Method, device and computer equipment for identifying target types in infrared remote sensing images

Abstract

The embodiments of the present invention provide a method, device, and computer equipment for identifying target types in infrared remote sensing images. The method extracts the radiation and motion characteristics of targets in infrared remote sensing images, plots characteristic curves as experimental data sets for this technology, feeds the data into a feature extraction network, and adds an attention module to the network structure to obtain the discriminative key features of the characteristic curves for each category. The loss function is designed to improve the imbalance problem in the data. Currently, relatively good experimental results can be obtained for high-speed targets such as rockets, ballistic missiles, and hypersonic aircraft. Using characteristic curves with large amounts of data to classify targets solves the difficulty of identifying target types, which only professionals can do, and is also superior to the latest existing technologies in terms of recognition accuracy.

Description

Technical Field

[0001] The present invention relates to the field of computers, and in particular to a method and device for identifying target types in infrared remote sensing images, and computer equipment. Background Art

[0002] The processing of high-resolution remote sensing imagery has had a significant impact on military, commercial, and agricultural fields. Improving the resolution and resolution of remote sensing images can drive the development of related industries. There is a significant demand for remote sensing image segmentation and extraction in areas such as land use, environmental change monitoring, and transportation. Automatically extracting and segmenting remote sensing images, while obtaining precise edge and contour information, has always been a goal of automated remote sensing image interpretation.

[0003] Remote sensing images are characterized by high resolution, complex backgrounds, inconsistent object scales, and high dimensionality. Traditional segmentation methods are unable to accurately capture these characteristics and consume significant human and material resources in image interpretation. In recent years, deep learning has rapidly developed in the field of remote sensing imagery due to its high level of intelligence.

[0004] Although deep convolutional neural networks have powerful learning and feature representation capabilities, semantic segmentation in complex scenes still faces many difficulties and challenges. Summary of the Invention

[0005] In view of this, the infrared remote sensing image target type recognition method, device and computer equipment provided in the embodiments of the present invention use characteristic curve diagrams with large amounts of data to classify targets, solving the difficulty that only professionals can identify target types, and are also superior to the latest existing technologies in terms of recognition accuracy.

[0006] In a first aspect, an embodiment of the present invention provides a method for identifying target types in infrared remote sensing images, comprising:

[0007] Acquire infrared remote sensing images containing radiation characteristics and motion characteristics to construct a target type recognition dataset;

[0008] Pre-building a target type recognition model based on an attention mechanism, wherein the target type recognition model is used to characterize the corresponding relationship between the radiation characteristics, the motion characteristics, and the target type;

[0009] Inputting a portion of the target type recognition data set as training data into the target type recognition model for training to obtain a target type recognition model with optimal parameters;

[0010] An infrared remote sensing image to be identified is acquired and input into the target type identification model with optimal parameters to obtain a target type identification result of the infrared remote sensing image to be identified.

[0011] As an optional solution, it also includes:

[0012] Determine the target type to identify low-frequency and small-sample data in the dataset;

[0013] Increase the weight of low-frequency and small-sample data in the loss function of the target type recognition model.

[0014] As an optional solution, the target type recognition model pre-built based on the attention mechanism includes:

[0015] Construct a convolutional neural network and superimpose the attention model including spatial attention and channel attention on the last layer of the convolutional neural network;

[0016] Extract the radiation characteristics and motion characteristics of the target in the infrared remote sensing image, feed the extracted target characteristic data into the convolutional neural network, extract and learn the characteristics of the target to obtain a feature map;

[0017] The feature map weights the learned features through the attention model to obtain the key features of the target category.

[0018] As an optional solution, the feature map weights the learned features through an attention model to obtain key features of the target category, including:

[0019] The feature map first adds channel attention and then adds spatial attention, so that the target is focused on the key features.

[0020] As an optional solution, the target categories include at least two of rockets, ballistic missiles, and hypersonic aircraft.

