A remote sensing countermeasure defense method and system based on consistency regularization

Abstract

The present invention discloses a remote sensing adversarial defense method and system based on consistency regularization, the method comprising: obtaining remote sensing image information to be processed; pre-processing the remote sensing image information to be processed to obtain initial processed remote sensing image information; processing the initial processed remote sensing image information using a remote sensing adversarial defense processing model to obtain target detection image result information.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a remote sensing confrontation defense method and system based on consistency regularization. Background Art

[0002] In recent years, methods based on convolutional neural networks have achieved great success in remote sensing images. Compared with traditional recognition methods, they have better feature extraction and object positioning capabilities, faster speed and higher accuracy, and have broad application prospects in various fields. However, convolutional neural networks are susceptible to adversarial attacks. By adding small perturbations to the input image, generating adversarial images and inputting them into the network, the network will output incorrect prediction results. Therefore, a remote sensing adversarial defense method and system based on consistency regularization is provided to resist the degradation of network prediction performance caused by adversarial attacks, improve the image prediction accuracy under adversarial attacks, and reduce the incorrect prediction results of adversarial images. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a remote sensing adversarial defense method and system based on consistency regularization, which is beneficial to resist the degradation of network prediction performance caused by adversarial attacks, improve the image prediction accuracy under adversarial attacks, and reduce the erroneous prediction results of adversarial images.

[0004] In order to solve the above technical problems, a first aspect of an embodiment of the present invention discloses a processing method for remote sensing countermeasure defense, the method comprising:

[0005] Obtain remote sensing image information to be processed;

[0006] Pre-processing the remote sensing image information to be processed to obtain initial processed remote sensing image information;

[0007] The initial processed remote sensing image information is processed using a remote sensing adversarial defense processing model to obtain target detection image result information.

[0008] A second aspect of an embodiment of the present invention discloses a processing system for remote sensing countermeasure defense, the system comprising:

[0009] An acquisition module is used to acquire remote sensing image information to be processed;

[0010] A first processing module is used to pre-process the remote sensing image information to be processed to obtain initial processed remote sensing image information;

[0011] The second processing module is used to process the initial processed remote sensing image information using the remote sensing countermeasure defense processing model to obtain target detection image result information.

[0012] The third aspect of the present invention discloses another processing system for remote sensing countermeasure defense, the system comprising:

[0013] A memory storing executable program code;

[0014] a processor coupled to the memory;

[0015] The processor calls the executable program code stored in the memory to execute part or all of the steps in the processing method for remote sensing countermeasure defense disclosed in the first aspect of the embodiment of the present invention.

[0016] The fourth aspect of the present invention discloses a computer-readable storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute some or all of the steps in the processing method for remote sensing countermeasure defense disclosed in the first aspect of an embodiment of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

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

[0018] Figure 1 is a schematic diagram of a scenario of a processing system for remote sensing countermeasure defense provided by an embodiment of the present invention;

[0019] Figure 2 It is a flowchart of a processing method for remote sensing countermeasure defense disclosed in an embodiment of the present invention;

[0020] Figure 3 It is a structural schematic diagram of a processing system for remote sensing countermeasure defense disclosed in an embodiment of the present invention;

[0021] Figure 4 is a schematic diagram of the structure of another processing system for remote sensing countermeasure defense disclosed in an embodiment of the present invention;

[0022] Figure 5 It is a flow chart of a remote sensing countermeasure defense processing model disclosed in an embodiment of the present invention;

[0023] Figure 6 It is a schematic diagram of a flow chart of a detection and activation module disclosed in an embodiment of the present invention;

[0024] Figure 7 It is a flow chart of a second processing model disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or devices.

[0027] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0028] In this application, the word "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described in this application as "exemplary" is not necessarily to be construed as being preferred or advantageous over other embodiments. The following description is given to enable any technician in the field to implement and use the present application. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present application can be implemented without using these specific details. In other instances, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present application with unnecessary details. Therefore, the present application is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.

[0029] It should be noted that since the method of the embodiment of the present application is executed in a computer device, the processing objects of each computer device exist in the form of data or information. For example, time is actually time information. It can be understood that if size, quantity, position, etc. are mentioned in subsequent embodiments, they are all corresponding data for processing by the computer device. The details will not be repeated here.

[0030] The embodiments of the present application provide a processing method, system, computer device, and computer-readable storage medium for remote sensing countermeasure defense, which are described in detail below.

[0031] See also Figure 1 , Figure 1 A schematic diagram of a scenario of a processing system for remote sensing countermeasure defense provided in an embodiment of the present application. The processing system for remote sensing countermeasure defense may include a computer device 100, in which the processing system for remote sensing countermeasure defense is integrated, such as Figure 1 Computer equipment in.

[0032] In the embodiment of the present application, the computer device 100 is mainly used to obtain remote sensing image information to be processed;

[0033] Pre-processing the remote sensing image information to be processed to obtain initial processed remote sensing image information;

[0034] The initial processed remote sensing image information is processed using a remote sensing countermeasure defense processing model to obtain target detection image result information.

[0035] It can resist the degradation of network prediction performance caused by adversarial attacks, improve the image prediction accuracy under adversarial attacks, and reduce the erroneous prediction results of adversarial images.

[0036] In the embodiment of the present application, the computer device 100 may be an independent server, or a server network or server cluster composed of servers. For example, the computer device 100 described in the embodiment of the present application includes but is not limited to a computer, a network host, a single network server, a plurality of network server sets or a cloud server composed of a plurality of servers. The cloud server is composed of a large number of computers or network servers based on cloud computing.

[0037] It is understandable that the computer device 100 used in the embodiments of the present application may be a device including both receiving and transmitting hardware, that is, a device having receiving and transmitting hardware capable of performing two-way communication on a two-way communication link. Such a device may include: a cellular or other communication device having a single-line display or a multi-line display or a cellular or other communication device without a multi-line display. The specific computer device 100 may be a desktop terminal or a mobile terminal, and the computer device 100 may also be one of a mobile phone, a tablet computer, a laptop computer, etc.

