Specific camouflage target identification method and system based on optical remote sensing image

Abstract

The invention relates to the technical field of target positioning and recognition, in particular to a specific camouflage target recognition method and system based on an optical remote sensing image, and aims to mark a specific target in a collected optical remote sensing image data set, grade and sort the camouflage effect of the specific target by using a target camouflage evaluation system, and improve the target positioning and recognition efficiency. Constructing an optical remote sensing image sample set according to a rating and sorting result; training a search and recognition network by using the optical remote sensing image sample set to obtain a target recognition module, the search and recognition network comprising a search unit used for searching camouflage targets in the optical remote sensing image and a recognition unit used for carrying out identity confirmation on all searched camouflage targets, the search unit generates global features for camouflage target identity confirmation by using a texture enhancement mechanism and an attention mechanism; and inputting the to-be-detected optical remote sensing image into the target identification model, and obtaining a specific camouflage target in the to-be-detected optical remote sensing image by using the target identification model. The method can improve the accuracy and effectiveness of camouflage target recognition in a specific task environment, and can be deployed and implemented in an actual confrontation scene.

Description

Technical Field

[0001] The present invention relates to the technical field of target positioning and recognition, and in particular to a specific camouflage target recognition method and system based on optical remote sensing images. Background Art

[0002] Target camouflage and identification have always been research hotspots in the field of information confrontation. The struggle between the two revolves around the difference between the target and the background, and the difference between the target and the background is the fundamental reason for the target to be exposed. Common adversarial targets use camouflage paint, camouflage nets, and other means to change the surface characteristics of the target by using color, texture, and various patterns to achieve the effect of blending into the surrounding complex environment and achieving the purpose of invisibility. Some camouflage effects are even unrecognizable to the human eye. For example, in certain background environments, camouflage nets can achieve the effect of "different objects with the same spectrum" between the target and the background, which poses a great challenge to typical target identification. Camouflaged object detection (COD) can better solve the above problems. It aims to quickly and accurately detect targets hidden in complex backgrounds.

[0003] Satellite remote sensing is an important means of reconnaissance because of its large-scale synchronous observation, high timeliness, and few restrictions on information acquisition, especially without national borders. Remote sensing technology can achieve all-weather, all-day, all-round, and highly dynamic reconnaissance, and can accurately identify and locate typical camouflaged targets. Compared with the rapid development of imaging technology, the recognition of typical camouflaged targets based on high-resolution optical remote sensing images is still in its infancy, and the ability to obtain high-value target information is weak. It is still difficult to meet the application requirements of confrontation tasks in terms of accuracy and real-time performance. To this end, how to use advanced artificial intelligence technology to achieve efficient and intelligent interpretation of massive high-resolution optical remote sensing image data, extract high-value typical camouflaged target information, and improve the intelligence level of intelligence analysis is a key technical problem that needs to be solved urgently. Summary of the invention

[0004] To this end, the present invention provides a specific camouflaged target recognition method and system based on optical remote sensing images, which improves the accuracy and effectiveness of camouflaged target recognition in a specific mission environment.

[0005] According to the design scheme provided by the present invention, on the one hand, a method for identifying a specific camouflaged target based on an optical remote sensing image is provided, comprising:

[0006] The specific targets in the collected optical remote sensing image data set are marked, the camouflage effects of the specific targets are ranked and sorted using the target camouflage evaluation system, and the optical remote sensing image sample set is constructed based on the ranking results;

[0007] The search and recognition network is trained using an optical remote sensing image sample set to obtain a target recognition module, wherein the search and recognition network comprises: a search unit for searching for camouflaged targets in the optical remote sensing image and a recognition unit for confirming the identities of all searched camouflaged targets, and the search unit generates global features for confirming the identities of camouflaged targets using a texture enhancement mechanism and an attention mechanism;

[0008] The optical remote sensing image to be tested is input into the target recognition model, and the target recognition model is used to obtain the specific camouflaged target in the optical remote sensing image to be tested.

