ISAR image classification method and system based on CycleMLP model

Abstract

The invention discloses an ISAR image classification method and system based on a CycleMLP model, and relates to the field of inverse synthetic aperture radar ISAR image classification, and the method comprises the steps: carrying out the preprocessing of a plurality of ISAR images, and obtaining a minimum data set; expanding the minimum data set to obtain an expanded data set; a CycleMLP model comprising a block extraction layer, a CycleMLP group and a classifier is constructed, the CycleMLP model is trained based on the expanded data set, and the trained CycleMLP model is obtained; and the trained CycleMLP model is used to classify the target ISAR image. The ISAR image classification accuracy and timeliness can be improved, and high practicability is achieved.

Description

Background Art

[0002] Radar is a detection device that can detect objects in all weather conditions, day and night, and is not blocked by fog, clouds, or rain. In terms of space situational awareness, large-scale inverse synthetic aperture radars with large bandwidth, high frequency band, and long distance can achieve precise detection of space objects and high-resolution ISAR imaging, and depict information such as the rich electromagnetic scattering characteristics of the target. ISAR image classification is an important part of realizing the cognition of space objects. At present, the classification of space objects using two-dimensional ISAR images of space objects is at a turning point from relying on manual feature extraction to intelligent classification. Although traditional methods can realize space object recognition, the process is complicated and consumes a lot of time and energy. In order to realize the intelligent development of space object classification, deep learning technology has emerged in the task of space object classification. Compared with the manual feature extraction method, the deep learning algorithm has the advantages of automatically learning deep modal features and obtaining higher classification accuracy due to its multi-layer network structure.

[0003] As deep learning models are constantly evolving, the Vit series of models represented by the Transformer architecture have performed well in the CV field. However, due to the characteristics of the global attention calculation mechanism, when inputting high-resolution images, the amount of calculation is huge. In recent years, the series of models represented by CycleMLP replace the attention calculation module inside the Transformer with the MLP fully connected structure. Under this model framework, it can reduce the model calculation complexity without affecting the task performance, improve the model's reasoning speed, and have good image size adaptability. Summary of the invention

[0004] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art and specifically provide an ISAR image classification method and system based on the CycleMLP model to improve the accuracy and timeliness of ISAR image classification.

[0005] 1) In the first aspect, the present invention provides an ISAR image classification method based on the CycleMLP model, and the specific technical solution is as follows:

[0006] Preprocess multiple ISAR images to obtain the minimum data set;

[0007] Expand the minimum data set to obtain an expanded data set;

[0008] A CycleMLP model including a block extraction layer, a CycleMLP group and a classifier is constructed. The block extraction layer is used to process each ISAR image in the expanded dataset to obtain multiple final image blocks. The CycleMLP group is used to perform multi-dimensional feature fusion processing on the final image blocks. The classifier is used to receive and process the fused features and output the classification results.

[0009] The CycleMLP model is trained based on the expanded data set to obtain a trained CycleMLP model;

[0010] The trained CycleMLP model is used to classify the target ISAR image.

[0011] On the basis of the above scheme, the ISAR image classification method based on the CycleMLP model of the present invention can also be improved as follows.

[0012] Furthermore, the ISAR images are preprocessed to obtain the minimum data set, including:

[0013] The ISAR image is processed by edge removal to remove the parts irrelevant to the target in the ISAR image, and the minimum data set is obtained. The minimum data set contains P1 type of targets, each type of target contains no less than P2 subdivided targets, and each subdivided target has no less than P3 images in different postures.

[0014] Furthermore, the target-irrelevant parts include the white edges and annotation information of the ISAR image;

[0015] The annotation information includes the x-axis, y-axis, distance, azimuth, and elevation.

[0016] Further, the minimum data set is expanded to include:

[0017] The ISAR images are expanded by adding noise, non-uniform scaling, flipping and color gamut distortion.

[0018] Furthermore, each ISAR image in the expanded dataset is processed to obtain multiple final image blocks, including:

[0019] Each ISAR image in the expanded dataset is stacked and divided according to the partition window to obtain multiple initial image blocks. The channel features of each initial image block are dimensionally upgraded through a linear coding layer, and then each initial image block after dimension upgrade is downsampled four times using strided convolution to obtain multiple final image blocks.

