Deception jamming method and system for infrared image target recognition

Abstract

The invention provides a deception jamming method and system for infrared image target recognition, electronic equipment and a storage medium. The method comprises the following steps: representing local invariance features of an infrared image; tiny disturbance is added to the represented infrared image based on a differential evolution algorithm, the disturbance cannot be recognized by a model by means of local invariant features of the image, and the model is made to make misclassification. According to the scheme provided by the invention, a certain theoretical basis can be provided for an anti-intelligent identification system, and a set of deception jamming scheme for intelligent identification of the anti-infrared seeker is established.

Description

Technical Field

[0001] The invention belongs to the technical field of infrared modeling, and in particular relates to a deception interference method, system, electronic equipment and storage medium for infrared image target recognition. Background Art

[0002] At present, intelligent technologies represented by deep learning are in a stage of rapid development, and their wide application in target recognition has made many progresses. In the infrared field, intelligent target recognition technology can process images of different resolutions, use computers, artificial intelligence, image processing technology and other means to identify targets in infrared images and output recognition results, which can be used for target detection and recognition in complex environments.

[0003] In the military field, infrared image target recognition technology can be used to identify enemy equipment and hidden targets, improving military security. In the industrial field, this technology can be used to detect product defects and reduce the missed detection rate of manual inspection. With the development of artificial intelligence related technologies, infrared image target recognition can achieve high-precision target recognition by deeply analyzing and extracting target shape, edge, texture and other features in infrared images.

[0004] However, with the widespread application of infrared image recognition technology, how to effectively counter possible intelligent recognition technology and achieve anti-intelligent deception interference has become an equally important research topic. Anti-intelligent recognition deception interference algorithm aims to interfere with and mislead the opponent's intelligent recognition system through various means, making it unable to accurately identify the target in the infrared image.

[0005] Therefore, how to provide a deception interference method, system, electronic device and storage medium for infrared image target recognition has become a technical problem that urgently needs to be solved in this field. Summary of the invention

[0006] The purpose of the present invention is to provide a deception interference method, system, electronic equipment and storage medium for infrared image target recognition.

[0007] According to a first aspect of the present invention, a deception jamming method for infrared image target recognition is provided, comprising:

[0008] Step S1, characterizing the local invariance features of the infrared image;

[0009] Step S2: adding a tiny disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image, so that the model cannot recognize the disturbance and makes the model make an incorrect classification.

[0010] According to the method of the first aspect of the present invention, in step S1, characterizing the local invariant features of the infrared image includes:

[0011] Preprocess the infrared image;

[0012] According to the invariant feature detection function under the scale space theory, the local feature invariance of the preprocessed infrared image is described to obtain the feature points in the scale space;

[0013] The feature points in the scale space are processed by corner, edge, spot and region detection algorithms to obtain local feature points;

[0014] Applying SIFT algorithm to extract the local feature points and generate descriptors;

[0015] The descriptors are matched by the FLANN approximate nearest neighbor search algorithm.

[0016] According to the method of the first aspect of the present invention, in step S1, characterizing the local invariance features of the infrared image further comprises:

[0017] The high-dimensional features after feature matching are reduced in dimension through PCA to obtain a low-dimensional feature space, thus completing the construction of local feature spaces of infrared images at different scales.

[0018] According to the method of the first aspect of the present invention, in step S2, adding a small disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image, so that the model cannot recognize the disturbance and causes the model to make an incorrect classification includes:

[0019] Based on the differential evolution algorithm, find the pixel modification direction of the perturbation of the represented image;

[0020] According to the pixel modification direction, designing a feature similarity disturbance direction of the image;

[0021] According to the feature-similar perturbation direction, a predefined number of pixels are modified to achieve an image deception effect.

[0022] The second aspect of the present invention discloses a deception jamming system for infrared image target recognition; the system comprises:

[0023] A first processing module is configured to characterize local invariant features of the infrared image;

[0024] The second processing module is configured to add a small disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image so that the model cannot recognize the disturbance and makes the model make an incorrect classification.

[0025] According to the system of the second aspect of the present invention, the first processing module is specifically configured as follows:

[0026] Preprocess the infrared image;

[0027] According to the invariant feature detection function under the scale space theory, the local feature invariance of the preprocessed infrared image is described to obtain the feature points in the scale space;

[0028] The feature points in the scale space are processed by corner, edge, spot and region detection algorithms to obtain local feature points;

[0029] Applying SIFT algorithm to extract the local feature points and generate descriptors;

[0030] The descriptors are matched by the FLANN approximate nearest neighbor search algorithm.