[0021] In a second aspect, an embodiment of the present invention provides an infrared remote sensing image target type recognition device, comprising:

[0022] The acquisition module is used to acquire infrared remote sensing images containing radiation characteristics and motion characteristics and construct a target type recognition dataset;

[0023] A construction module is used to pre-construct a target type recognition model based on an attention mechanism, wherein the target type recognition model is used to characterize the corresponding relationship between the radiation characteristics, the motion characteristics and the target type;

[0024] A model training module is used to input a portion of the data in the target type recognition data set as training data into the target type recognition model to train the target type recognition model to obtain a target type recognition model with optimal parameters;

[0025] The recognition module is used to obtain the infrared remote sensing image to be recognized and input the target type recognition model with optimal parameters to obtain the target type recognition result of the infrared remote sensing image to be recognized.

[0026] As an optional solution, it also includes:

[0027] The loss function adjustment module is used to determine the low-frequency and small-sample data in the target type recognition data set and increase the weight of the low-frequency and small-sample data in the loss function of the target type recognition model.

[0028] As an optional solution, the building block is specifically used to:

[0029] Construct a convolutional neural network and superimpose the attention model including spatial attention and channel attention on the last layer of the convolutional neural network;

[0030] Extract the radiation characteristics and motion characteristics of the target in the infrared remote sensing image, feed the extracted target characteristic data into the convolutional neural network, extract and learn the characteristics of the target to obtain a feature map;

[0031] The feature map weights the learned features through the attention model to obtain the key features of the target category.

[0032] As an optional solution, the building block is specifically used to:

[0033] The feature map first adds channel attention and then adds spatial attention, so that the target is focused on the key features.

[0034] In a third aspect, an embodiment of the present invention provides a computer device, including:

[0035] at least one processor; and

[0036] a memory communicatively connected to the at least one processor; wherein,

[0037] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor to enable the at least one processor to execute the above-mentioned infrared remote sensing image target type recognition method.

[0038] In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the above-mentioned infrared remote sensing image target type recognition method.

[0039] The embodiments of the present invention provide a method, device, and computer equipment for identifying target types in infrared remote sensing images. The method extracts the radiation and motion characteristics of targets in infrared remote sensing images, plots characteristic curves as experimental data sets for this technology, feeds the data into a feature extraction network, and adds an attention module to the network structure to obtain the discriminative key features of the characteristic curves for each category. The loss function is designed to improve the imbalance problem in the data. Currently, relatively good experimental results can be obtained for high-speed targets such as rockets, ballistic missiles, and hypersonic aircraft. Using characteristic curves with large amounts of data to classify targets solves the difficulty of identifying target types, which only professionals can do, and is also superior to the latest existing technologies in terms of recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A flowchart of a method for identifying target types in infrared remote sensing images is provided in an embodiment of the present invention;

[0041] Figure 2 A schematic flow chart of a method for identifying target types in infrared remote sensing images is provided in an embodiment of the present invention;

[0042] Figure 3 A schematic diagram of an attention mechanism in a method for identifying target types in infrared remote sensing images is provided in an embodiment of the present invention;

[0043] Figure 4 A structural block diagram of an infrared remote sensing image target type recognition device is provided in an embodiment of the present invention;

[0044] Figure 5 A structural block diagram of a computer device is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions 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 embodiments described 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 making creative efforts should fall within the scope of protection of the present invention.