[0020] Furthermore, the CycleMLP group includes N concatenated CycleMLP blocks, and each CycleMLP block includes: a CycleFC layer and a ChannelMLP layer.

[0021] 2) In the second aspect, the present invention also provides an ISAR image classification system based on the CycleMLP model, and the specific technical solution is as follows:

[0022] It includes a data set construction module, a data set expansion module, a model construction module, a training module and a classification module;

[0023] The dataset construction module is used to: pre-process multiple ISAR images to obtain the minimum dataset;

[0024] The data set expansion module is used to: expand the minimum data set to obtain an expanded data set;

[0025] The model building module is used to build a CycleMLP model including a block extraction layer, a CycleMLP group and a classifier. The block extraction layer is used to process each ISAR image in the expanded data set to obtain multiple final image blocks. The CycleMLP group is used to perform multi-dimensional feature fusion processing on the final image blocks. The classifier is used to receive and process the fused features and output the classification results.

[0026] The training module is used to: train the CycleMLP model based on the expanded data set to obtain a trained CycleMLP model;

[0027] Classification module: Use the trained CycleMLP model to classify the target ISAR image.

[0028] On the basis of the above scheme, the ISAR image classification system based on the CycleMLP model of the present invention can also be improved as follows.

[0029] Furthermore, the dataset construction module is specifically used to:

[0030] The ISAR image is processed by edge removal to remove the parts irrelevant to the target in the ISAR image, and the minimum data set is obtained. The minimum data set contains P1 type of targets, each type of target contains no less than P2 subdivided targets, and each subdivided target has no less than P3 images in different postures.

[0031] Furthermore, the target-irrelevant parts include the white edges and annotation information of the ISAR image;

[0032] The annotation information includes the x-axis, y-axis, distance, azimuth, and elevation.

[0033] Further, the minimum data set is expanded to include:

[0034] The ISAR images are expanded by adding noise, non-uniform scaling, flipping and color gamut distortion.

[0035] Furthermore, each ISAR image in the expanded dataset is processed to obtain multiple final image blocks, including:

[0036] Each ISAR image in the expanded dataset is stacked and divided according to the partition window to obtain multiple initial image blocks. The channel features of each initial image block are dimensionally upgraded through a linear coding layer, and then each initial image block after dimension upgrade is downsampled four times using strided convolution to obtain multiple final image blocks.

[0037] Furthermore, the CycleMLP group includes N concatenated CycleMLP blocks, and each CycleMLP block includes: a CycleFC layer and a ChannelMLP layer.

[0038] 3) In a third aspect, the present invention further provides an electronic device, comprising a processor, the processor being coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor, so that the electronic device implements any one of the above-mentioned ISAR image classification methods based on the CycleMLP model.

[0039] 4) In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, any one of the above-mentioned ISAR image classification methods based on the CycleMLP model is implemented.

[0040] The beneficial technical effects of the present invention are:

[0041] The present invention introduces an ISAR image expansion mechanism and realizes image classification through a trained CycleMLP network model, thereby improving the accuracy and timeliness of ISAR image classification and having strong practicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic diagram of a process of an ISAR image classification method based on a CycleMLP model according to an embodiment of the present invention;

[0043] Figure 2 This is a processing flow framework diagram of the ISAR image classification method based on the CycleMLP model;

[0044] Figure 3 The figure is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] An ISAR image classification method based on a CycleMLP model according to an embodiment of the present invention (an ISAR image classification method based on a CycleMLP model) has the following process: Figure 1 and Figure 2 As shown, the specific steps of the method are:

[0046] S1. Preprocess multiple ISAR images to obtain a minimum data set. Specifically:

[0047] The ISAR image is subjected to edge culling processing to remove the parts of the ISAR image that are irrelevant to the target and obtain a minimum data set, wherein the parts that are irrelevant to the target include the white edges and annotation information of the ISAR image; the annotation information includes the x-axis, y-axis, distance, azimuth, and pitch angle. This step can reduce the impact of irrelevant information on target evaluation.

[0048] All ISAR images in the minimum dataset are required to have obvious features and high differentiation between images. The minimum dataset contains P1 class targets, each class of targets contains no less than P2 subdivided targets, and the number of images of each subdivided target in different postures is no less than P3. The minimum dataset is divided into training set, validation set and test set in a ratio of 7:2:1.