[0009] As a specific camouflaged target recognition method based on optical remote sensing images of the present invention, further, a target camouflage evaluation system is used to rank the camouflage effects of specific targets, including:

[0010] Setting characteristic indicators of camouflage effect of specific mission targets in optical remote sensing images, and determining weights of each characteristic indicator using a dynamic weight allocation algorithm, wherein the characteristic indicators include image brightness indicator, image color indicator, image texture indicator and target edge contour indicator;

[0011] The similarity of specific targets in optical remote sensing images is obtained by weighted fusion based on the weights of feature indicators, and a multi-indicator target camouflage evaluation system is established based on the similarity.

[0012] According to the multi-index target camouflage evaluation system, the recognition difficulty of specific targets marked in optical remote sensing images is finely graded and sorted to obtain an optical remote sensing image sample set for camouflaged target recognition.

[0013] As a specific camouflaged target recognition method based on optical remote sensing images of the present invention, further, a dynamic weight allocation algorithm is used to determine the weight of each feature index, including:

[0014] Formulate a weight update strategy based on the influencing factors of the characteristic indicators on the camouflaged target recognition task, wherein the influencing factors include the influence degree of each characteristic indicator, the contribution of each characteristic indicator and the interaction relationship between each characteristic indicator;

[0015] Based on the weight update strategy and using the dynamic weight allocation algorithm, the weight of each feature indicator is determined.

[0016] As a specific camouflaged target recognition method based on optical remote sensing images of the present invention, further, the process of searching for camouflaged targets in the optical remote sensing images by the search unit includes:

[0017] Perform feature extraction on the input optical remote sensing image to obtain a feature map with a preset resolution;

[0018] The texture enhancement mechanism is used to capture the fine-grained texture with amplified context clues in the feature map, the neighbor connection decoder is used to extract the camouflaged target location information, and the fused feature map is obtained by fusing adjacent features.

[0019] Based on the pixels contained in the feature map, the global spatial context of the feature map is aggregated using a self-attention mechanism to obtain images containing specific disguised targets.

[0020] As a specific camouflaged target recognition method based on optical remote sensing images of the present invention, further, the process of aggregating the global spatial context of the feature map using the self-attention mechanism is expressed as:

[0021] Among them, P i and Q i represents the feature map corresponding to feature map i, N i is the number of pixels contained in feature map i, j, m, n∈{1, Ni} represents the matrix index of each pixel, and w k and w v is the linear transformation matrix used to project the feature map, For P i and Q i Weight matrices of the same shape.

[0022] As a specific camouflaged target recognition method based on optical remote sensing images of the present invention, further, the process of the recognition unit confirming the identity of all searched camouflaged targets includes:

[0023] All the searched camouflaged target images are inverted and cascaded grouped and embedded to obtain a combined feature map;

[0024] The combined feature map is input into an adaptive detection head, which is used to perform decoupled classification and locate the output camouflaged target. The adaptive detection head adopts a decoupled method and uses parallel convolution branches to perform camouflaged target classification and regression frame positioning.

[0025] On the other hand, the present invention also provides a specific camouflaged target recognition system based on optical remote sensing images, comprising: a sample collection module, a model training module and a target recognition module, wherein:

[0026] The sample collection module is used to mark specific targets in the collected optical remote sensing image data set, rank the camouflage effects of specific targets using the target camouflage evaluation system, and construct an optical remote sensing image sample set based on the ranking results;

[0027] A model training module is used to train a search and recognition network using an optical remote sensing image sample set to obtain a target recognition module, wherein the search and recognition network includes: a search unit for searching for camouflaged targets in the optical remote sensing image and a recognition unit for confirming the identities of all searched camouflaged targets, and the search unit generates global features for confirming the identities of camouflaged targets using a texture enhancement mechanism and an attention mechanism;

[0028] The target recognition module is used to input the optical remote sensing image to be tested into the target recognition model, and use the target recognition model to obtain a specific camouflaged target in the optical remote sensing image to be tested.