[0038] Those skilled in the art will understand that Figure 1 The application environment shown in the figure is only one application scenario of the present application solution and does not constitute a limitation on the application scenario of the present application solution. Other application environments may also include Figure 1More or less computer equipment as shown in Figure 1 Only one computer device is shown in the figure. It can be understood that the processing system for remote sensing countermeasure defense can also include one or more other services, which are not limited here.

[0039] In addition, if Figure 1 As shown, the processing system for remote sensing countermeasure defense may also include a memory 200 for storing data, such as image data, location information, etc.

[0040] It should be noted that Figure 1 The scenario diagram of the processing system for remote sensing countermeasure defense shown is merely an example. The processing system and scenario for remote sensing countermeasure defense described in the embodiment of the present application are intended to more clearly illustrate the technical solution of the embodiment of the present application, and do not constitute a limitation on the technical solution provided in the embodiment of the present application. A person of ordinary skill in the art can know that with the evolution of the processing system for remote sensing countermeasure defense and the emergence of new business scenarios, the technical solution provided in the embodiment of the present application is also applicable to similar technical problems.

[0041] The present invention discloses a remote sensing adversarial defense method and system based on consistency regularization, which is beneficial to resist the degradation of network prediction performance caused by adversarial attacks, improve the image prediction accuracy under adversarial attacks, and reduce the erroneous prediction results of adversarial images. The following are detailed descriptions.

[0042] Embodiment 1

[0043] See also Figure 2 , Figure 2 : is a flowchart of a processing method for remote sensing countermeasure defense disclosed in an embodiment of the present invention. Figure 2 The processing method for remote sensing countermeasure defense described is applied to a management system, such as a local server or a cloud server for management, and the embodiments of the present invention do not limit this. Figure 2 As shown, the processing method for remote sensing countermeasure defense may include the following operations:

[0044] 101. Obtain remote sensing image information to be processed.

[0045] 102. Pre-process the remote sensing image information to be processed to obtain initial processed remote sensing image information.

[0046] 103. The remote sensing adversarial defense processing model is used to process the initial remote sensing image information to obtain the target detection image result information.

[0047] It should be noted that the processing method for remote sensing adversarial defense in this application adopts multi-dimensional adversarial convolution to solve the problem of feature changes in remote sensing images caused by complex adversarial attacks. Feature extraction is performed on clean images and adversarial images in all dimensions to obtain rich contextual information. The model parameters and convolution kernels are adaptively adjusted to effectively extract consistent features of different images while controlling the network scale, reduce the interference of adversarial attacks, and significantly improve the feature extraction capability of convolution.

[0048] In this optional embodiment, as an optional implementation manner, the above-mentioned pre-processing of the remote sensing image information to be processed to obtain the initial processed remote sensing image information includes:

[0049] Performing a first filtering process on the remote sensing image information to be processed to obtain first filtered remote sensing image information;

[0050] Performing a second filtering process on the first filtered remote sensing image information to obtain second filtered remote sensing image information;

[0051] The second filtered remote sensing image information is subjected to image scale adjustment processing to obtain initial processed remote sensing image information.

[0052] It should be noted that the filtering frequencies of the first filtering and the second filtering are inconsistent, wherein the first filtering may be a high-frequency filtering and the second filtering may be a low-frequency filtering, which is not limited in the embodiment of the present invention.

[0053] It should be noted that the above image resizing process is to resize the image to a size of 256*256, which is not limited in the embodiment of the present invention.

[0054] It can be seen that implementing the processing method for remote sensing adversarial defense described in the embodiment of the present invention is beneficial to resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0055] In an optional embodiment, if Figure 5 As shown, the remote sensing countermeasure defense processing model includes a first feature extraction module, a second feature extraction module, a third feature extraction module, a fourth feature extraction module, a first convolution module, a second convolution module, a third convolution module, a fourth convolution module, a fifth convolution module, a sixth convolution module, a seventh convolution module, an eighth convolution module, a ninth convolution module, a tenth convolution module, an eleventh convolution module, a twelfth convolution module, a thirteenth convolution module, a first activation module, a second activation module, a third activation module, a fourth activation module, a fifth activation module, a sixth activation module, a seventh activation module, a first fusion module and a detection activation module; wherein,

[0056] The input end of the first fusion module is connected to the first model input and the second model input of the remote sensing countermeasure defense processing model, and the output end of the first fusion module is connected to the input end of the first feature extraction module; the output end of the first feature extraction module is connected to the input end of the second feature extraction module; the output end of the second feature extraction module is connected to the input end of the first convolution module; the output end of the first convolution module is connected to the input end of the first activation module; the output end of the first activation module is connected to the input end of the third feature extraction module; the output end of the third feature extraction module is connected to the input end of the second convolution module; the output end of the second convolution module is connected to the input end of the second activation module; the output end of the second activation module is connected to the input end of the fourth feature extraction module; the output end of the fourth feature extraction module is respectively connected to the input end of the third convolution module and the input end of the detection activation module; the output end of the third convolution module is connected to the input end of the third activation module; the output end of the third activation module is connected to the input end of the fourth convolution module; the output end of the fourth convolution module is connected to the input end of the fourth activation module; the output end of the fourth activation module is connected to the input end of the fifth convolution module; the output end of the fifth convolution module is connected to the input end of the fifth activation module ; The output end of the fifth activation module is connected to the input end of the first connection module; the output end of the first connection module is connected to the input end of the sixth activation module; the output end of the sixth activation module is connected to the input end of the second connection module; the output end of the second connection module is connected to the input end of the seventh activation module; the output end of the seventh activation module is respectively connected to the input end of the sixth convolution module and the input end of the detection activation module; the output end of the sixth convolution module is connected to the input end of the seventh convolution module; the output end of the seventh convolution module is connected to the input end of the eighth convolution module; the output end of the eighth convolution module is respectively connected to the input end of the ninth convolution module and the input end of the detection activation module; the output end of the ninth convolution module is connected to the input end of the tenth convolution module; the output end of the tenth convolution module is respectively connected to the input end of the eleventh convolution module and the input end of the detection activation module; the output end of the eleventh convolution module is connected to the input end of the twelfth convolution module; the output end of the twelfth convolution module is respectively connected to the input end of the thirteenth convolution module and the input end of the detection activation module; the output end of the thirteenth convolution module is connected to the input end of the detection activation module; the output end of the detection activation module is connected to the first model output of the remote sensing countermeasure defense processing model.