[0049] The subdivided targets and the number of subdivided targets included in each type of target can be set according to actual conditions.

[0050] In one embodiment, there are 2400 ISAR images in the minimum data set, which are divided into 6 categories of targets according to morphological differences: parabolic antenna targets, space shuttle targets, drum targets, double-sided sailboard targets, single-sided sailboard targets and cuboid targets. The subdivided targets and the number of subdivided targets included in the parabolic antenna targets, space shuttle targets, drum targets, double-sided sailboard targets, single-sided sailboard targets and cuboid targets can be set according to actual conditions. The specific classification is shown in Table 1.

[0051] Table 1:

[0052]

[0053] S2. Expand the minimum data set to obtain an expanded data set, specifically:

[0054] The ISAR images are expanded by adding noise, non-uniform scaling, flipping and color gamut distortion.

[0055] Data set expansion is a very important means to improve the stability of the algorithm. The present invention uses four methods to achieve ISAR image data set expansion: noise expansion, non-proportional scaling expansion, flip expansion and color gamut distortion expansion. Through expansion, ISAR images can be made more diverse, the robustness of the classifier can be improved, and the influence of various additional factors on classification can be reduced. The specific implementation methods of expansion are as follows:

[0056] 1) Noise expansion: If the ISAR images in the data set are clean and perfect, the model may not be able to adapt to the noise and interference factors in some actual scenes. By expanding the data set by adding noise, the model will learn to better deal with these noise and interference factors, thereby improving its robustness and generalization ability. The present invention uses Gaussian noise to process pixel values, and sets the probability density function of the noise to obey the Gaussian distribution (ie, normal distribution), with a mean of 0 and a variance of a constant σ. Gaussian random numbers are superimposed on any pixel value, and the superposition result is scaled to [0, xs max ] range, xs max The maximum pixel value to set.

[0057] 2) Non-uniform scaling: Since the high-performance network model of the present invention does not require images of fixed size as input, non-uniform scaling can effectively introduce image data of different scale transformations to increase data diversity. Let the original ISAR image size be (w i ,h i ), the scaled size is (w n ,h n ), non-uniform scaling has two random coefficients s1 and s2. The non-uniform scaling of the image is achieved by formula (1) and formula (2), which are as follows:

[0058]

[0059] 3) Flip expansion: In horizontal flipping, the pixel at position (x0, y0) in the original ISAR image has coordinates (w-1-x0, y0) after flipping. The pixel at position (x0, y0) in the original ISAR image has coordinates (x0, h-y0-1) after vertical flipping, which is implemented by formula (3) and formula (4):

[0060]

[0061] 4) Color gamut distortion and expansion: The original ISAR image is usually RGB three-channel data. Image color gamut distortion does not change the color range of the image in the RGB domain, but converts the RGB domain to the HSV domain, distorts the hue H, saturation and lightness V in the HSV domain, and converts the distorted HSV domain data back to the RGB domain to obtain color gamut distorted image data.

[0062] S3. Construct a CycleMLP model including a block extraction layer, a CycleMLP group and a classifier. The block extraction layer is used to process each ISAR image in the expanded data set to obtain multiple final image blocks; the CycleMLP group is used to perform multi-dimensional feature fusion processing on the final image blocks, and the classifier is used to receive and process the fused features and output the classification results, as follows:

[0063] 1) Each ISAR image in the expanded data set is processed to obtain multiple final image blocks. The specific implementation method is as follows:

[0064] Each ISAR image in the expanded dataset is stacked and divided according to the partition window to obtain multiple initial image blocks. The channel features of each initial image block are dimensionally upgraded through a linear coding layer, and then each initial image block after dimension upgrade is downsampled four times using strided convolution to obtain multiple final image blocks.

[0065] Among them, the size of each ISAR image in the expanded dataset is (w, h, 3), the window size of the partition window is 7, the step size is 4, the size of each final image block obtained is (w / 4, h / 4, C1), and the number of final image blocks is recorded as S.