[0031] According to the system of the second aspect of the present invention, the first processing module is specifically configured as follows:

[0032] The characterization of the local invariance features of the infrared image also includes:

[0033] The high-dimensional features after feature matching are reduced in dimension through PCA to obtain a low-dimensional feature space, thus completing the construction of local feature spaces of infrared images at different scales.

[0034] According to the system of the second aspect of the present invention, the second processing module is specifically configured as follows:

[0035] The method of adding a small disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image, makes the model unable to recognize the disturbance and causes the model to make an incorrect classification, including:

[0036] Based on the differential evolution algorithm, find the pixel modification direction of the perturbation of the represented image;

[0037] According to the pixel modification direction, designing a feature similarity disturbance direction of the image;

[0038] According to the feature-similar perturbation direction, a predefined number of pixels are modified to achieve an image deception effect.

[0039] The third aspect of the present invention discloses an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any one of the infrared image target recognition deception jamming methods in the first aspect of the present disclosure are implemented.

[0040] The fourth aspect of the present invention discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any one of the infrared image target recognition deception jamming methods in the first aspect of the present disclosure are implemented.

[0041] The beneficial effects brought by the present invention are as follows:

[0042] It can be seen from the above scheme that the embodiments of the present invention provide a deception interference method, system, electronic device and storage medium for infrared image target recognition, which has the following beneficial effects: it can provide a certain theoretical basis for the anti-intelligent recognition system and establish a set of deception interference schemes for anti-infrared seeker intelligent recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flowchart of a deception jamming method for infrared image target recognition provided according to an embodiment;

[0044] Figure 2 A structural diagram of a deception jamming system for infrared image target recognition according to an embodiment of the present invention;

[0045] Figure 3 The figure is a structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all 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.

[0047] Embodiment 1:

[0048] According to a first aspect of the present invention, the present invention discloses a deception interference method for infrared image target recognition. Figure 1 FIG. 1 is a flow chart of a deception jamming method for infrared image target recognition according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0049] Step S1, characterizing the local invariance features of the infrared image;

[0050] Step S2: adding a tiny disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image, so that the model cannot recognize the disturbance and makes the model make an incorrect classification.

[0051] In step S1, the local invariant features of the infrared image are characterized.

[0052] In some embodiments, in step S1, characterizing the local invariance features of the infrared image includes:

[0053] Preprocess the infrared image;

[0054] According to the invariant feature detection function under the scale space theory, the local feature invariance of the preprocessed infrared image is described to obtain the feature points in the scale space;

[0055] The feature points in the scale space are processed by corner, edge, spot and region detection algorithms to obtain local feature points;

[0056] Applying SIFT algorithm to extract the local feature points and generate descriptors;

[0057] The descriptors are matched by the FLANN approximate nearest neighbor search algorithm.

[0058] The characterization of the local invariance features of the infrared image also includes:

[0059] The high-dimensional features after feature matching are reduced in dimension through PCA to obtain a low-dimensional feature space, thus completing the construction of local feature spaces of infrared images at different scales.

[0060] In step S2, a small disturbance is added to the characterized infrared image based on the differential evolution algorithm, and the local invariant features of the image are relied upon to make the model unable to recognize the disturbance and cause the model to make an incorrect classification.

[0061] In some embodiments, in step S2, adding a small disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image so that the model cannot recognize the disturbance and causes the model to make an incorrect classification includes:

[0062] Based on the differential evolution algorithm, find the pixel modification direction of the perturbation of the represented image;

[0063] According to the pixel modification direction, designing a feature similarity disturbance direction of the image;

[0064] According to the feature-similar perturbation direction, a predefined number of pixels are modified to achieve an image deception effect. Specific embodiments

[0066] In order to meet the needs of intelligent infrared stealth for targets in the sea battlefield, this patent conducts research on adversarial anti-intelligent recognition technology for the recognition algorithm of typical infrared seekers. Based on the local invariant feature description of the image and the deep neural network learning algorithm, the optimization goal is to reduce the recognition probability of the target. By changing the image pixels, texture or target contour and other features, the survival probability of the ship target under the anti-ship missile strike operation is improved. It is mainly divided into the following two steps.