[0057] It should be noted that the remote sensing adversarial defense processing model can adaptively and dynamically perform feature extraction, effectively learn the robust features of clean images and corresponding adversarial images, not only obtain the remote sensing image context information in multiple dimensions, but also efficiently extract the shared features and specific features between clean images and adversarial images, significantly improving the feature extraction capability of objects under adversarial interference.

[0058] It should be noted that the model structures of the first feature extraction module, the second feature extraction module, the third feature extraction module, and the fourth feature extraction module are consistent, and the embodiment of the present invention does not limit this.

[0059] It should be noted that the first convolution module and the second convolution module are both multi-dimensional adversarial convolution layers, and the convolution kernel size is 3×3, which is not limited in the embodiment of the present invention. Further, the first convolution module and the second convolution module are used to learn the objective difference between the clean image and the adversarial image.

[0060] It should be noted that the above-mentioned third convolution module, fourth convolution module, fifth convolution module, seventh convolution module, ninth convolution module, eleventh convolution module and thirteenth convolution module are all constructed based on ordinary convolution layers, and their convolution kernel size is 3×3, which is not limited in the embodiments of the present invention.

[0061] It should be noted that the sixth convolution module, the eighth convolution module, the tenth convolution module, and the twelfth convolution module are all constructed based on ordinary convolution layers, and the convolution kernel size is 1×1, which is not limited in the embodiment of the present invention.

[0062] It should be noted that the above-mentioned first activation module, second activation module, third activation module, fourth activation module, fifth activation module, sixth activation module and seventh activation module are all constructed based on the ReLU activation function, which is not limited in the embodiment of the present invention.

[0063] It should be noted that the above-mentioned first fusion module is an element-by-element addition fusion operation, which is not limited in the embodiment of the present invention.

[0064] It can be seen that implementing the processing method for remote sensing adversarial defense described in the embodiment of the present invention is beneficial to resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0065] In another optional embodiment, the first feature extraction module includes a first convolution unit, a second convolution unit, a first activation unit, a second activation unit and a first pooling unit; wherein,

[0066] The output end of the first fusion module is connected to the input end of the first convolution unit, the output end of the first convolution unit is connected to the input end of the first activation unit; the output end of the first activation unit is connected to the input end of the second convolution unit; the output end of the second convolution unit is connected to the input end of the second activation unit; the output end of the second activation unit is connected to the input end of the first pooling unit; the output end of the first pooling unit is connected to the input end of the second feature extraction module.

[0067] It should be noted that the first convolution unit and the second convolution unit are both multi-dimensional adversarial convolution layers, and the convolution kernel size is 3×3, which is not limited in the embodiment of the present invention. Furthermore, the first convolution unit and the second convolution unit are used to learn the objective difference between the clean image and the adversarial image. The multi-dimensional adversarial convolution has more layers than the ordinary convolution operation, so it is not applied to all convolution layers, and effectively improves the feature extraction capability under the premise of balancing the model scale and performance.

[0068] It should be noted that the above-mentioned first activation unit and second activation unit are constructed based on the ReLU activation function, which is not limited in the embodiment of the present invention.

[0069] It should be noted that the above-mentioned first pooling unit is constructed based on the maximum pooling layer, which is not limited in the embodiment of the present invention.

[0070] It should be noted that when the image is disturbed by complex adversarial attacks, the features in the generated adversarial image will change. The above-mentioned first feature extraction module is different from the conventional convolution layer in that it uses the same convolution kernel for all inputs. Instead, it adaptively adjusts the parameters and convolution kernels of different images for feature extraction.

[0071] It can be seen that implementing the processing method for remote sensing adversarial defense described in the embodiment of the present invention is beneficial to resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0072] In yet another optional embodiment, Figure 6 As shown, the detection activation module includes a first normalization unit, a third convolution unit, a fourth convolution unit, a first classification unit, a second classification unit, a filtering unit, and a third activation unit; wherein,

[0073] The input end of the first normalization unit is respectively connected to the first module input, the second module input, the third module input, the fourth module input, the fifth module input and the sixth module input of the detection activation module, and the output end of the first normalization unit is respectively connected to the input end of the third convolution unit and the input end of the fourth convolution unit; the output end of the third convolution unit is connected to the input end of the first classification unit; the output end of the fourth convolution unit is connected to the input end of the second classification unit; the output end of the first classification unit and the output end of the second classification unit are both connected to the input end of the filtering and screening unit; the output end of the filtering and screening unit is connected to the input end of the third activation unit; the output end of the third activation unit is connected to the first model output of the remote sensing countermeasure defense processing model.

[0074] It should be noted that the above-mentioned first module input, second module input, third module input, fourth module input, fifth module input and sixth module input are respectively the output end of the fourth feature extraction module, the output end of the seventh activation module, the output end of the eighth convolution module, the output end of the tenth convolution module, the output end of the twelfth convolution module and the output end of the thirteenth convolution module, which is not limited in the embodiment of the present invention.

