium image; generating a plurality of three-dimensional scattering centers of the smooth surface of the metal material of the non-medium image by using a GTD model; wherein the target surface of the non-medium image is free of medium coating; inputting the three-dimensional scattering centers of the non-medium image into a pre-trained model to adjust the scattering amplitude coefficient of each three-dimensional scattering center, and obtaining a non-cooperative target medium image. The present application can accurately generate a medium image according to a non-medium image by fully utilizing the prior information of a known cooperative target to train a neural network, and can effectively improve the accuracy of the non-cooperative target medium image compared with the non-cooperative target non-medium image generated by a traditional method.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, and in particular to a method, apparatus, device, and medium for generating media images of non-cooperative targets. Background Technology

[0002] The Geometric Theory of Diffraction (GTD) model is a commonly used scattering center model. It expresses the frequency dependence of the scattering center as a power function of a half-integer exponent, and the frequency dependence factor corresponds to the geometric structure of the target's scattering center, providing rich information about the target's structure. However, for non-cooperative targets, due to limited intelligence gathering, it is difficult to accurately model their surface roughness and coating materials. Considering the target's surface roughness and coating materials, it is difficult to generate a medium image of a non-cooperative target using the traditional GTD model, i.e., it is difficult to obtain a two-dimensional image with a high similarity to the measured two-dimensional image.

[0003] Therefore, there is an urgent need for a method to generate media images of non-cooperative targets. Summary of the Invention

[0004] To address the current problem of inaccurate generation of media images for non-cooperative targets, embodiments of the present invention provide a method, apparatus, device, and medium for generating media images of non-cooperative targets.

[0005] In a first aspect, embodiments of the present invention provide a method for generating a medium image of a non-cooperative target, the method comprising:

[0006] Acquire media-free images of non-cooperative targets;

[0007] A number of three-dimensional scattering centers of the smooth metallic surface in the medium-free image are generated using the GTD model; wherein the target surface of the medium-free image has no medium coating.

[0008] The three-dimensional scattering centers of the medium-free image are input into a pre-trained model to adjust the scattering amplitude coefficient of each three-dimensional scattering center, thereby obtaining the medium image of the non-cooperative target.

[0009] Secondly, embodiments of the present invention also provide an apparatus for generating a medium image of a non-cooperative target, the apparatus comprising:

[0010] Acquisition unit, used to acquire media-free images of non-cooperative targets;

[0011] The generation unit is used to generate several three-dimensional scattering centers of the smooth metallic surface of the medium-free image using a GTD model; wherein the target surface of the medium-free image has no medium coating.

[0012] An adjustment unit is used to input the three-dimensional scattering centers of the medium-free image into a pre-trained model to adjust the scattering amplitude coefficient of each of the three-dimensional scattering centers, thereby obtaining the medium image of the non-cooperative target.

[0013] Thirdly, embodiments of the present invention also provide a computing device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the method described in any embodiment of this specification.

[0014] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the methods described in any embodiment of this specification.

[0015] This invention provides a method, apparatus, device, and medium for generating media images of non-cooperative targets. By fully utilizing prior information of known cooperative targets to train a neural network, it can generate media images based on media-free images with relatively high accuracy. Compared with media-free images of non-cooperative targets generated by traditional methods, this solution can effectively improve the accuracy of media images of non-cooperative targets. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a method for generating a medium image of a non-cooperative target according to an embodiment of the present invention;

[0018] Figure 2 This is a hardware architecture diagram of a computing device provided in an embodiment of the present invention;

[0019] Figure 3 This is a structural diagram of a device for generating a medium image of a non-cooperative target according to an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] The specific implementation of the above concept is described below.

[0022] Please refer to Figure 1 This invention provides a method for generating a medium image of a non-cooperative target, the method comprising:

[0023] Step 100: Obtain a medium-free image of the non-cooperative target;

[0024] Step 102: Use the GTD model to generate several three-dimensional scattering centers of a smooth metallic surface in a medium-free image; wherein, the target surface in the medium-free image has no medium coating.

[0025] Step 104: Input the three-dimensional scattering centers of the medium-free image into the pre-trained model to adjust the scattering amplitude coefficient of each three-dimensional scattering center to obtain the medium image of the non-cooperative target.