[0029] Beneficial effects of the present invention:

[0030] Aiming at the situation that traditional camouflage effect evaluation mostly relies on subjective judgment of judges and the conclusions obtained are greatly affected by the environment, the present invention sets objective standard camouflage effect evaluation indicators of typical specific task targets of remote sensing images to reduce the influence of environmental factors, and aims at the problems that the imaging scale and illumination of remote sensing images vary greatly, the background is complex, the size of remote sensing camouflage targets is small, and the camouflage degree is high. The texture enhancement mechanism (TEM) and the nearest neighbor connection decoder (NCD) in the search and recognition network SINet are introduced to capture more recognizable features and position information; after feature fusion, the attention mechanism is used to further optimize the fused features, the significance of the features at different scales is adaptively learned, and the small target detection performance is enhanced; the decoupled classification and positioning tasks are completed by an adaptive detection head (ADH) based on a task learning mechanism (TLM), the target recognition accuracy and positioning precision are enhanced, and the present invention can be deployed and implemented in actual confrontation scenarios, and has good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a schematic diagram of a specific camouflaged target recognition process based on optical remote sensing images in an embodiment;

[0032] Figure 2 This is a schematic diagram of the SINet framework of the target recognition model in the embodiment;

[0033] Figure 3 This is an illustration of the context aggregation principle in the embodiment;

[0034] Figure 4 This is the evaluation of the multi-attribute camouflage effect in the embodiment. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solutions and advantages of the present invention clearer and more understandable, the present invention is further described in detail below in conjunction with the accompanying drawings and technical solutions.

[0036] Remote sensing image target recognition is an important branch of computer vision tasks. It is similar to the method of image target recognition in natural scenes. However, remote sensing images are significantly different from natural images in terms of imaging mode, image quality, image size, etc. The type, perspective, background, etc. of the targets in the images are also significantly different from those in natural images, which makes remote sensing image target recognition more difficult. Due to the differences from natural background images, when applying target recognition and detection methods to remote sensing images, certain improvements need to be made to adapt to the characteristics of remote sensing images, so as to achieve better target recognition effects. For an embodiment of the present invention, see Figure 1 As shown, a specific camouflaged target recognition method based on optical remote sensing images is provided, comprising:

[0037] S101, annotating specific targets in the collected optical remote sensing image data set, using a target camouflage evaluation system to rank the camouflage effects of the specific targets, and constructing an optical remote sensing image sample set based on the ranking results.

[0038] Collect satellite remote sensing images containing specific scenes, manually annotate the camouflaged targets in the images, rate and rank the camouflaged effects of the camouflaged targets under the guidance of the camouflage evaluation system, and establish a sample data set.

[0039] Specifically, the target camouflage evaluation system is used to rank the camouflage effects of specific targets, which can be designed to include:

[0040] Setting characteristic indicators of camouflage effect of specific mission targets in optical remote sensing images, and determining weights of each characteristic indicator using a dynamic weight allocation algorithm, wherein the characteristic indicators include image brightness indicator, image color indicator, image texture indicator and target edge contour indicator;

[0041] The similarity of specific targets in optical remote sensing images is obtained by weighted fusion based on the weights of feature indicators, and a multi-indicator target camouflage evaluation system is established based on the similarity.

[0042] According to the multi-index target camouflage evaluation system, the recognition difficulty of specific targets marked in optical remote sensing images is finely graded and sorted to obtain an optical remote sensing image sample set for camouflaged target recognition.

[0043] The dynamic weight allocation algorithm is used to determine the weight of each characteristic indicator, which may include:

[0044] Formulate a weight update strategy based on the influencing factors of the characteristic indicators on the camouflaged target recognition task, wherein the influencing factors include the influence degree of each characteristic indicator, the contribution of each characteristic indicator and the interaction relationship between each characteristic indicator;

[0045] Based on the weight update strategy and using the dynamic weight allocation algorithm, the weight of each feature indicator is determined.