[0075] It should be noted that the above-mentioned first normalization unit is constructed based on regularization, which is not limited in the embodiment of the present invention.

[0076] It should be noted that the first classification unit is constructed based on softmax, which is not limited in the embodiment of the present invention. Further, the first classification unit can perform regression classification prediction on the type corresponding to the prediction box of each grid point of the convolution feature layer output by the third convolution unit.

[0077] It should be noted that the second classification unit is constructed based on bounding box regrssion, which is not limited in the embodiment of the present invention. Further, the second classification unit can perform regression classification prediction on the type corresponding to the prediction box of each grid point of the convolution feature layer output by the fourth convolution unit.

[0078] It should be noted that the above-mentioned filtering and screening unit is constructed based on the non-maximum suppression module, which is not limited in the embodiment of the present invention. Further, the threshold of the non-maximum suppression module is a positive number not less than 0.5 and not greater than 1, so as to prevent misscreening, which is not limited in the embodiment of the present invention.

[0079] It should be noted that the third activation unit is constructed based on the sigmoid activation function, which is not limited in the embodiment of the present invention.

[0080] It should be noted that the convolution kernels of the third convolution unit and the fourth convolution unit are both 1×1, and the number of channels are 21 and 4 respectively, which is not limited in the embodiment of the present invention.

[0081] It can be seen that implementing the processing method for remote sensing adversarial defense described in the embodiment of the present invention is beneficial to resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0082] In yet another optional embodiment, the remote sensing countermeasure defense processing model is obtained based on the following method:

[0083] Obtain initial training sample information;

[0084] Based on the initial training sample information, determining the target training sample information;

[0085] The initial processing model is trained using the target training sample information to obtain a training processing model and training result information; the initial processing model includes a first processing model and a second processing model; the training processing model includes a first training processing model and a second training processing model; the first processing model is consistent with the model architecture of the remote sensing countermeasure defense processing model; the training result information includes first training result information corresponding to the first training processing model and second training result information corresponding to the second training processing model;

[0086] The loss function is used to calculate and process the training result information to obtain the loss function value;

[0087] Determine whether the loss function value meets the training termination condition and obtain the training judgment result;

[0088] When the training judgment result is no, the training processing model is determined as a new initial processing model, and the execution is triggered to determine the target training sample information based on the initial training sample information;

[0089] When the training judgment result is yes, the first training processing model is determined as the remote sensing confrontation defense processing model.

[0090] It should be noted that the above-mentioned acquisition of initial training sample information is to select clear optical remote sensing image slices, annotate the object category and position in the image, mark a rectangular horizontal bounding box along the edge of the object, and save the object category and the coordinates of the upper left corner and lower right corner as labels. For example, the public remote sensing dataset HRSC is used as a sample dataset, which contains a high-resolution optical ship dataset, including 2976 ship images, including two scenes of offshore ships and nearshore ships. The image size ranges from 300×300 to 1500×900 pixels. The dataset is divided into training set, validation set and test set, which contain 436 images and 1207 samples, 181 images and 541 samples, and 444 images and 1228 samples, respectively.

[0091] It should be noted that the target training sample information is determined based on the initial training sample information by randomly selecting one of the initial image sample information in the initial training sample information as the target training sample information, which is not limited in the embodiment of the present invention.

[0092] It should be noted that the above initial processing model is composed of a first processing model and a second processing model, wherein the output end of the fifth activation module of the first processing model extends a port as the input end of the second processing model, which is not limited in the embodiment of the present invention.

[0093] It should be noted that the first training result information represents the output information of the first processing model; further, the second training result information represents the output information of the second processing model. Further, the output information is the image vector corresponding to the image, which is not limited in the embodiment of the present invention.

[0094] It should be noted that the above training termination condition represents the convergence of the data composed of the loss function value and the historical loss function value. Further, the above convergence can represent the fluctuation of the value between 0.001, which is not limited in the embodiment of the present invention.

[0095] It can be seen that implementing the processing method for remote sensing adversarial defense described in the embodiment of the present invention is beneficial to resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0096] In an optional embodiment, if Figure 7 As shown, the second processing model includes a first normalization module, a second normalization module, a third normalization module, a fourteenth convolution module, a fifteenth convolution module, a sixteenth convolution module, a seventeenth convolution module, an eighteenth convolution module, an eighth activation module, a ninth activation module, a tenth activation module and a scaling module; wherein,

[0097] The input end of the first normalization module is connected to the output end of the fifth activation module, and the output end of the first normalization module is connected to the input end of the fourteenth convolution module; the output end of the fourteenth convolution module is connected to the input end of the second normalization module; the output end of the second normalization module is connected to the input end of the eighth activation module; the output end of the eighth activation module is connected to the input end of the sixteenth convolution module; the output end of the sixteenth convolution module is connected to the input end of the third normalization module; the output end of the third normalization module is connected to the input end of the ninth activation module; the output end of the ninth activation module is connected to the input end of the seventeenth convolution module; the output end of the seventeenth convolution module is connected to the input end of the eighteenth convolution module; the output end of the eighteenth convolution module is connected to the input end of the tenth activation module; the output end of the tenth activation module is connected to the input end of the scaling module; the output end of the scaling module is connected to the second model output of the second processing model.

[0098] It should be noted that the above-mentioned first normalization module is constructed based on the L2 regularization rule, which is not limited in the embodiment of the present invention.

[0099] It should be noted that the above-mentioned second normalization module and the third normalization module are constructed based on the batch normalization layer, which is not limited in the embodiment of the present invention.

[0100] It should be noted that the above-mentioned fourteenth convolution module, sixteenth convolution module, and eighteenth convolution module are constructed based on ordinary convolution layers, and the convolution kernel size is 3×3, which is not limited in the embodiment of the present invention.