[0046] The terms "first," "second," "third," "fourth," and the like in the specification and claims of the present invention and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. 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 apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0047] In the electromagnetic spectrum, the wavelength range of 0.76 to 1000 microns is generally referred to as the infrared region. This range is further divided into the near infrared (0.76 to 3.0 microns), mid-infrared (3.0 to 6.0 microns), far infrared (6.0 to 15.0 microns), and ultra-far infrared (15.0 to 1000 microns). Near-infrared and mid-infrared are also collectively referred to as reflected infrared, while far infrared is referred to as thermal infrared (8 to 14 microns) or emitted infrared. Although the infrared spectrum is broad, due to atmospheric absorption, only a few infrared "windows" are actually usable. Commonly used infrared bands include the near and mid-infrared bands of 0.3 to 1.3 microns, 1.5 to 1.8 microns, 2.0 to 2.6 microns, 3.0 to 3.6 microns, and 4.2 to 5.0 microns, and the far infrared band of 7.0 to 15.0 microns. Near-infrared (NIR) is primarily used for optical photography, such as infrared or color infrared photography, and operates only during daylight hours. It is also used for multi-band photography or multi-band scanning. Far-infrared (thermal) is primarily used for nighttime infrared scanning and imaging, as it is emitted by the ground itself. Infrared remote sensing is widely used in military reconnaissance, detecting volcanoes, geothermal heat, groundwater, soil temperature, identifying geological structures, and pollution monitoring.

[0048] Combine Figure 1 As shown, an embodiment of the present invention provides a method for identifying target types in infrared remote sensing images, comprising:

[0049] S101. Acquire infrared remote sensing images containing radiation characteristics and motion characteristics, and construct a target type recognition dataset.

[0050] Infrared remote sensing refers to remote sensing in which sensors operate within the infrared wavelength range. The detection band generally ranges from 0.76 to 1000 microns. It uses infrared remote sensors (such as infrared cameras and infrared scanners) to detect differences in the infrared characteristics reflected or radiated by distant vegetation and other ground objects in order to determine the nature, state, and dynamics of these objects. Using infrared remote sensing technology, image data of the target is acquired to construct a target type recognition dataset. This dataset is then plotted as a graph of the radiation characteristics versus motion characteristics, with all scale values ​​and annotations removed from the image.

[0051] Infrared remote sensing images are very sensitive to temperature changes. Rockets, missiles, and other targets experience significant temperature fluctuations during launch. Furthermore, the shutdown and stage separation times of the same type of rocket or missile are programmed to occur at pre-set times. This results in noticeable grayscale changes in infrared remote sensing images. Therefore, using the target's radiation characteristics for identification offers relatively high accuracy. Furthermore, different targets operate at different speeds during different processes, so combining speed and radiation characteristics for identification provides a more stable approach.

[0052] Radiation characteristic analysis: different targets radiate different amounts of heat in different time periods. For example, a rocket's flight time is relatively long, and obvious secondary engine ignition events may occur during the flight, which is quite different from the programmed separation moment of a ballistic missile. An aircraft may not experience obvious ignition during flight. These obvious heat changes will be sensitively captured by an infrared camera and reflected in the generated radiation characteristic curve, which can serve as a key basis for distinguishing the type.

[0053] S102. Pre-construct a target type recognition model based on an attention mechanism, wherein the target type recognition model is used to characterize the correspondence between the radiation characteristics, the motion characteristics, and the target type.

[0054] The attention mechanism originates from the study of human vision. In cognitive science, due to information processing bottlenecks, humans selectively focus on a portion of all information while ignoring other visible information. This mechanism is often referred to as the attention mechanism. Attention mechanisms are divided into spatial and channel domains. Not all regions of an image contribute equally to the task; only task-relevant regions are of interest. For example, in the main body of a classification task, the spatial attention model identifies the most important areas in the network for processing. Since the area of ​​interest is often only a small part of the image, the essence of spatial attention is to locate the target and perform some transformation or derive weights. For a CNN fed with a two-dimensional image as input, one dimension is the image's scale space, i.e., its length and width, and the other dimension is the channel.

[0055] The target type recognition model adopts a convolutional neural network, and the attention model including spatial attention and channel attention is superimposed on the last layer of the convolutional neural network. The radiation characteristics and motion characteristics of the target in the infrared remote sensing image are extracted, and the extracted target characteristic data is sent to the convolutional neural network. The characteristics of the target are extracted and learned to obtain a feature map. The feature map is weighted by the learned features through the attention model to obtain the key features of the target category.