[0066] 2) The CycleMLP group includes N concatenated CycleMLP blocks, each of which includes: a CycleFC layer and a ChannelMLP layer, specifically:

[0067] ①CycleFC layer: The CycleMLP group mainly uses the CycleFC layer to extract the spatial features of the image block and strengthen the extraction of the contextual features of the image block. The working principle of the CycleFC layer is shown in formula (5), where U represents the output of the CycleFC layer, CycleFC represents the data processed by the Cycle fully connected layer, LayerNorm represents layer normalization, X represents the image block output by the block extraction layer as the input of the CycleFC layer, LayerNorm(X) represents layer normalization of X, and CycleFC(LayerNorm(X)) represents the CycleFC layer processing of X after layer normalization:

[0068] U=CycleFC(LayerNorm(X))+X,i=1,2,…,C2 (5)

[0069] In the specific implementation, the CycleFC layer is composed of three parallel CycleFC sz Processing operator implementation, its operator step size (S H,S W ) are 1×7, 7×1 and 1×1 respectively. The operator calculation formula is:

[0070]

[0071] δ i (c) = (cmodS H )-1

[0072]

[0073] Among them, V is the normalized image block, CycleFC sz (V) indicates CycleFC sz The operator processes the image block V, C in represents the number of channel features, Indicates the c-th channel feature, spatial position (i+δ i (c),j+δ j (c), the value at c), and b are the weight parameter and bias parameter of the CyCleFC network layer, respectively. i (c) and δ j (c) are the spatial moving intervals of the image block in the w direction and the h direction respectively. CycleFC can expand the overall receptive field of the model without increasing the computational complexity. H ,S W ) is the maximum stride size introduced in the h and w directions to expand the receptive field.

[0074] ②ChannelMLP layer: The feature quantity U output by the CycleFC layer is mixed with information in the channel domain through the MLP structure to obtain the feature quantity Y. The MLP structure is composed of two fully connected layers and a GELU activation function stacked sequentially between the two fully connected layers.

[0075] The working principle of the ChannelMLP layer is shown in formula (7), where Y represents the output of the ChannelMLP layer, W1 and W2 correspond to the weights of the two fully connected layers in the MLP, σ represents the GELU activation function, LayerNorm represents layer normalization, U represents the output of the CycleFC layer, and LayerNorm(U) represents the normalization of the feature quantity U. The ChannelMLP operation is to fuse different channels of the image block and ensure that the dimension of the image block finally output by the ChannelMLP layer is consistent with the dimension of the image block at the time of input. Formula (6) is as follows:

[0076] Y=W1σ(W2LayerNorm(U))+U (6)

[0077] 3) The classifier design is composed of the traditional global average pooling layer, fully connected layer, and Softmax layer. The standard classification head is used in conjunction with the global average pooling layer, and then a linear classifier is used to implement classification. The global average pooling layer calculates an average value for all pixels in the feature map of each channel of the output, and transforms the feature dimension through the fully connected layer. Finally, softmax is used to classify the target and obtain the category to which the target belongs.

[0078] S4. Train the CycleMLP model based on the expanded data set to obtain a trained CycleMLP model.

[0079] Among them, each ISAR image in the expanded dataset is annotated using Table 1. In the process of training the CycleMLP model, the Adam optimizer is used, and the initial learning rate can be set to 10 -3 , the attenuation factor is 6×10 -6 , the training batch is 1024 times. In order to verify the accuracy of the model, the trained CycleMLP model, CNN, and ResNet are selected to conduct experiments on the ISAR classification data set (which can be a reset data set or the expanded data set mentioned above, etc., which can be set according to actual conditions). The experimental results are shown in Table 2. According to Table 2, the trained Cycle MLP model of the present invention shows the highest accuracy in both the training set and the test set.

[0080] Table 2:

[0081]

[0082] S5. Classify the target ISAR image using the trained CycleMLP model. The target ISAR image may be an ISAR image specified by a user.

[0083] In the above embodiments, although the steps are numbered S1, S2, etc., these are only specific embodiments given by the present invention. Those skilled in the art may adjust the execution order of S1, S2, etc. according to actual conditions, which is also within the protection scope of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0084] The present invention also provides an ISAR image classification system based on the CycleMLP model (an ISAR image classification system based on the CycleMLP model), and the specific technical scheme is as follows:

[0085] It includes a data set construction module, a data set expansion module, a model construction module, a training module and a classification module;

[0086] The dataset construction module is used to: pre-process multiple ISAR images to obtain the minimum dataset;

[0087] The data set expansion module is used to: expand the minimum data set to obtain an expanded data set;

[0088] The model building module is used to build a CycleMLP model including a block extraction layer, a CycleMLP group and a classifier. The block extraction layer is used to process each ISAR image in the expanded data set to obtain multiple final image blocks. The CycleMLP group is used to perform multi-dimensional feature fusion processing on the final image blocks. The classifier is used to receive and process the fused features and output the classification results.