[0101] It should be noted that the fifteenth convolution module and the seventeenth convolution module are constructed based on the transposed convolution layer, and the convolution kernel size is 3×3, which is used for upsampling / deconvolution operations, and the embodiment of the present invention does not limit this.

[0102] It should be noted that the eighth activation module and the ninth activation module are constructed based on the ReLU activation function, which is not limited in the embodiment of the present invention.

[0103] It should be noted that the tenth activation module is constructed based on the Tanh activation function, which is not limited in the embodiment of the present invention.

[0104] It should be noted that the above-mentioned scaling module is used to perform scaling operations on the input data, and is used to regularize, standardize or normalize the feature map or input data in order to improve the training effect of the model.

[0105] It should be noted that the above-mentioned second processing model is used to constrain the reconstruction features. By extracting consistent features from the mixed distribution, the reconstructed image is used for object classification and positioning, which effectively reduces the interference of adversarial attacks and improves the adversarial robustness of the model.

[0106] It can be seen that implementing the processing method for remote sensing adversarial defense described in the embodiment of the present invention is beneficial to resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0107] In another optional embodiment, the loss function is:

[0108]

[0109] Where LOSS is the loss function value; X1 is the image vector corresponding to the target training sample information; X2 is the image vector corresponding to the second training result information; x i is the i-th vector element in the image vector corresponding to the second training result information; X3 is the image vector corresponding to the first training result information; KL(·) is the Kullback-Leibler divergence; λ1, λ2, λ3 and λ4 are the first loss coefficient, the second loss coefficient, the third loss coefficient and the fourth loss coefficient respectively.

[0110] It should be noted that the function part in the above KL(·) can be constructed based on consistency regularization to compare the distribution of the clean image and the output of the second processing model and perform regularization to ensure their consistency, which is not limited in the embodiment of the present invention.

[0111] It should be noted that the loss function is used to train the network to obtain the remote sensing anti-defense processing model. The model obtained after training is robust, can reduce the impact of adversarial attacks on network prediction results, and effectively resist adversarial attacks.

[0112] It should be noted that the first loss coefficient, the second loss coefficient, the third loss coefficient and the fourth loss coefficient may be set by the user or by a default value given by the system, which is not limited in the embodiment of the present invention. Preferably, the first loss coefficient, the second loss coefficient, the third loss coefficient and the fourth loss coefficient may be 1, 1, 0.16 and 5 respectively, which is not limited in the embodiment of the present invention.

[0113] It can be seen that implementing the processing method for remote sensing adversarial defense described in the embodiment of the present invention is beneficial to resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0114] Embodiment 2

[0115] See also Figure 3 , Figure 3 : is a schematic diagram of a processing system for remote sensing countermeasure defense disclosed in an embodiment of the present invention. Figure 3 The described system can be applied to a management system, such as a local server or a cloud server for management, etc., which is not limited in the embodiments of the present invention. Figure 3 As shown, the system may include:

[0116] An acquisition module 201 is used to acquire remote sensing image information to be processed;

[0117] The first processing module 202 is used to pre-process the remote sensing image information to be processed to obtain initial processed remote sensing image information;

[0118] The second processing module 203 is used to process the initial processed remote sensing image information using the remote sensing countermeasure defense processing model to obtain target detection image result information.

[0119] It can be seen that the implementation Figure 3 The described processing system for remote sensing adversarial defense is beneficial for resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0120] In another optional embodiment, Figure 3As shown, the remote sensing countermeasure defense processing model includes a first feature extraction module, a second feature extraction module, a third feature extraction module, a fourth feature extraction module, a first convolution module, a second convolution module, a third convolution module, a fourth convolution module, a fifth convolution module, a sixth convolution module, a seventh convolution module, an eighth convolution module, a ninth convolution module, a tenth convolution module, an eleventh convolution module, a twelfth convolution module, a thirteenth convolution module, a first activation module, a second activation module, a third activation module, a fourth activation module, a fifth activation module, a sixth activation module, a seventh activation module, a first fusion module and a detection activation module; wherein,

[0121] The input end of the first fusion module is connected to the first model input and the second model input of the remote sensing countermeasure defense processing model, and the output end of the first fusion module is connected to the input end of the first feature extraction module; the output end of the first feature extraction module is connected to the input end of the second feature extraction module; the output end of the second feature extraction module is connected to the input end of the first convolution module; the output end of the first convolution module is connected to the input end of the first activation module; the output end of the first activation module is connected to the input end of the third feature extraction module; the output end of the third feature extraction module is connected to the input end of the second convolution module; the output end of the second convolution module is connected to the input end of the second activation module; the output end of the second activation module is connected to the input end of the fourth feature extraction module; the output end of the fourth feature extraction module is respectively connected to the input end of the third convolution module and the input end of the detection activation module; the output end of the third convolution module is connected to the input end of the third activation module; the output end of the third activation module is connected to the input end of the fourth convolution module; the output end of the fourth convolution module is connected to the input end of the fourth activation module; the output end of the fourth activation module is connected to the input end of the fifth convolution module; the output end of the fifth convolution module is connected to the input end of the fifth activation module ; The output end of the fifth activation module is connected to the input end of the first connection module; the output end of the first connection module is connected to the input end of the sixth activation module; the output end of the sixth activation module is connected to the input end of the second connection module; the output end of the second connection module is connected to the input end of the seventh activation module; the output end of the seventh activation module is respectively connected to the input end of the sixth convolution module and the input end of the detection activation module; the output end of the sixth convolution module is connected to the input end of the seventh convolution module; the output end of the seventh convolution module is connected to the input end of the eighth convolution module; the output end of the eighth convolution module is respectively connected to the input end of the ninth convolution module and the input end of the detection activation module; the output end of the ninth convolution module is connected to the input end of the tenth convolution module; the output end of the tenth convolution module is respectively connected to the input end of the eleventh convolution module and the input end of the detection activation module; the output end of the eleventh convolution module is connected to the input end of the twelfth convolution module; the output end of the twelfth convolution module is respectively connected to the input end of the thirteenth convolution module and the input end of the detection activation module; the output end of the thirteenth convolution module is connected to the input end of the detection activation module; the output end of the detection activation module is connected to the first model output of the remote sensing countermeasure defense processing model.