[0056] S103: Input a portion of the target type recognition data set as training data into the target type recognition model for training to obtain a target type recognition model with optimal parameters.

[0057] The target type recognition data set can be divided into a part as a training set and a part as a test set. The division ratio can be selected as needed. In this embodiment, it can be divided into a training set and a test set according to a ratio of 9:1. The pre-built target type recognition model is trained using the training set until the model converges and the training is completed. At this time, the target type recognition model with the optimal parameters is obtained. The target type recognition model with the optimal parameters is then verified by the test set to determine whether the model meets the usage requirements.

[0058] S104 , obtaining an infrared remote sensing image to be identified and inputting the target type identification model with the optimal parameters to obtain a target type identification result of the infrared remote sensing image to be identified.

[0059] The trained target type recognition model with optimal parameters is used to identify the infrared remote sensing image to determine the target type of the target object in the image.

[0060] The present invention provides a method for identifying target types in infrared remote sensing images. The method extracts the radiation characteristics and motion characteristics of targets in infrared remote sensing images, plots characteristic curves as experimental data sets for this technology, feeds the data into a feature extraction network, and adds an attention module to the network structure to obtain the discriminative key features of the characteristic curves for each category. The loss function is designed to improve the imbalance problem in the data. Currently, relatively good experimental results can be obtained for high-speed targets such as rockets, ballistic missiles, and hypersonic aircraft. Using characteristic curves with large amounts of data to classify targets solves the difficulty of identifying target types, which can only be done by professionals, and is also superior to the latest existing technologies in terms of recognition accuracy.

[0061] Analyzing the target's motion characteristics, ballistic missiles typically travel at speeds exceeding Mach 10, approximately 700 meters per second. Ballistic missiles, especially long-range ones, can reach the fastest speeds at the end of their power phase, reaching the first cosmic velocity, approximately 22 to 25 times the speed of sound. A rocket's flight speed depends on the thrust of its engine and the rocket's mass ratio. The greater the engine's thrust, the faster the rocket's flight speed; the greater the rocket's mass ratio, the higher the speed it can reach. Generally speaking, for a launch into a sun-synchronous orbit at an altitude of 500 kilometers, a rocket enters orbit at a speed of 7.6 kilometers per second; for a launch into a geosynchronous transfer orbit, the rocket can reach a speed of 10.1 kilometers per second; and for a launch into a cis-lunar transfer orbit, the rocket can even reach a speed of 10.8 kilometers per second. Hypersonic vehicles are winged or wingless aircraft, such as aircraft, missiles, and artillery shells, that travel at speeds exceeding five times the speed of sound. Their speeds are similar to those of ballistic missiles, requiring subsequent analysis of their radiation characteristics.

[0062] The input data is first plotted as a curve graph of radiation and motion characteristics. All scale values ​​and annotations in the image are removed. The processed image is then fed into a feature extraction network to extract curve features. An attention mechanism is incorporated into the network structure. The attention model weights the learned features to capture key characteristics of the category. Next, a weighted loss function is designed to give low-frequency samples a higher weight in the loss function, thereby increasing their influence during classification. The output of the final, fully connected layer of the network is the target category.

[0063] Since there is a certain degree of imbalance in the target categories, when designing the loss function, the loss function is weighted so that the weight of low-frequency samples in the loss function is larger, thereby increasing the influence of low-frequency samples during classification. Specifically, the method provided in this embodiment also includes:

[0064] Determine the target type to identify low-frequency and small-sample data in the dataset;

[0065] Increase the weight of low-frequency and small-sample data in the loss function of the target type recognition model.