[0067] (1) Characterization of local invariance features of infrared images

[0068] Focusing on infrared images of targets in naval battlefield environments, we analyze typical features and their changing patterns in infrared images, study the invariant feature detection function under scale space theory, describe the invariance of local features of infrared images, and study the detection effects of mainstream corner, edge, spot and region detection algorithms on local invariant features of infrared images. We design infrared image local feature descriptors to quantitatively describe the local structural features of images, study the shape and texture structure characteristics of local images near feature points, form feature descriptors with high robustness, uniqueness and high matching speed for infrared images, and complete the construction of local feature spaces of infrared images at different scales.

[0069] The basic steps are as follows:

[0070] 1) Infrared image preprocessing

[0071] In order to improve the effect of feature extraction, the infrared image needs to be preprocessed first, including denoising, contrast enhancement and standardization.

[0072] Denoising: For infrared images, a Gaussian filter is used to smooth the image and remove noise. The Gaussian filter is defined as:

[0073]

[0074] Among them, σ is the standard deviation, which is used to control the degree of smoothing, and x, y are pixel coordinates.

[0075] 2) Feature extraction based on scale space theory

[0076] Scale-Space Theory allows the analysis of images at different scales and can effectively extract scale-invariant features.

[0077] The first step is to construct the scale space by using Gaussian convolution filter. The construction of scale space is achieved by continuous Gaussian blur operation, which can be expressed as

[0078] L(x,y,σ)=G(x,y,σ)*I(x,y),

[0079] Among them, L(x,y,σ) is the image at scale σ, I(x,y) is the original image, and G(x,y,σ) is the Gaussian kernel.

[0080] Secondly, calculate the difference between Gaussian images at different scales (DoG) to obtain the feature points in the scale space

[0081] D(x,y,σ)=L(x,y,kσ)-L(x,y,kσ),

[0082] Among them, k is the scale factor and σ is the current scale.

[0083] 3) Corner, edge, spot and region detection algorithms

[0084] According to the characteristics of the target infrared image, select an appropriate local feature detection algorithm, such as Harris corner detection or Canny edge detection algorithm.

[0085] Harris corner detection algorithm identifies corners by calculating the autocorrelation matrix of the local area of ​​the image. The core of the algorithm is to calculate the image gradient matrix M, which is calculated by

[0086]

[0087] Among them, I x ,I y are the gradients of the image in the x and y directions, respectively.

[0088] The corner point response function R is

[0089]

[0090] Where det(M) is the determinant of the matrix, trace(M) is the trace of the matrix, and k is an empirical constant, usually between 0.04 and 0.06.

[0091] The Canny edge detection algorithm calculates the image gradient and performs non-maximum suppression and double threshold processing to finally detect the edge of the image. The gradient of edge detection is

[0092]

[0093] Among them, I x ,I y are the gradients of the image in the x and y directions, respectively.

[0094] 4) Local feature descriptor

[0095] The local features of the image are quantitatively described by designing descriptors, such as the SIFT descriptor.

[0096] The SIFT algorithm extracts local feature points and generates descriptors. The descriptor of each feature point is constructed by calculating the gradient direction and magnitude of the region. The gradient direction is calculated as

[0097]

[0098] Among them, I x ,I y are the gradients of the image in the x and y directions, respectively.

[0099] The descriptor is usually a histogram consisting of the gradient direction and amplitude of the local area, and the descriptor of the feature point is usually calculated within a 16×16 window.

[0100] 5) Feature matching and matching speed optimization

[0101] Feature matching is the key to target recognition. In order to improve matching speed and accuracy, accelerated feature descriptors (such as ORB) and optimized matching methods are used.

[0102] FLANN (Fast Library for Approximate Nearest Neighbors) is an efficient approximate nearest neighbor search algorithm that can accelerate the feature matching process. The matching steps are as follows:

[0103] Match(d1,d2)=argmin||d1-d2||,

[0104] Among them, d1 and d2 are vectors of two feature descriptors respectively. represents the Euclidean distance.

[0105] In order to improve the matching accuracy, the ratio test method is used to filter out false matches:

[0106]

[0107] Among them, d1, d2 are two matching distances, and τ is the ratio threshold, which is usually 0.7.

[0108] 6) Construction of local feature space

[0109] In order to solve the problem of feature matching at different scales, a local feature space can be constructed to optimize the matching efficiency through feature clustering or dimensionality reduction.

[0110] PCA is used to reduce the dimension of high-dimensional features to obtain a low-dimensional feature space, thereby improving matching efficiency. The calculation formula of PCA is:

[0111] Z=XW,

[0112] Among them, X is the feature matrix, W is the feature vector matrix, and Z is the feature matrix after dimensionality reduction.