[0122] It can be seen that the implementation Figure 3 The described processing system for remote sensing adversarial defense is beneficial for resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0123] In yet another optional embodiment, Figure 2 As shown, the first feature extraction module includes a first convolution unit, a second convolution unit, a first activation unit, a second activation unit and a first pooling unit; wherein,

[0124] The output end of the first fusion module is connected to the input end of the first convolution unit, the output end of the first convolution unit is connected to the input end of the first activation unit; the output end of the first activation unit is connected to the input end of the second convolution unit; the output end of the second convolution unit is connected to the input end of the second activation unit; the output end of the second activation unit is connected to the input end of the first pooling unit; the output end of the first pooling unit is connected to the input end of the second feature extraction module.

[0125] It can be seen that the implementation Figure 3 The described processing system for remote sensing adversarial defense is beneficial for resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0126] In yet another optional embodiment, Figure 2 As shown, the detection activation module includes a first normalization unit, a third convolution unit, a fourth convolution unit, a first classification unit, a second classification unit, a filtering unit, and a third activation unit; wherein,

[0127] The input end of the first normalization unit is respectively connected to the first module input, the second module input, the third module input, the fourth module input, the fifth module input and the sixth module input of the detection activation module, and the output end of the first normalization unit is respectively connected to the input end of the third convolution unit and the input end of the fourth convolution unit; the output end of the third convolution unit is connected to the input end of the first classification unit; the output end of the fourth convolution unit is connected to the input end of the second classification unit; the output end of the first classification unit and the output end of the second classification unit are both connected to the input end of the filtering and screening unit; the output end of the filtering and screening unit is connected to the input end of the third activation unit; the output end of the third activation unit is connected to the first model output of the remote sensing countermeasure defense processing model.

[0128] It can be seen that the implementation Figure 2 The described processing system for remote sensing adversarial defense is beneficial for resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0129] In yet another optional embodiment, Figure 3 As shown, the remote sensing countermeasure defense processing model is obtained by the second processing module 203 executing the following method:

[0130] Obtain initial training sample information;

[0131] Based on the initial training sample information, determining the target training sample information;

[0132] The initial processing model is trained using the target training sample information to obtain a training processing model and training result information; the initial processing model includes a first processing model and a second processing model; the training processing model includes a first training processing model and a second training processing model; the first processing model is consistent with the model architecture of the remote sensing countermeasure defense processing model; the training result information includes first training result information corresponding to the first training processing model and second training result information corresponding to the second training processing model;

[0133] The loss function is used to calculate and process the training result information to obtain the loss function value;

[0134] Determine whether the loss function value meets the training termination condition and obtain the training judgment result;

[0135] When the training judgment result is no, the training processing model is determined as a new initial processing model, and the execution is triggered to determine the target training sample information based on the initial training sample information;

[0136] When the training judgment result is yes, the first training processing model is determined as the remote sensing confrontation defense processing model.

[0137] It can be seen that the implementation Figure 3 The described processing system for remote sensing adversarial defense is beneficial for resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0138] In yet another optional embodiment, Figure 2 As shown, the second processing model includes a first normalization module, a second normalization module, a third normalization module, a fourteenth convolution module, a fifteenth convolution module, a sixteenth convolution module, a seventeenth convolution module, an eighteenth convolution module, an eighth activation module, a ninth activation module, a tenth activation module and a scaling module; wherein,

[0139] The input end of the first normalization module is connected to the output end of the fifth activation module, and the output end of the first normalization module is connected to the input end of the fourteenth convolution module; the output end of the fourteenth convolution module is connected to the input end of the second normalization module; the output end of the second normalization module is connected to the input end of the eighth activation module; the output end of the eighth activation module is connected to the input end of the sixteenth convolution module; the output end of the sixteenth convolution module is connected to the input end of the third normalization module; the output end of the third normalization module is connected to the input end of the ninth activation module; the output end of the ninth activation module is connected to the input end of the seventeenth convolution module; the output end of the seventeenth convolution module is connected to the input end of the eighteenth convolution module; the output end of the eighteenth convolution module is connected to the input end of the tenth activation module; the output end of the tenth activation module is connected to the input end of the scaling module; the output end of the scaling module is connected to the second model output of the second processing model.

[0140] It can be seen that the implementation Figure 3 The described processing system for remote sensing adversarial defense is beneficial for resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0141] In yet another optional embodiment, Figure 2 As shown, the loss function is:

[0142]

[0143] Where LOSS is the loss function value; X1 is the image vector corresponding to the target training sample information; X2 is the image vector corresponding to the second training result information; x i is the i-th vector element in the image vector corresponding to the second training result information; X3 is the image vector corresponding to the first training result information; KL(·) is the Kullback-Leibler divergence; λ1, λ2, λ3 and λ4 are the first loss coefficient, the second loss coefficient, the third loss coefficient and the fourth loss coefficient respectively.

[0144] It can be seen that the implementation Figure 3 The described processing system for remote sensing adversarial defense is beneficial for resisting the degradation of network prediction performance caused by adversarial attacks, improving the image prediction accuracy under adversarial attacks, and reducing the erroneous prediction results of adversarial images.

[0145] Embodiment 3

[0146] See also Figure 4 , Figure 4 is a schematic diagram of the structure of another processing system for remote sensing countermeasure defense disclosed in an embodiment of the present invention. Figure 4 The described system can be applied to a management system, such as a local server or a cloud server for management, etc., which is not limited in the embodiments of the present invention. Figure 3 As shown, the system may include:

[0147] A memory 301 storing executable program codes;

[0148] a processor 302 coupled to the memory 301;

[0149] The processor 302 calls the executable program code stored in the memory 301 to execute the steps in the processing method for remote sensing countermeasure defense described in the first embodiment.