[0066] The loss function is used to measure the degree of inconsistency between your model's predicted value f(x) and the true value Y. It is a non-negative real-valued function, usually expressed as L(Y, f(x)). The smaller the loss function, the better the robustness of the model. The loss function is the core part of the empirical risk function and an important component of the structural risk function. The structural risk function of the model includes the empirical risk term and the regularization term, and can usually be expressed as the following formula (1):

[0067]

[0068] The mean function in the front represents the empirical risk function, L represents the loss function, and Φ is the regularizer or penalty term, which can be L1, L2, or other regularization functions. The entire formula means finding the value of θ that minimizes the objective function.

[0069] Because the target dataset is imbalanced, simple data flipping and translation during data augmentation can alter the target's characteristic curve, hindering network learning. Therefore, the loss function design must address this data imbalance. Intuitively, the problem caused by sample imbalance is that classification of categories with fewer samples is more difficult. Therefore, considering the difficulty of classifying samples, the loss function is designed to focus on difficult samples, addressing the issue of low classification accuracy for categories with fewer samples.

[0070] The following formula (2) is the loss function L fl A form of low frequency class to high frequency class: p t = 0 to 1, p t reflects the closeness to the true value category, p t A larger value indicates a closer approximation to the true category, meaning a more accurate classification. γ is an adjustable factor. Setting γ > 0 results in different loss function curves. For inaccurately classified samples, the loss remains unchanged, while for accurately classified samples, the loss decreases. Overall, this is equivalent to increasing the weight of inaccurately classified samples in the loss function.

[0071] L fl =-(1-p t ) γ log(p t ) (2).

[0072] Combine Figure 2 As shown, the input data is first plotted as a curve graph of radiation and motion characteristics. All scale values ​​and labels in the image are deleted, and the processed image is fed into a feature extraction network. The feature extraction network uses a convolutional neural network (Conv) to extract curve features. An attention mechanism is incorporated into the network structure. The attention model can weight the learned features to obtain key characteristics of the category. Next, a weighted loss function is designed to give low-frequency samples a higher weight in the loss function, thereby increasing their influence during classification. The output of the final, fully connected layer of the network is the target category.

[0073] To enhance the extraction of salient features in convolutional neural networks, attention modules are introduced. Attention mechanisms operate in both spatial and channel domains. Not all regions of an image contribute equally to the task; only those relevant to the task are of interest. For example, in the main body of a classification task, the spatial attention model identifies the most important areas in the network for processing. Since the region of interest is often only a small part of the image, the essence of spatial attention is to locate the target and perform some transformation or derive weights. For a CNN with a two-dimensional image as input, one dimension is the image's scale space, i.e., its length and width, and the other is the channel.

[0074] Combine Figure 3 As shown, in some embodiments, the feature map weights the learned features through an attention model to obtain key features of the target category, including:

[0075] The feature map first adds channel attention and then adds spatial attention, so that the target is focused on the key features.

[0076] As an optional solution, the target categories include at least two of rockets, ballistic missiles, and hypersonic vehicles. Hypersonic vehicles refer to winged or wingless aircraft, such as aircraft, missiles, and artillery shells, that fly at speeds exceeding five times the speed of sound. It should be noted that the target type can be adjusted based on the target to be identified and is not limited to this.

[0077] Combine Figure 4 As shown, an embodiment of the present invention provides an infrared remote sensing image target type recognition device 400, comprising:

[0078] An acquisition module 401 is used to acquire infrared remote sensing images including radiation characteristics and motion characteristics and construct a target type recognition data set;

[0079] A construction module 402 is configured to pre-construct a target type recognition model based on an attention mechanism, wherein the target type recognition model is configured to characterize a correspondence between the radiation characteristics, the motion characteristics, and the target type;

[0080] A model training module 403 is configured to input a portion of the target type recognition data set as training data into the target type recognition model for training to obtain a target type recognition model with optimal parameters;

[0081] The recognition module 404 is configured to obtain an infrared remote sensing image to be recognized and input the target type recognition model with the optimal parameters to obtain a target type recognition result of the infrared remote sensing image to be recognized.