[0113] (2) Implementing deception jamming strategy based on differential evolution algorithm

[0114] In view of the sensitivity of deep neural networks to tiny perturbations of input data, image pixels are perturbed based on the differential evolution algorithm. According to the local invariant features of infrared images, a pixel attack strategy with similar perturbation directions of image features is designed. The deception and stealth confrontation of image recognition results are achieved through a few-pixel attack method that can hide the modification effect.

[0115] By constructing image invariant features of typical infrared scenes, we study the dependence of mainstream target detection and recognition algorithms on image invariant features, and deceive typical classification networks by adding small and imperceptible pixel interference to the image, inducing typical target detection and classification algorithms to misclassify targets, and accumulating effective adversarial samples to counter intelligent learning attacks.

[0116] In deep neural networks, tiny pixel perturbations can have a huge impact on the output of the model. The goal of this patent is to design a pixel perturbation strategy that keeps the visual effect of the image unchanged, but can induce the classification network or target detection model to produce incorrect classification or recognition results. Specifically, adding tiny perturbations to infrared images, relying on the local invariant features of the image (such as corners, edges, etc.), makes the model unable to recognize these perturbations, and causes the target detection algorithm to make incorrect classifications.

[0117] 1) Differential Evolution Algorithm

[0118] The differential evolution algorithm is a population-based random optimization algorithm that is suitable for solving global optimization problems in continuous space and can be used to find the most effective pixel modification direction for image perturbations. Its basic principle is to update the solution by mutation, crossover and selection in the population.

[0119] The basic steps are as follows:

[0120] i) Initialize the population. Generate the initial population X0 = {x1, x2, ..., x N}, where x i is the i-th solution;

[0121] ii) Mutation operation. Generate mutation vector v through differential mutation operation i =x r 1+F·(x r 2-x r 3), where x r 1,x r 2,x r 3 are different individuals randomly selected, and F is the scaling factor to control the amplitude of variation;

[0122] iii) Crossover operation. Cross the mutation vector and the target solution

[0123]

[0124] Among them, rand (0,1) is a random number, C is the crossover probability, j rand is a randomly selected dimension.

[0125] iv) Select operation. Select the optimal solution based on the fitness function:

[0126]

[0127] When constructing the objective function, the impact of the perturbed image on the neural network output can be designed as a trade-off between minimizing the image perturbation and maximizing the model classification error, that is,

[0128]

[0129] Among them, x is the perturbed image, x o is the original image, is the size of the disturbance, L adv (x) is the adversarial loss, which indicates the degree of misclassification of the model’s output, and α and β are trade-off parameters.

[0130] 3) Design of perturbations similar to local invariant features and image features

[0131] The perturbation direction is designed to be similar to the image features so that the perturbation can keep the local invariant features of the image (such as corners, edges, etc.) as much as possible, while preventing the target detection and recognition algorithms from misclassifying.

[0132] The invariant features (such as SIFT, etc.) are extracted through the scale space theory of the image, and the descriptors of the feature points are constructed. For each feature point, the perturbation objective function can be defined as

[0133]

[0134] Among them, d(x i ,x′ i ) is the difference metric of the descriptors, and ||·||2 is the Euclidean distance.

[0135] The direction of perturbation needs to be selected based on the invariance of local features of the image. For each feature point, the perturbation strategy is designed by minimizing the impact of perturbation on the feature (maintaining similarity) and maximizing the classification error. Specifically, the goal of perturbation is to make:

[0136]

[0137] In this way, the perturbation affects the classification result without changing the local structure of the image.

[0138] 4) Few-pixel attack method

[0139] Minimal Pixel Attack aims to deceive images by modifying a small number of pixels while keeping the modification imperceptible. The specific strategy is as follows:

[0140] i) Pixel selection. Select the pixels that have the greatest impact on the target classification network for modification. Determine the modification location by analyzing the gradient or using important areas of the image (such as the edge of the target)

[0141]

[0142] in, is the gradient of the loss function, indicating the sensitivity to image perturbations.

[0143] ii) Apply perturbations. Perturbations are performed by calculating the pixel locations that are most sensitive to the target network. Only a few pixels in the image are modified each time, the impact of the perturbation is calculated, and the number of modifications is gradually increased until the maximum deception effect is achieved.

[0144] Through the above strategy, a set of adversarial samples are generated and used to train or evaluate the target detection and classification algorithms. The accumulation of adversarial samples can help improve the effectiveness of adversarial attacks and verify the robustness of attacks in different environments. Through multiple rounds of iterative optimization of perturbations, the number of adversarial samples is gradually increased, and the impact of these samples on the classification network is verified. After each round of attack, the perturbation strategy is adjusted according to the classification results to make the adversarial samples more effective.