[0150] Embodiment 4

[0151] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the processing method for remote sensing countermeasure defense described in the first embodiment.

[0152] Embodiment 5

[0153] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps in the processing method for remote sensing countermeasure defense described in Embodiment 1.

[0154] The system embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative labor.

[0155] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0156] Finally, it should be noted that the remote sensing countermeasures defense method and system based on consistency regularization disclosed in the embodiment of the present invention discloses only the preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, ordinary technicians in this field should understand that it is still possible to modify the technical solutions recorded in the aforementioned embodiments, or to make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A processing method for remote sensing countermeasure defense, characterized in that the method include: Obtain remote sensing image information to be processed; Pre-processing the remote sensing image information to be processed to obtain initial processed remote sensing image information; The pre-processing of the remote sensing image information to be processed to obtain the initial processed remote sensing image information includes: Performing a first filtering process on the remote sensing image information to be processed to obtain first filtered remote sensing image information; Performing a second filtering process on the first filtered remote sensing image information to obtain second filtered remote sensing image information; Performing image scale adjustment processing on the second filtered remote sensing image information to obtain initial processed remote sensing image information; The remote sensing countermeasure defense processing model is used to process the initial remote sensing image information to obtain the target detection image result information; Among them, the remote sensing countermeasure defense processing model includes a first feature extraction module, a second feature extraction module, a third feature extraction module, a fourth feature extraction module, a first convolution module, a second convolution module, a third convolution module, a fourth convolution module, a fifth convolution module, a sixth convolution module, a seventh convolution module, an eighth convolution module, a ninth convolution module, a tenth convolution module, an eleventh convolution module, a twelfth convolution module, a thirteenth convolution module, a first activation module, a second activation module, a third activation module, a fourth activation module, a fifth activation module, a sixth activation module, a seventh activation module, a first fusion module and a detection activation module; the model structures of the first feature extraction module, the second feature extraction module, the third feature extraction module and the fourth feature extraction module are consistent; the first convolution module, the second convolution module, the third convolution module, the fourth convolution module, the fifth convolution module, the sixth convolution module, the seventh convolution module, the thirteenth convolution module, the first activation module, the second activation module, the third activation module, the fourth activation module, the fifth activation module, the sixth activation module, the seventh activation module, the first fusion module and the ... The convolution modules are all multi-dimensional adversarial convolutional layers, and their convolution kernel size is 3×3. The first convolution module and the second convolution module are used to learn the objective difference between clean images and adversarial images; the third convolution module, the fourth convolution module, the fifth convolution module, the seventh convolution module, the ninth convolution module, the eleventh convolution module, and the thirteenth convolution module are all built based on ordinary convolutional layers, and the convolution kernel size is 3×3; the sixth convolution module, the eighth convolution module, the tenth convolution module, and the twelfth convolution module are all built based on ordinary convolutional layers, and the convolution kernel size is 1×1; the first activation module, the second activation module, the third activation module, the fourth activation module, the fifth activation module, the sixth activation module, and the seventh activation module are all built based on the ReLU activation function; the first fusion module is an element-by-element addition fusion operation; in, The input end of the first fusion module is connected to the first model input and the second model input of the remote sensing countermeasure defense processing model, and the output end of the first fusion module is connected to the input end of the first feature extraction module; the output end of the first feature extraction module is connected to the input end of the second feature extraction module; the output end of the second feature extraction module is connected to the input end of the first convolution module; the output end of the first convolution module is connected to the input end of the first activation module; the output end of the first activation module is connected to the input end of the third feature extraction module; the output end of the third feature extraction module is connected to the input end of the second convolution module; the output end of the second convolution module is connected to the input end of the second activation module; the output end of the second activation module is connected to the input end of the fourth feature extraction module; the output end of the fourth feature extraction module is respectively connected to the input end of the third convolution module and the input end of the detection activation module; the output end of the third convolution module is connected to the input end of the third activation module; the output end of the third activation module is connected to the input end of the fourth convolution module; the output end of the fourth convolution module is connected to the input end of the fourth activation module; the output end of the fourth activation module is connected to the input end of the fifth convolution module; the output end of the fifth convolution module is connected to the input end of the fifth activation module ; The output end of the fifth activation module is connected to the input end of the first connection module; the output end of the first connection module is connected to the input end of the sixth activation module; the output end of the sixth activation module is connected to the input end of the second connection module; the output end of the second connection module is connected to the input end of the seventh activation module; the output end of the seventh activation module is respectively connected to the input end of the sixth convolution module and the input end of the detection activation module; the output end of the sixth convolution module is connected to the input end of the seventh convolution module; the output end of the seventh convolution module is connected to the input end of the eighth convolution module; the output end of the eighth convolution module is respectively connected to the input end of the ninth convolution module and the input end of the detection activation module; the output end of the ninth convolution module is connected to the input end of the tenth convolution module; the output end of the tenth convolution module is respectively connected to the input end of the eleventh convolution module and the input end of the detection activation module; the output end of the eleventh convolution module is connected to the input end of the twelfth convolution module; the output end of the twelfth convolution module is respectively connected to the input end of the thirteenth convolution module and the input end of the detection activation module; the output end of the thirteenth convolution module is connected to the input end of the detection activation module; the output end of the detection activation module is connected to the first model output of the remote sensing countermeasure defense processing model; Among them, the first feature extraction module includes a first convolution unit, a second convolution unit, a first activation unit, a second activation unit and a first pooling unit; the first convolution unit and the second convolution unit are both multi-dimensional adversarial convolution layers, and the convolution kernel size is 3×3; the first activation unit and the second activation unit are constructed based on the ReLU activation function; the first pooling unit is constructed based on the maximum pooling layer; in, The output end of the first fusion module is connected to the input end of the first convolution unit, the output end of the first convolution unit is connected to the input end of the first activation unit; the output end of the first activation unit is connected to the input end of the second convolution unit; the output end of the second convolution unit is connected to the input end of the second activation unit; the output end of the second activation unit is connected to the input end of the first pooling unit; the output end of the first pooling unit is connected to the input end of the second feature extraction module; Among them, the detection activation module includes a first normalization unit, a third convolution unit, a fourth convolution unit, a first classification unit, a second classification unit, a filtering and screening unit, and a third activation unit; the first module input, the second module input, the third module input, the fourth module input, the fifth module input and the sixth module input are respectively the output end of the fourth feature extraction module, the output end of the seventh activation module, the output end of the eighth convolution module, the output end of the tenth convolution module, the output end of the twelfth convolution module and the output end of the thirteenth convolution module; the first normalization unit is constructed based on regularization; the first classification unit is constructed based on softmax; the second classification unit is constructed based on bounding box regrssion; the filtering and screening unit is constructed based on the non-maximum suppression module; the third activation unit is constructed based on the sigmoid activation function; the convolution kernels of the third convolution unit and the fourth convolution unit are both 1×1; in, The input end of the first normalization unit is respectively connected to the first module input, the second module input, the third module input, the fourth module input, the fifth module input and the sixth module input of the detection activation module, and the output end of the first normalization unit is respectively connected to the input end of the third convolution unit and the input end of the fourth convolution unit; the output end of the third convolution unit is connected to the input end of the first classification unit; the output end of the fourth convolution unit is connected to the input end of the second classification unit; the output end of the first classification unit and the output end of the second classification unit are both connected to the input end of the filtering and screening unit; the output end of the filtering and screening unit is connected to the input end of the third activation unit; the output end of the third activation unit is connected to the first model output of the remote sensing countermeasure defense processing model.