[0046] In the scope of optical camouflage, the factors affecting the camouflage effect include target brightness coefficient, color, texture, geometric characteristics and spectral reflection characteristics. Among them, the spectral reflection characteristics are related to the reflection of the color characteristics of the object surface. The spectral reflection characteristics of the target can be reflected by extracting the color characteristics, and the geometric characteristics can be reflected by extracting the edge contour characteristics of the target itself. In terms of the human eye visual perception system, the evaluation of the camouflage effect under visible light mainly affects the subjective perception of human vision. The target recognition characteristics under visible light include brightness, color, edge, shape, boundary contrast and texture difference. The selection of index features should be based on the comprehensiveness of the overall index, while ensuring the independence of individual indicators. According to current research, the brightness index, color index and texture index have greater visual stimulation to the human eye. The edge of the object can reflect its external shape characteristics. At the same time, whether the object can be directly integrated into the background at the boundary is the decisive factor in whether the target can be discovered. Therefore, these four independent indicators can be selected to construct an evaluation index system.

[0047] Generally speaking, the brightness of the target will directly affect the fusion effect between the target and the background. The smaller the difference between the overall brightness of the target and the surrounding background environment, the better the target can blend into the environment and achieve a good camouflage purpose. In extracting the target brightness index, we must first select a suitable color space. The color of a normal color image is in the RGB color space, which is not suitable for calculating and processing images. Combined with specific feature extraction, the CIE Lab color space

[35] is often used in the processing of brightness features. The Lab color space can fully retain the brightness information. The CIE Lab mode is a color mode based on the physiological characteristics of the human eye and can specifically express the visual perception of the human eye. It contains three basic elements: L (brightness), a (green and red color channels), and b (blue and yellow color channels). Brightness is a separate information channel, so it can be processed separately to extract the brightness index of the image.

[0048] The relationship between the brightness value L and the XYZ color space conversion is as follows:

[0049]

[0050] In the formula, the reference white light radiation brightness value is Y 0 =100.

[0051] As a widely used image feature, color feature extraction is relatively easy compared to other features, and is less affected by geometric transformations and has strong robustness. As one of the widely used color feature extraction methods, the color histogram is a method that describes the proportion of different colors in an image to the color of the entire image through statistical distribution characteristics. This method can only describe the proportion of the number of pixels of the same color, and cannot record the spatial position of the color, so it is not affected by the geometric transformation of the image. Normalizing the histogram can facilitate the statistical difference in color features, which refers to the ratio of the number of pixels of the same color to the total number of pixels, as shown in the following formula:

[0052] H(k)=n k / N,k=0,1,…L-1

[0053] Where k is the color value; L is the number of all color values; n k is the number of pixels with color value k; N is the total number of pixels.

[0054] Texture in an image is often used to describe the grayscale changes of each pixel in spatial distribution, including visual features such as roughness, periodicity, and directionality. Whether the target and background can be well integrated and whether the overall distribution characteristics are coordinated visually can be described by comparing the texture differences between the two. When the texture difference between the target and the background is within a certain range, it is considered that the texture similarity between the target and the background is high. The gray level co-occurrence matrix can well describe the spatial distribution of texture features and calculate the texture similarity between images.

[0055] Gray-level co-occurrence matrix is ​​a second-order statistical method that can not only reflect the spatial distribution difference of pixel gray levels, but also describe the spatial information of the relative position between pixels. Pixels in an image can be expressed by position coordinates and gray levels. Suppose the image has X pixels in the horizontal direction, Y pixels in the vertical direction, and the gray level is w. The image information expression is as follows:

[0056]

[0057] The extraction of the boundary line between the target and the background is called edge detection. The commonly used edge operators are the classic template operator method and the linear filtering method. The classic template operator belongs to a nonlinear filtering algorithm with simple calculations and is suitable for edge extraction of images with simple shapes. Linear filtering methods include the Canny operator, wavelet edge detection operator, LOG operator, etc. Among them, the Canny operator is more commonly used in edge detection, and has a better effect on edge contour extraction, which is suitable for edge extraction of images with complex noise. In the embodiment of this case, the Canny operator can be used to obtain the contour edge of the target in the remote sensing image, the image can be smoothed by using a Gaussian filter to filter out the noise in the remote sensing image, and the Sobel operator is used to calculate the horizontal and vertical gradient values ​​of each pixel in the image, and the gradient amplitude and direction of each pixel are obtained according to the gradient value; non-maximum suppression is performed on the obtained gradient amplitude image to narrow the edge width, making the edge more refined and clear; according to the set high threshold and low threshold, the pixels in the gradient amplitude image are divided into three categories: strong edge, weak edge and non-edge, and the complete edge image is obtained by connecting the strong edge pixels and the weak edge pixels adjacent to them.