[0145] In summary, the solution proposed in the present invention can provide a certain theoretical basis for the anti-intelligent recognition system and establish a deception interference solution for the intelligent recognition of anti-infrared seeker.

[0146] Embodiment 2:

[0147] The invention discloses a deception jamming system for infrared image target recognition. Figure 2 FIG. 4 is a structural diagram of a deception jamming system for infrared image target recognition according to an embodiment of the present invention; Figure 2 As shown, the system 100 includes:

[0148] The first processing module 101 is configured to characterize the local invariance features of the infrared image;

[0149] The second processing module 102 is configured to add a small disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image so that the model cannot recognize the disturbance and makes the model make an incorrect classification.

[0150] According to the system of the second aspect of the present invention, the first processing module 101 is specifically configured as follows:

[0151] Preprocess the infrared image;

[0152] According to the invariant feature detection function under the scale space theory, the local feature invariance of the preprocessed infrared image is described to obtain the feature points in the scale space;

[0153] The feature points in the scale space are processed by corner, edge, spot and region detection algorithms to obtain local feature points;

[0154] Applying SIFT algorithm to extract the local feature points and generate descriptors;

[0155] The descriptors are matched by the FLANN approximate nearest neighbor search algorithm.

[0156] The characterization of the local invariance features of the infrared image also includes:

[0157] The high-dimensional features after feature matching are reduced in dimension through PCA to obtain a low-dimensional feature space, thus completing the construction of local feature spaces of infrared images at different scales.

[0158] According to the system of the second aspect of the present invention, the second processing module 102 is specifically configured as follows:

[0159] The method of adding a small disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image, makes the model unable to recognize the disturbance and causes the model to make an incorrect classification, including:

[0160] Based on the differential evolution algorithm, find the pixel modification direction of the perturbation of the represented image;

[0161] According to the pixel modification direction, designing a feature similarity disturbance direction of the image;

[0162] According to the feature-similar perturbation direction, a predefined number of pixels are modified to achieve an image deception effect.

[0163] Embodiment 3:

[0164] The present application discloses an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the infrared image target recognition deception jamming method in any one of the embodiments 1 disclosed in the present invention are implemented.

[0165] Figure 3 is a structural diagram of an electronic device according to an embodiment of the present invention, such as Figure 3As shown, the electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the electronic device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, near field communication (NFC) or other technologies. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic device can be a touch layer covered on the display screen, or a button, a trackball or a touch pad set on the housing of the electronic device, or an external keyboard, touch pad or mouse, etc.

[0166] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a structural diagram of the part related to the technical solution of the present disclosure, and does not constitute a limitation on the electronic device to which the technical solution of the present application is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0167] Embodiment 4:

[0168] The present invention discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the deception interference method for infrared image target recognition in any one of the embodiments of the present invention are implemented.

[0169] Please note that the technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. The above embodiments only express several implementation methods of the present application, and their descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the present application, several variations and improvements can be made, which all belong to the scope of protection of the present application. Therefore, the scope of protection of the patent in this application shall be based on the attached claims.

[0170] The embodiments of the subject matter and functional operations described in this specification may be implemented in the following: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or a combination of one or more of them. The embodiments of the subject matter described in this specification may be implemented as one or more computer programs, i.e., one or more modules in computer program instructions encoded on a tangible non-temporary program carrier to be executed by a data processing device or to control the operation of the data processing device. Alternatively or additionally, the program instructions may be encoded on an artificially generated propagation signal, such as a machine-generated electrical, optical or electromagnetic signal, which is generated to encode information and transmit it to a suitable receiver device for execution by a data processing device. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them.

[0171] The processes and logic flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform corresponding functions by operating on input data and generating output. The processes and logic flows can also be performed by special purpose logic circuits, such as FPGAs (field programmable gate arrays) or ASICs (application-specific integrated circuits), and the apparatus can also be implemented as special purpose logic circuits.

[0172] Computers suitable for executing computer programs include, for example, general and / or special microprocessors, or any other type of central processing unit. Typically, the central processing unit will receive instructions and data from a read-only memory and / or a random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, the computer will also include one or more large-capacity storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, or the computer will be operably coupled to this large-capacity storage device to receive data from it or to transmit data to it, or both. However, the computer does not necessarily have such a device. In addition, the computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0173] Computer readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including, for example, semiconductor memory devices (e.g., EPROM, EEPROM and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD ROM and DVD-ROM disks. The processor and memory may be supplemented by, or incorporated in, special purpose logic circuitry.