2. The processing method for remote sensing countermeasure defense according to claim 1, characterized in that: The remote sensing countermeasure defense processing model is based on the following method: Obtain initial training sample information; Based on the initial training sample information, determining the target training sample information; The initial processing model is trained using target training sample information to obtain a training processing model and training result information; the initial processing model includes a first processing model and a second processing model; the training processing model includes a first training processing model and a second training processing model; the first processing model is consistent with the model architecture of the remote sensing countermeasure defense processing model; The training result information includes first training result information corresponding to the first training processing model and second training result information corresponding to the second training processing model; The initial processing model is composed of a first processing model and a second processing model, wherein the output end of the fifth activation module of the first processing model extends a port as an input end of the second processing model; The loss function is used to calculate and process the training result information to obtain the loss function value; Determine whether the loss function value meets the training termination condition and obtain the training judgment result; When the training judgment result is no, the training processing model is determined as a new initial processing model, and the execution is triggered to determine the target training sample information based on the initial training sample information; When the training judgment result is yes, the first training processing model is determined as the remote sensing confrontation defense processing model.

3. The processing method for remote sensing countermeasure defense according to claim 2, characterized in that: The second processing model includes a first normalization module, a second normalization module, a third normalization module, a fourteenth convolution module, a fifteenth convolution module, a sixteenth convolution module, a seventeenth convolution module, an eighteenth convolution module, an eighth activation module, a ninth activation module, a tenth activation module and a scaling module; the first normalization module is constructed based on the L2 regularization rule; the second normalization module and the third normalization module are constructed based on the batch normalization layer; the fourteenth convolution module, the sixteenth convolution module and the eighteenth convolution module are constructed based on the ordinary convolution layer, and the convolution kernel size is 3×3; the fifteenth convolution module and the seventeenth convolution module are constructed based on the transposed convolution layer, and the convolution kernel size is 3×3; the eighth activation module and the ninth activation module are constructed based on the ReLU activation function; the tenth activation module is constructed based on the Tanh activation function; the scaling module is used to scale the input data, and is used to normalize, standardize or normalize the feature map or the input data; in, The input end of the first normalization module is connected to the output end of the fifth activation module, and the output end of the first normalization module is connected to the input end of the fourteenth convolution module; the output end of the fourteenth convolution module is connected to the input end of the second normalization module; the output end of the second normalization module is connected to the input end of the eighth activation module; the output end of the eighth activation module is connected to the input end of the sixteenth convolution module; the output end of the sixteenth convolution module is connected to the input end of the third normalization module; the output end of the third normalization module is connected to the input end of the ninth activation module; the output end of the ninth activation module is connected to the input end of the seventeenth convolution module; the output end of the seventeenth convolution module is connected to the input end of the eighteenth convolution module; the output end of the eighteenth convolution module is connected to the input end of the tenth activation module; the output end of the tenth activation module is connected to the input end of the scaling module; the output end of the scaling module is connected to the second model output of the second processing model.

4. The processing method for remote sensing countermeasure defense according to claim 2, characterized in that: The loss function is: Where LOSS is the loss function value; X1 is the image vector corresponding to the target training sample information; X2 is the image vector corresponding to the second training result information; x i is the i-th vector element in the image vector corresponding to the second training result information; X3 is the image vector corresponding to the first training result information; KL(·) is the Kullback-Leibler divergence; λ1, λ2, λ3 and λ4 are the first loss coefficient, the second loss coefficient, the third loss coefficient and the fourth loss coefficient respectively.

5. A processing system for remote sensing countermeasure defense, characterized in that: The system includes: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the processing method for remote sensing countermeasure defense as claimed in any one of claims 1 to 4.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, which, when called, are used to execute the processing method for remote sensing countermeasure defense as claimed in any one of claims 1 to 4.

Images