[0089] The training module is used to: train the CycleMLP model based on the expanded data set to obtain a trained CycleMLP model;

[0090] Classification module: Use the trained CycleMLP model to classify the target ISAR image.

[0091] Optionally, in the above technical solution, the data set construction module is specifically used to:

[0092] The ISAR image is processed by edge removal to remove the parts irrelevant to the target in the ISAR image, and the minimum data set is obtained. The minimum data set contains P1 type of targets, each type of target contains no less than P2 subdivided targets, and each subdivided target has no less than P3 images in different postures.

[0093] Optionally, in the above technical solution, the target-irrelevant part includes a white edge and annotation information of the ISAR image;

[0094] The annotation information includes the x-axis, y-axis, distance, azimuth, and elevation.

[0095] Optionally, in the above technical solution, expanding the minimum data set includes:

[0096] The ISAR images are expanded by adding noise, non-uniform scaling, flipping and color gamut distortion.

[0097] Optionally, in the above technical solution, each ISAR image in the expanded data set is processed to obtain a plurality of final image blocks, including:

[0098] Each ISAR image in the expanded dataset is stacked and divided according to the partition window to obtain multiple initial image blocks. The channel features of each initial image block are dimensionally upgraded through a linear coding layer, and then each initial image block after dimension upgrade is downsampled four times using strided convolution to obtain multiple final image blocks.

[0099] Optionally, in the above technical solution, the CycleMLP group includes N spliced ​​CycleMLP blocks, and each CycleMLP block includes: a CycleFC layer and a ChannelMLP layer.

[0100] It should be noted that the beneficial effects of the ISAR image classification system based on the CycleMLP model provided in the above embodiment are the same as the beneficial effects of the ISAR image classification method based on the CycleMLP model, which will not be described in detail here. In addition, when the system provided in the above embodiment realizes its functions, only the division of the above functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to actual conditions to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be described in detail here.

[0101] Among them, the CycleMLP model-based ISAR image classification system of the present invention can be a computer program (including program code) running in a computer device. For example, the CycleMLP model-based ISAR image classification system of the present invention is an application software that can be used to execute the corresponding steps in the CycleMLP model-based ISAR image classification method of the present invention.

[0102] In some embodiments, the CycleMLP model-based ISAR image classification system of the present invention can be implemented in a combination of software and hardware. As an example, the CycleMLP model-based ISAR image classification system of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the CycleMLP model-based ISAR image classification method of the present invention. For example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs) or other electronic components.

[0103] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The name of a module does not limit the module itself in some cases.

[0104] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any one of the above-mentioned ISAR image classification methods based on the CycleMLP model is implemented. That is to say, an electronic device according to an embodiment of the present invention may include but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the ISAR image classification method based on the CycleMLP model shown in any embodiment of the present invention by calling the computer program.

[0105] In an alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, such as through a bus 4002. Optionally, the electronic device 4000 may also include a transceiver 4004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0106] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of the present invention. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0107] The bus 4002 may include a path to transmit information between the above components. The bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 In the figure, only one thick line is used to represent the bus 4002, but this does not mean that there is only one bus or one type of bus.

[0108] The memory 4003 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to this.

[0109] The memory 4003 is used to store the application code (computer program) for executing the solution of the present invention, and the execution is controlled by the processor 4001. The processor 4001 is used to execute the application code stored in the memory 4003 to implement the content shown in the above method embodiment.

[0110] Among them, the electronic device can also be a terminal device, and the terminal device can be any terminal device that can install applications and access web pages through applications, including at least one of a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a smart TV, and a smart car device.

[0111] It should be noted that Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0112] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, and when the computer program is executed by a processor, any one of the above-mentioned ISAR image classification methods based on the CycleMLP model is implemented.

[0113] Optionally, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0114] In an exemplary embodiment, a computer program product or a computer program is also provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device performs any of the above-mentioned ISAR image classification methods based on the CycleMLP model.