[0174] Although this specification includes many specific implementation details, these should not be interpreted as limiting the scope of any invention or the scope of protection claimed, but are mainly used to describe the features of the specific embodiments of specific inventions. Certain features described in multiple embodiments in this specification may also be implemented in combination in a single embodiment. On the other hand, the various features described in a single embodiment may also be implemented separately in multiple embodiments or in any suitable sub-combination. In addition, although features may work in certain combinations as described above and even initially claim protection, one or more features from the claimed combination may be removed from the combination in some cases, and the claimed combination may point to a sub-combination or a variation of a sub-combination.

[0175] Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that these operations be performed in the particular order shown or performed sequentially, or requiring that all illustrated operations be performed to achieve the desired results. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system modules and components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product, or packaged into multiple software products.

[0176] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some implementations, multitasking and parallel processing may be advantageous.

[0177] The above are preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A deception jamming method for infrared image target recognition, characterized in that: include: Step S1, characterizing the local invariance features of the infrared image; Step S2: adding a tiny disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image, so that the model cannot recognize the disturbance and makes the model make an incorrect classification.

2. The deception jamming method for infrared image target recognition according to claim 1 is characterized in that: In the step S1, characterizing the local invariance features of the infrared image includes: Preprocess the infrared image; According to the invariant feature detection function under the scale space theory, the local feature invariance of the preprocessed infrared image is described to obtain the feature points in the scale space; The feature points in the scale space are processed by corner, edge, spot and region detection algorithms to obtain local feature points; Applying SIFT algorithm to extract the local feature points and generate descriptors; The descriptors are matched by the FLANN approximate nearest neighbor search algorithm.

3. The deception jamming method for infrared image target recognition according to claim 2 is characterized in that: In the step S1, characterizing the local invariance features of the infrared image further includes: The high-dimensional features after feature matching are reduced in dimension through PCA to obtain a low-dimensional feature space, thus completing the construction of local feature spaces of infrared images at different scales.

4. The deception jamming method for infrared image target recognition according to claim 1 is characterized in that: In step S2, adding a small disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image, making the model unable to recognize the disturbance and causing the model to make an incorrect classification includes: Based on the differential evolution algorithm, find the pixel modification direction of the perturbation of the represented image; According to the pixel modification direction, designing a feature similarity disturbance direction of the image; According to the feature-similar perturbation direction, a predefined number of pixels are modified to achieve an image deception effect.

5. A deception jamming system for infrared image target recognition, characterized in that: The system comprises: A first processing module is configured to characterize local invariant features of the infrared image; The second processing module is configured to add a small disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image so that the model cannot recognize the disturbance and makes the model make an incorrect classification.

6. The infrared image target recognition deception jamming system according to claim 5 is characterized in that: The first processing module is specifically configured to: Preprocess the infrared image; According to the invariant feature detection function under the scale space theory, the local feature invariance of the preprocessed infrared image is described to obtain the feature points in the scale space; The feature points in the scale space are processed by corner, edge, spot and region detection algorithms to obtain local feature points; Applying SIFT algorithm to extract the local feature points and generate descriptors; The descriptors are matched by the FLANN approximate nearest neighbor search algorithm.

7. The infrared image target recognition deception jamming system according to claim 6 is characterized in that: The first processing module is specifically configured to: The characterization of the local invariance features of the infrared image also includes: The high-dimensional features after feature matching are reduced in dimension through PCA to obtain a low-dimensional feature space, thus completing the construction of local feature spaces of infrared images at different scales.

8. The infrared image target recognition deception jamming system according to claim 5 is characterized in that: The second processing module is specifically configured to: The method of adding a small disturbance to the characterized infrared image based on the differential evolution algorithm, relying on the local invariant features of the image, makes the model unable to recognize the disturbance and causes the model to make an incorrect classification, including: Based on the differential evolution algorithm, find the pixel modification direction of the perturbation of the represented image; According to the pixel modification direction, designing a feature similarity disturbance direction of the image; According to the feature-similar perturbation direction, a predefined number of pixels are modified to achieve an image deception effect.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps in the deception interference method for infrared image target recognition described in any one of claims 1 to 4 are 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 steps in the deception interference method for infrared image target recognition described in any one of claims 1 to 4 are implemented.

Images