[0058] S102. Using an optical remote sensing image sample set to train a search and recognition network to obtain a target recognition module, the search and recognition network includes: a search unit for searching for camouflaged targets in the optical remote sensing image and a recognition unit for confirming the identities of all searched camouflaged targets, and the search unit generates global features for confirming the identities of camouflaged targets using a texture enhancement mechanism and an attention mechanism.

[0059] The process of searching for camouflaged targets in the optical remote sensing image by the search unit can be designed to include:

[0060] Perform feature extraction on the input optical remote sensing image to obtain a feature map with a preset resolution;

[0061] The texture enhancement mechanism is used to capture the fine-grained texture with amplified context clues in the feature map, the neighbor connection decoder is used to extract the camouflaged target location information, and the fused feature map is obtained by fusing adjacent features.

[0062] Based on the pixels contained in the feature map, the global spatial context of the feature map is aggregated using a self-attention mechanism to obtain images containing specific disguised targets.

[0063] The search and recognition network SINet is inspired by biological vision and includes the first two stages of animal hunting, namely search and recognition. Specifically, the search stage is responsible for searching for hidden targets, while the recognition stage accurately detects hidden targets in a cascaded manner. Figure 2As shown, in the search stage, the texture enhancement mechanism (TEM) is used to extend the context clues to capture fine-grained textures, and the nearest neighbor decoder (NCD) can provide positioning information.

[0064] The self-attention mechanism is a special attention mechanism that imitates biological visual saliency detection and selective attention, performs attention calculation on the sequence itself, assigns different weight information to different elements to obtain the internal connection of the sequence, makes the network pay more attention to useful information, alleviates the interference caused by complex background information, and improves the accuracy and robustness of the detection model.

[0065] The context aggregation mechanism CABlock is a plug-and-play self-attention mechanism that can effectively improve the network detection accuracy. It is to enhance the network's ability to locate targets in complex backgrounds while slightly increasing the number of network parameters. The structure of CABlock is as follows Figure 3 As shown in the figure, represents the Kronecker product of the matrix, and ⊙ represents the Hadamard product of the matrix. The spatial context calculation formula for the input feature map i is as follows:

[0066]

[0067] Where P i and Q i Represents the feature map corresponding to the feature map, N i is the number of pixels contained in feature map i. j,m,n∈{1,N i} represents the matrix index of each pixel. k and w v is the linear transformation matrix used to project the feature map, is with P i and Q i A weight matrix of the same shape that balances how much each pixel aggregates the global spatial context.

[0068] The process of the recognition unit confirming the identity of all the camouflaged targets found can be designed to include:

[0069] All the searched camouflaged target images are inverted and cascaded grouped and embedded to obtain a combined feature map;

[0070] The combined feature map is input into an adaptive detection head, which is used to perform decoupled classification and locate the output camouflaged target. The adaptive detection head adopts a decoupled method and uses parallel convolution branches to perform camouflaged target classification and regression frame positioning.

[0071] The adaptive detection head (ADH) based on the task learning mechanism (TLM) is used to decouple the classification and positioning tasks, thereby enhancing the target recognition accuracy and positioning precision.

[0072] S103, inputting the optical remote sensing image to be tested into a target recognition model, and using the target recognition model to obtain a specific camouflaged target in the optical remote sensing image to be tested.