[0115] Computer program code for performing the operations of the present invention may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0116] It should be understood that the flow charts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the module, the program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0117] The computer-readable storage medium provided by the embodiment of the present invention may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device, or component.

[0118] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0119] The above description is only a preferred embodiment of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the disclosure scope involved in the present invention is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present invention (but not limited to) by each other.

[0120] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects and represent the definition of a specific order or sequence. The order of use of similar objects can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than the order shown or described.

[0121] Those skilled in the art know that the present invention can be implemented as a system, method or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: it can be complete hardware, it can be complete software (including firmware, resident software, microcode, etc.), or it can be a combination of hardware and software, which is generally referred to as "circuit", "module" or "system" herein. In addition, in some embodiments, the present invention can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium contains computer-readable program code.

[0122] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A CycleMLP model based ISAR image classification method, characterized in that: The steps include: Preprocess multiple ISAR images to obtain the minimum data set; Expanding the minimum data set to obtain an expanded data set; A CycleMLP model including a block extraction layer, a CycleMLP group and a classifier is constructed, wherein the block extraction layer is used to process each ISAR image in the expanded data set to obtain a plurality of final image blocks; the CycleMLP group is used to perform multi-dimensional feature fusion processing on the final image blocks; and the classifier is used to receive and process the fused features and output classification results; Training the CycleMLP model based on the expanded data set to obtain a trained CycleMLP model; The trained CycleMLP model is used to classify the target ISAR image.

2. The ISAR image classification method based on the CycleMLP model according to claim 1, characterized in that: Preprocess the ISAR image to obtain the minimum data set, including: The ISAR image is processed by edge removal to remove the parts irrelevant to the target in the ISAR image, and the minimum data set is obtained. The minimum data set contains P1 type of targets, each type of target contains no less than P2 subdivided targets, and each subdivided target has no less than P3 images in different postures.

3. The ISAR image classification method based on the CycleMLP model according to claim 2, characterized in that: The target-irrelevant parts include white edges and annotation information of the ISAR image; The annotation information includes an x-axis, a y-axis, a distance, an azimuth angle, and a pitch angle.

4. The ISAR image classification method based on the CycleMLP model according to claim 3, characterized in that: Expanding the minimum data set includes: The ISAR images are expanded by adding noise, non-uniform scaling, flipping and color gamut distortion.

5. The ISAR image classification method based on the CycleMLP model according to claim 4, characterized in that: Each ISAR image in the expanded dataset is processed to obtain multiple final image blocks, including: Each ISAR image in the expanded dataset is stacked and divided according to the partition window to obtain multiple initial image blocks. The channel features of each initial image block are dimensionally upgraded through a linear coding layer, and then each initial image block after dimension upgrade is downsampled four times using strided convolution to obtain multiple final image blocks.

6. The ISAR image classification method based on the CycleMLP model according to claim 1, characterized in that: The CycleMLP group includes N concatenated CycleMLP blocks, and each CycleMLP block includes: a CycleFC layer and a ChannelMLP layer.

7. An ISAR image classification system based on the CycleMLP model, characterized in that: It includes a data set construction module, a data set expansion module, a model construction module, a training module and a classification module; The data set construction module is used to: pre-process multiple ISAR images to obtain a minimum data set; The data set expansion module is used to: expand the minimum data set to obtain an expanded data set; The model building module is used to build a CycleMLP model including a block extraction layer, a CycleMLP group and a classifier, wherein the block extraction layer is used to process each ISAR image in the expanded data set to obtain a plurality of final image blocks; the CycleMLP group is used to perform multi-dimensional feature fusion processing on the final image blocks, and the classifier is used to receive and process the fused features and output classification results; The training module is used to: train the CycleMLP model based on the expanded data set to obtain a trained CycleMLP model; The classification module uses the trained CycleMLP model to classify the target ISAR image.

8. The ISAR image classification system based on the CycleMLP model according to claim 7, characterized in that: The data set construction module is specifically used for: The ISAR image is processed by edge removal to remove the parts irrelevant to the target in the ISAR image, and the minimum data set is obtained. The minimum data set contains P1 type of targets, each type of target contains no less than P2 subdivided targets, and each subdivided target has no less than P3 images in different postures.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the ISAR image classification method based on the CycleMLP model as claimed in any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for ISAR image classification based on the CycleMLP model described in any one of claims 1 to 6 is implemented.

Images