[0026] In this embodiment of the invention, by fully utilizing the prior information of known cooperative targets to train the neural network, it is possible to generate medium images based on mediumless images with relatively high accuracy. Compared with the mediumless images of non-cooperative targets generated by traditional methods, this solution can effectively improve the accuracy of medium images of non-cooperative targets.

[0027] For step 100:

[0028] In this embodiment, since there is limited intelligence gathering on non-cooperative targets, it is difficult to accurately model their surface roughness level and surface coating material. Only medium-free images of non-cooperative targets can be obtained. The surface of a target without a medium coating is usually a smooth surface. The surface of a target with a medium coating has a medium coating and can reflect information such as the surface roughness level and surface coating material of the target. The measured two-dimensional image is the medium image.

[0029] Regarding step 102:

[0030] Since non-cooperative targets are usually made of metal, several three-dimensional scattering centers of a smooth metallic surface can be generated directly using the GTD (Geometrical Theory of Diffraction) model without a medium.

[0031] In some implementations, the three-dimensional scattering center of the medium-free image is generated by the following formula:

[0032]

[0033] in,

[0034]

[0035]

[0036] In the formula, y is the radar echo signal corresponding to the mediumless image, k is the wave number, and k c Here, f is the center wave number, and f is the radar observation frequency. c Here, c is the center frequency, c is the speed of light, M is the number of scattering centers, and A is the center frequency. m r is the scattering amplitude coefficient of the m-th scattering center. m Let α be the radial position of the m-th scattering center. m This is a type parameter, taking values ​​that are integer multiples of 0.5, where ε represents noise and j is an imaginary number. It is a unit vector.

[0037] In this embodiment, there are M three-dimensional scattering centers in the medium-free image, and the scattering amplitude coefficient of each scattering center is the corresponding A. m The radial position of each scattering center is r. m .

[0038] Regarding step 104:

[0039] In some implementations, the model is generated based on training a pre-built neural network, which consists of a product layer, a first convolutional layer, a second convolutional layer, a pooling layer, an activation function layer, and an output layer connected in series. The product layer is used to modulate the intensity of each three-dimensional scattering center, the first convolutional layer is used to transform the three-dimensional scattering center to a two-dimensional image space to obtain a two-dimensional image matrix, and the second convolutional layer is used to adjust the dimension of the two-dimensional image matrix.

[0040] In this embodiment of the invention, the kernel size of the product layer is M*N*P, which is used to modulate the intensity of each three-dimensional scattering center. The kernel size of the first convolutional layer is 1*1 and the number of channels is P. The kernel size of the second convolutional layer is 3*3 and the number of channels is 1. The pooling layer window size is 3*3 and the stride is 3. The activation function in the activation function layer is the ReLU function.

[0041] The ReLU function is shown in the following formula:

[0042]

[0043] In some implementations, step 104 may include:

[0044] S1, input the three-dimensional scattering centers of the mediumless image into the pre-trained model, and the model's product layer modulates the intensity of each three-dimensional scattering center;

[0045] S2, the first convolutional layer transforms the modulated three-dimensional scattering center into a two-dimensional image space to obtain a two-dimensional image matrix;

[0046] S3, after adjusting the dimensions of the two-dimensional image matrix in the second convolutional layer, it passes through the pooling layer and the activation function layer in sequence to obtain the activated two-dimensional image matrix;

[0047] S4, determine whether the dimension of the activated two-dimensional image matrix is ​​consistent with the preset dimension of the medium image pixels;

[0048] S5. If there is no consistency, the activated two-dimensional image matrix is ​​input into the second convolutional layer, and S3-S4 are repeated to continue adjusting the dimension of the two-dimensional image matrix.

[0049] S6. If they match, the activated two-dimensional image matrix is ​​input to the output layer to obtain the medium image of the non-cooperative target.

[0050] In this embodiment, by continuously adjusting the dimensions of the two-dimensional image matrix, the obtained medium image of the non-cooperative target is made to a preset medium image size, making the medium image clearer and more accurate.

[0051] In some implementations, the medium image of a non-cooperative target is represented as:

[0052]

[0053] in,

[0054]

[0055]

[0056] In the formula, y′ is the radar echo signal corresponding to the medium image, k is the wave number, and k c Here, f is the center wave number, and f is the radar observation frequency. c Here, c is the center frequency, c is the speed of light, M is the number of scattering centers, and A is the center frequency. m r is the scattering amplitude coefficient of the m-th scattering center. m Let α be the radial position of the m-th scattering center. m This is a type parameter, taking values ​​that are integer multiples of 0.5, where ε represents noise and j is an imaginary number. B is a unit vector. m is the coefficient of variation of the scattering amplitude coefficient at the m-th scattering center.