[0073] In view of the current lack of typical remote sensing image data sets of camouflaged targets for specific tasks and the existing camouflage effect evaluation methods in the field of optical reconnaissance that rely on the subjective judgment of the judge, are greatly affected by the environment, and cannot achieve rapid, objective and accurate evaluation, in the embodiment of this case, by analyzing and extracting multiple indicators that affect the camouflage effect of specific task targets in remote sensing images, according to the degree of influence, contribution and interaction relationship of each feature indicator, a dynamic weight allocation algorithm that conforms to the camouflage performance is used to determine the size of the weight of each indicator, and based on the weighted similarity of each feature, a camouflage effect evaluation system based on multiple indicators is established; the basic idea of ​​constructing a camouflage effect evaluation system is to evaluate the four indicator attribute values ​​between remote sensing images of a limited number of different camouflage effects of a specific target (specifically including brightness contrast value, color similarity, texture similarity and edge contour similarity), and complete the ranking of the target schemes by weight allocation and calculation of the evaluation coefficients of each scheme, and select the image with the largest evaluation coefficient as the best camouflage effect. The multi-attribute decision evaluation step includes the following three steps: first, the decision attribute matrix of the initial scheme is normalized, second, the weight of the indicator scheme is determined by the entropy method, and finally all schemes are comprehensively evaluated and ranked. Steps such as Figure 4 As shown, the specific steps are as follows:

[0074] (1) Establishing the initial decision attribute matrix

[0075] Assume that the total number of camouflaged images of a specific camouflaged target is M, and the camouflaged image set U = {u 1 ,u 2 ,…,u m}; record the four characteristic index set V = {v 1 ,v 2 ,v 3 ,v 4}; Record the disguised image U i For the index V j The attribute value is a ij , establish the initial decision attribute matrix set A = (A ij ) m×4 .

[0076] (2) Normalization of the initial decision attribute matrix

[0077] The initial matrix is ​​normalized to avoid the situation where the indicators cannot be compared due to different dimensions and orders of magnitude. The range transformation method is used to process the data. The normalized standard decision attribute matrix X = (x ij ) m×4 , the expression is as follows:

[0078]

[0079] (3) Calculate weights using entropy method

[0080] In the weight distribution of camouflage indicators, the entropy method is used to calculate the weight distribution between each indicator. The initial attribute decision matrix A is composed of m camouflage images and 4 evaluation feature indicators. ij ) m×4 , after normalization, we get the standard decision attribute matrix X = (X ij ) m×4 . Then the weight y of the index value under the jth index of the i-th disguised image is ij :

[0081]

[0082] In the formula, represents the sum of all disguised image attribute values ​​in the jth feature index, and thus the standardized matrix Y of the data can be obtained = {y ij} m×4 , where 0≤yij≤1. The expression is as follows:

[0083]

[0084] The information entropy value of the i-th feature index is expressed as e j . Index value x ij The greater the difference, the greater the evaluation effect on the image camouflage effect. At this time, the information entropy value e j The smaller the information entropy e j The calculation formula is as follows:

[0085]

[0086] Where k>0, e j ≥0.

[0087] Thus, the information entropy value of the four characteristic indicators is obtained as follows: 1 ,e 2 ,e 3 ,e 4}.

[0088] The weight value of the jth indicator:

[0089]

[0090] The weight vector w = {w 1 ,w 2 ,w 3 ,w 4}.

[0091] (4) TOPSIS method comprehensive evaluation and ranking

[0092] After determining the attribute values ​​and indicator weight distribution of the characteristic indicators of the evaluation system, the TOPSIS method is used to conduct a comprehensive evaluation and ranking of the camouflage effects by measuring the distance between each sample and the positive and negative ideal solutions.

[0093] Then, based on open source intelligence and optical remote sensing image data, remote sensing images are collected and fine-grained typical specific mission targets are manually annotated to determine the fine-grained attributes of each target, such as field command posts, tanks, special vehicles, etc. Finally, based on the camouflage effect evaluation system, the difficulty of camouflaged target recognition is graded and sorted, and a fine-grained typical camouflaged target remote sensing image dataset is established. The dataset is used to train and optimize the target recognition model, and the trained target recognition model is used to efficiently and real-time detect and identify camouflaged targets in optical remote sensing images in the current task.