[0082] An infrared remote sensing image target type recognition device provided in an embodiment of the present invention extracts the radiation characteristics and motion characteristics of the target in the infrared remote sensing image, draws a characteristic curve as the experimental data set of this technology, feeds the data into a feature extraction network, and adds an attention module to the network structure to obtain the discriminative key features of the characteristic curve of each category. The imbalance problem of the data is improved by designing a loss function. Currently, relatively good experimental results can be obtained for high-speed targets such as rockets, ballistic missiles, and hypersonic aircraft. Using large amounts of characteristic curves to classify targets solves the difficulty of identifying target types that only professionals can do, and is also superior to the latest existing technologies in terms of recognition accuracy.

[0083] As an optional solution, it also includes:

[0084] The loss function adjustment module 405 is used to determine the low-frequency and small sample data in the target type recognition data set, and increase the weight of the low-frequency and small sample data in the loss function of the target type recognition model.

[0085] As an optional solution, the construction module 402 is specifically used to:

[0086] Construct a convolutional neural network and superimpose the attention model including spatial attention and channel attention on the last layer of the convolutional neural network;

[0087] Extract the radiation characteristics and motion characteristics of the target in the infrared remote sensing image, feed the extracted target characteristic data into the convolutional neural network, extract and learn the characteristics of the target to obtain a feature map;

[0088] The feature map weights the learned features through the attention model to obtain the key features of the target category.

[0089] As an optional solution, the construction module 402 is specifically used to:

[0090] The feature map first adds channel attention and then adds spatial attention, so that the target is focused on the key features.

[0091] Accordingly, according to embodiments of the present disclosure, the present disclosure also provides a computer device, a readable storage medium, and a computer program product.

[0092] Figure 5 FIG. 1 is a structural diagram of a computer device 12 provided in an embodiment of the present invention. Figure 5 A block diagram of an exemplary computer device 12 suitable for implementing embodiments of the present disclosure is shown. Figure 5The computer device 12 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.

[0093] like Figure 5 As shown, computer device 12 is represented in the form of a general-purpose computing device. Computer device 12 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0094] Components of computer device 12 may include, but are not limited to, one or more processors or processing units 16 , a system memory 28 , and a bus 18 that connects various system components, including system memory 28 and processing unit 16 .

[0095] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. Examples of these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0096] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0097] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be configured to read and write non-removable, non-volatile magnetic media ( Figure 5 Not shown, often called a "hard drive"). Although Figure 5Not shown, a magnetic disk drive for reading and writing to a removable non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of various embodiments of the present invention.

[0098] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may include an implementation of a network environment. Program modules 42 generally implement the functions and / or methods of the embodiments described herein.

[0099] The computer device 12 can also communicate with one or more external devices 14 (e.g., a keyboard, pointing device, display 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication can occur via an input / output (I / O) interface 22. Furthermore, the computer device 12 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with the other modules of the computer device 12 via a bus 18. It should be understood that, although not shown, other hardware and / or software modules can be used in conjunction with the computer device 12, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0100] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the infrared remote sensing image target type recognition method provided by the embodiment of the present invention.

[0101] An embodiment of the present invention also provides a non-transitory computer-readable storage medium storing computer instructions, on which a computer program is stored, wherein when the program is executed by a processor, the infrared remote sensing image target type recognition method provided in all the inventive embodiments of this application is implemented.

[0102] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. More specific examples (non-exhaustive list) of computer-readable storage media include: 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 of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device.

[0103] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0104] The program code included on the computer-readable medium can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, RF or the like, or any suitable combination thereof. The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or a combination thereof, and the programming language includes object-oriented programming languages ​​such as Java, Smalltalk, C++, and also includes conventional procedural programming languages ​​such as "C" language or similar programming languages. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can include a local area network (LAN) or a wide area network (WAN) connected to the user's computer through any type of network, or can be connected to an external computer (for example, utilizing an Internet service provider to connect through the Internet).