[0057] In this embodiment, by utilizing the medium image generation method of this scheme, a pre-trained model can be used to calculate the variation coefficient of the scattering amplitude coefficient of each scattering center based on the prior information of the known cooperative target. This allows for the adjustment of the scattering amplitude coefficient of the three-dimensional scattering center of the non-cooperative target without medium image, resulting in a more accurate two-dimensional medium image.

[0058] In some implementations, the model is trained and generated in the following manner:

[0059] Acquire several labeled media-free image samples; where the label is the measured two-dimensional image corresponding to the media-free image sample.

[0060] The medium-free image samples are sequentially input into the constructed neural network to generate the corresponding two-dimensional images;

[0061] By utilizing the difference between each two-dimensional image and the corresponding measured two-dimensional image, the network parameters of the neural network are adjusted until a model that meets the expectations is obtained.

[0062] In this embodiment, medium-free images of cooperative targets can be used as training samples, and measured 2D images of cooperative targets can be used as labels. A GTD model is used to generate the 3D scattering center of each medium-free image sample. This 3D scattering center is then input into a constructed neural network. A product layer, a first convolutional layer, a second convolutional layer, a pooling layer, and an activation function layer are sequentially used to process the 3D scattering center of each medium-free image sample, resulting in an activated 2D image matrix. The dimension of the activated 2D image matrix is ​​then compared with the preset pixel dimension of the medium image to determine if the convergence condition is met. If not, the process returns to the second convolutional layer until the convergence condition is met. The output layer calculates the difference between the generated 2D image and the corresponding measured 2D image, using a loss function to adjust the network parameters until a model with the expected accuracy is obtained. Therefore, during the inference phase, the 3D scattering center of a medium-free image of a non-cooperative target is input into the model, allowing the model to generate a medium image of the non-cooperative target.

[0063] like Figure 2 , Figure 3 As shown, this embodiment of the invention provides an apparatus for generating a medium image of a non-cooperative target. The apparatus embodiment can be implemented in software, hardware, or a combination of both. From a hardware perspective, as... Figure 2 The diagram shown is a hardware architecture diagram of a computing device housing a medium image generation apparatus for a non-cooperative target, as provided in an embodiment of the present invention. (Except for...) Figure 2In addition to the processor, memory, network interface, and non-volatile memory shown, the computing device in the embodiment may also include other hardware, such as a forwarding chip responsible for processing packets. Taking software implementation as an example, such as... Figure 3 As shown, as a logical device, it is formed by the CPU of its computing device reading the corresponding computer program from non-volatile memory into memory and running it. This embodiment provides a device for generating a medium image of a non-cooperative target, the device comprising:

[0064] Acquisition unit 301 is used to acquire media-free images of non-cooperative targets;

[0065] The generation unit 302 is used to generate several three-dimensional scattering centers of a smooth metallic surface in a medium-free image using a GTD model; wherein the target surface of the medium-free image has no medium coating.

[0066] The adjustment unit 303 is used to input the three-dimensional scattering center of the medium-free image into the pre-trained model to adjust the scattering amplitude coefficient of each three-dimensional scattering center to obtain the medium image of the non-cooperative target.

[0067] In one embodiment of the present invention, the three-dimensional scattering center of the medium-free image in the generation unit 302 is generated by the following formula:

[0068]

[0069] in,

[0070]

[0071]

[0072] In the formula, y is the radar echo signal corresponding to the mediumless image, k is the wave number, and k c Here, f is the center wave number, and f is the radar observation frequency. c Here, c is the center frequency, c is the speed of light, M is the number of scattering centers, and A is the center frequency. m r is the scattering amplitude coefficient of the m-th scattering center. m Let α be the radial position of the m-th scattering center. m This is a type parameter, taking values ​​that are integer multiples of 0.5, where ε represents noise and j is an imaginary number. It is a unit vector.

[0073] In one embodiment of the present invention, the model in the adjustment unit 303 is generated based on a pre-built neural network. The neural network consists of a product layer, a first convolutional layer, a second convolutional layer, a pooling layer, an activation function layer, and an output layer connected in series. The product layer is used to modulate the intensity of each three-dimensional scattering center, the first convolutional layer is used to transform the three-dimensional scattering center to a two-dimensional image space to obtain a two-dimensional image matrix, and the second convolutional layer is used to adjust the dimension of the two-dimensional image matrix.