[0105] The present disclosure also provides a computer program product, comprising a computer program, which implements the above-mentioned infrared remote sensing image target type recognition method when executed by a processor.

[0106] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0107] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A method for identifying target types in infrared remote sensing images, characterized in that: include: Acquire infrared remote sensing images containing radiation characteristics and motion characteristics to construct a target type recognition dataset; Pre-building a target type recognition model based on an attention mechanism, wherein the target type recognition model is used to characterize the corresponding relationship between the radiation characteristics, the motion characteristics, and the target type; The target type recognition model pre-built based on the attention mechanism includes: Construct a convolutional neural network and superimpose the attention model including spatial attention and channel attention on the last layer of the convolutional neural network; Extract the radiation characteristics and motion characteristics of the target in the infrared remote sensing image, feed the extracted target characteristic data into the convolutional neural network, extract and learn the characteristics of the target to obtain a feature map; The target characteristic data is a graph of radiation characteristics and motion characteristics; The feature map weights the learned features through the attention model to obtain the key features of the target category; Inputting a portion of the target type recognition data set as training data into the target type recognition model for training to obtain a target type recognition model with optimal parameters; Acquire an infrared remote sensing image to be identified and input the target type identification model with the optimal parameters to obtain a target type identification result of the infrared remote sensing image to be identified; The target categories include at least two of rockets, ballistic missiles, and hypersonic aircraft.

2. The infrared remote sensing image target type recognition method according to claim 1, characterized in that: Also includes: Determine the target type to identify low-frequency and small-sample data in the dataset; Increase the weight of low-frequency and small-sample data in the loss function of the target type recognition model.

3. The infrared remote sensing image target type recognition method according to claim 1, characterized in that: The feature map weights the learned features through the attention model to obtain the key features of the target category, including: The feature map first adds channel attention and then adds spatial attention, so that the target is focused on the key features.

4. An infrared remote sensing image target type recognition device, characterized in that: include: The acquisition module is used to acquire infrared remote sensing images containing radiation characteristics and motion characteristics and construct a target type recognition dataset; A construction module is used to pre-construct a target type recognition model based on an attention mechanism, wherein the target type recognition model is used to characterize the corresponding relationship between the radiation characteristics, the motion characteristics and the target type; The building blocks are specifically used for: Construct a convolutional neural network and superimpose the attention model including spatial attention and channel attention on the last layer of the convolutional neural network; Extract the radiation characteristics and motion characteristics of the target in the infrared remote sensing image, feed the extracted target characteristic data into the convolutional neural network, extract and learn the characteristics of the target to obtain a feature map; The target characteristic data is a graph of radiation characteristics and motion characteristics; The feature map weights the learned features through the attention model to obtain the key features of the target category; A model training module is used to input a portion of the data in the target type recognition data set as training data into the target type recognition model to train the target type recognition model to obtain a target type recognition model with optimal parameters; an identification module, configured to obtain an infrared remote sensing image to be identified and input the target type identification model with the optimal parameters to obtain a target type identification result of the infrared remote sensing image to be identified; The target categories include at least two of rockets, ballistic missiles, and hypersonic aircraft.

5. The infrared remote sensing image target type recognition device according to claim 4, characterized in that: Also includes: The loss function adjustment module is used to determine the low-frequency and small-sample data in the target type recognition data set, and increase the weight of the low-frequency and small-sample data in the loss function of the target type recognition model.

6. The infrared remote sensing image target type recognition device according to claim 5, characterized in that: The building blocks are specifically used for: The feature map first adds channel attention and then adds spatial attention, so that the target is focused on the key features.

7. A computer device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the infrared remote sensing image target type recognition method according to any one of claims 1 to 5.

Images