[0074] In one embodiment of the present invention, the adjustment unit 303 is configured to perform:

[0075] S1, input the three-dimensional scattering centers of the mediumless image into the pre-trained model, and the model's product layer modulates the intensity of each three-dimensional scattering center;

[0076] S2, the first convolutional layer transforms the modulated three-dimensional scattering center into a two-dimensional image space to obtain a two-dimensional image matrix;

[0077] S3, after adjusting the dimensions of the two-dimensional image matrix in the second convolutional layer, it passes through the pooling layer and the activation function layer in sequence to obtain the activated two-dimensional image matrix;

[0078] S4, determine whether the dimension of the activated two-dimensional image matrix is ​​consistent with the preset dimension of the medium image pixels;

[0079] S5. If there is no consistency, the activated two-dimensional image matrix is ​​input into the second convolutional layer, and S3-S4 are repeated to continue adjusting the dimension of the two-dimensional image matrix.

[0080] S6. If they match, the activated two-dimensional image matrix is ​​input to the output layer to obtain the medium image of the non-cooperative target.

[0081] In one embodiment of the present invention, the activation function in the activation function layer of the adjustment unit 303 is the ReLU function.

[0082] In one embodiment of the present invention, the medium image of the non-cooperative target in the adjustment unit 303 is represented as follows:

[0083]

[0084] in,

[0085]

[0086]

[0087] In the formula, y′ is the radar echo signal corresponding to the medium image, k is the wave number, and k c Here, f is the center wave number, and f is the radar observation frequency. cHere, c is the center frequency, c is the speed of light, M is the number of scattering centers, and A is the center frequency. m r is the scattering amplitude coefficient of the m-th scattering center. m Let α be the radial position of the m-th scattering center. m This is a type parameter, taking values ​​that are integer multiples of 0.5, where ε represents noise and j is an imaginary number. B is a unit vector. m is the coefficient of variation of the scattering amplitude coefficient at the m-th scattering center.

[0088] In one embodiment of the present invention, the model in the adjustment unit 303 is generated by training in the following manner:

[0089] Acquire several labeled media-free image samples; where the label is the measured two-dimensional image corresponding to the media-free image sample.

[0090] The medium-free image samples are sequentially input into the constructed neural network to generate the corresponding two-dimensional images;

[0091] By utilizing the difference between each two-dimensional image and the corresponding measured two-dimensional image, the network parameters of the neural network are adjusted until a model that meets the expectations is obtained.

[0092] It is understood that the structures illustrated in the embodiments of the present invention do not constitute a specific limitation on an apparatus for generating a medium image of a non-cooperative target. In other embodiments of the present invention, an apparatus for generating a medium image of a non-cooperative target may include more or fewer components than illustrated, or combine some components, or split some components, or arrange the components differently. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0093] The information interaction and execution process between the modules in the above-mentioned device are based on the same concept as the method embodiment of the present invention, and the specific details can be found in the description of the method embodiment of the present invention, and will not be repeated here.

[0094] This invention also provides a computing device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements a method for generating a medium image of a non-cooperative target according to any embodiment of this invention.

[0095] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform a method for generating a medium image of a non-cooperative target according to any embodiment of this invention.

[0096] Specifically, a system or apparatus equipped with a storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer (or CPU or MPU) of the system or apparatus may read and execute the program code stored in the storage medium.

[0097] In this case, the program code read from the storage medium can itself implement the function of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute part of the present invention.

[0098] Examples of storage media used to provide program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer via a communication network.

[0099] It should be clear that not only can the program code read by the computer be executed, but also the operating system or other components on the computer can be instructed based on the program code to perform some or all of the actual operations, thereby achieving the function of any of the embodiments described above.

[0100] Furthermore, it is understood that the program code read from the storage medium is written to the memory set in the expansion board inserted into the computer or to the memory set in the expansion module connected to the computer. Then, based on the instructions of the program code, the CPU or other components installed on the expansion board or expansion module execute some and all of the actual operations, thereby realizing the function of any of the above embodiments.

[0101] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0102] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as ROM, RAM, magnetic disk, or optical disk.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating a medium image of a non-cooperative target, characterized in that, include: Acquire media-free images of non-cooperative targets; A number of three-dimensional scattering centers of the smooth metallic surface in the medium-free image are generated using the GTD model; wherein the target surface of the medium-free image has no medium coating. The three-dimensional scattering centers of the medium-free image are input into a pre-trained model to adjust the scattering amplitude coefficient of each three-dimensional scattering center, thereby obtaining the medium image of the non-cooperative target. The model is generated based on a pre-built neural network trained on it. The neural network consists of a multiplication layer, a first convolutional layer, a second convolutional layer, a pooling layer, an activation function layer, and an output layer connected in series. The multiplication layer is used to modulate the intensity of each of the three-dimensional scattering centers. The first convolutional layer is used to transform the three-dimensional scattering centers into a two-dimensional image space to obtain a two-dimensional image matrix. The second convolutional layer is used to adjust the dimension of the two-dimensional image matrix. The step of inputting the three-dimensional scattering centers of the medium-free image into a pre-trained model to adjust the scattering amplitude coefficient of each three-dimensional scattering center to obtain the medium image of the non-cooperative target includes: S1, the three-dimensional scattering centers of the medium-free image are input into a pre-trained model, and the multiplication layer of the model modulates the intensity of each of the three-dimensional scattering centers; S2, the first convolutional layer transforms the modulated three-dimensional scattering center into a two-dimensional image space to obtain a two-dimensional image matrix; S3, after the second convolutional layer adjusts the dimension of the two-dimensional image matrix, it passes through the pooling layer and the activation function layer in sequence to obtain the activated two-dimensional image matrix; S4, determine whether the dimension of the activated two-dimensional image matrix is ​​consistent with the preset dimension of the medium image pixels; S5. If there is no consistency, the activated two-dimensional image matrix is ​​input into the second convolutional layer, and S3-S4 are repeated to continue adjusting the dimension of the two-dimensional image matrix. S6. If they match, the activated two-dimensional image matrix is ​​input to the output layer to obtain the medium image of the non-cooperative target.

2. The method according to claim 1, characterized in that, The three-dimensional scattering center of the medium-free image is generated by the following formula: in, In the formula, The radar echo signal corresponding to the mediumless image. For wave number, The central wave number, For radar observation frequency, Center frequency, Where is the speed of light, and M is the number of scattering centers. The scattering amplitude coefficient of the m-th scattering center. Let m be the radial position of the m-th scattering center. This is a type parameter, and its value must be a multiple of 0.

5. For noise, For imaginary numbers, It is a unit vector.

3. The method according to claim 1, characterized in that, The activation function in the activation function layer is the ReLU function.

4. The method according to claim 1, characterized in that, The medium image of the non-cooperative target is represented as follows: in, In the formula, The radar echo signal corresponding to the medium image. For wave number, The central wave number, For radar observation frequency, Center frequency, Where is the speed of light, and M is the number of scattering centers. The scattering amplitude coefficient of the m-th scattering center. Let m be the radial position of the m-th scattering center. This is a type parameter, and its value must be a multiple of 0.

5. For noise, For imaginary numbers, It is a unit vector. is the variation coefficient of the scattering amplitude coefficient at the m-th scattering center.

5. The method according to any one of claims 1-4, characterized in that, The model was trained and generated in the following manner: Acquire several labeled media-free image samples; wherein the labels are measured two-dimensional images corresponding to the media-free image samples; The media-free image samples are sequentially input into the constructed neural network to generate corresponding two-dimensional images; The network parameters of the neural network are adjusted by using the difference between each two-dimensional image and the corresponding measured two-dimensional image until a model that meets the expectations is obtained.

6. An apparatus for generating a medium image of a non-cooperative target, for implementing the method as described in any one of claims 1-5, characterized in that, include: Acquisition unit, used to acquire media-free images of non-cooperative targets; The generation unit is used to generate several three-dimensional scattering centers of the smooth metallic surface of the medium-free image using a GTD model; wherein the target surface of the medium-free image has no medium coating. An adjustment unit is used to input the three-dimensional scattering centers of the medium-free image into a pre-trained model to adjust the scattering amplitude coefficient of each of the three-dimensional scattering centers, thereby obtaining the medium image of the non-cooperative target.

7. A computing device comprising a memory and a processor, wherein the memory stores a computer program, and the processor, when executing the computer program, implements the method as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method of any one of claims 1-5.

Images