[0094] Furthermore, based on the above method, an embodiment of the present invention also provides a specific camouflaged target recognition system based on optical remote sensing images, comprising: a sample collection module, a model training module and a target recognition module, wherein:

[0095] The sample collection module is used to mark specific targets in the collected optical remote sensing image data set, rank the camouflage effects of specific targets using the target camouflage evaluation system, and construct an optical remote sensing image sample set based on the ranking results;

[0096] A model training module is used to train a search and recognition network using an optical remote sensing image sample set to obtain a target recognition module, wherein the search and recognition network includes: a search unit for searching for camouflaged targets in the optical remote sensing image and a recognition unit for confirming the identities of all searched camouflaged targets, and the search unit generates global features for confirming the identities of camouflaged targets using a texture enhancement mechanism and an attention mechanism;

[0097] The target recognition module is used to input the optical remote sensing image to be tested into the target recognition model, and use the target recognition model to obtain a specific camouflaged target in the optical remote sensing image to be tested.

[0098] Unless otherwise specifically stated, the relative steps, numerical expressions and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present invention.

[0099] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0100] The units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person of ordinary skill in the art may use different methods to implement the described functions for each specific application, but such implementation is not considered to be beyond the scope of the present invention.

[0101] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits, and accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software function modules. The present invention is not limited to any specific form of combination of hardware and software.

[0102] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for identifying a specific camouflaged target based on optical remote sensing images, characterized in that: Include: The specific targets in the collected optical remote sensing image data set are marked, the camouflage effects of the specific targets are ranked and sorted using the target camouflage evaluation system, and the optical remote sensing image sample set is constructed based on the ranking results; The search and recognition network is trained using an optical remote sensing image sample set to obtain a target recognition module, wherein the search and recognition network comprises: a search unit for searching for camouflaged targets in the optical remote sensing image and a recognition unit for confirming the identities of all searched camouflaged targets, and the search unit generates global features for confirming the identities of camouflaged targets using a texture enhancement mechanism and an attention mechanism; The optical remote sensing image to be tested is input into the target recognition model, and the target recognition model is used to obtain the specific camouflaged target in the optical remote sensing image to be tested.

2. The specific camouflage target recognition method based on optical remote sensing images according to claim 1 is characterized in that: The target camouflage evaluation system is used to rank the camouflage effects of specific targets, including: Setting characteristic indicators of camouflage effect of specific mission targets in optical remote sensing images, and determining weights of each characteristic indicator using a dynamic weight allocation algorithm, wherein the characteristic indicators include image brightness indicator, image color indicator, image texture indicator and target edge contour indicator; The similarity of specific targets in optical remote sensing images is obtained by weighted fusion based on the weights of feature indicators, and a multi-indicator target camouflage evaluation system is established based on the similarity. According to the multi-index target camouflage evaluation system, the recognition difficulty of specific targets marked in optical remote sensing images is finely graded and sorted to obtain an optical remote sensing image sample set for camouflaged target recognition.

3. The specific camouflage target recognition method based on optical remote sensing images according to claim 2 is characterized in that: The process of obtaining the similarity of specific targets in optical remote sensing images by weighted fusion based on feature index weights includes:

4. The specific camouflage target recognition method based on optical remote sensing images according to claim 2 is characterized in that: The dynamic weight allocation algorithm is used to determine the weight of each feature indicator, including: Formulate a weight update strategy based on the influencing factors of the characteristic indicators on the camouflaged target recognition task, wherein the influencing factors include the influence degree of each characteristic indicator, the contribution of each characteristic indicator and the interaction relationship between each characteristic indicator; Based on the weight update strategy and using the dynamic weight allocation algorithm, the weight of each feature indicator is determined.

5. The specific camouflage target recognition method based on optical remote sensing images according to claim 1 is characterized in that: The process of searching for camouflaged targets in optical remote sensing images by the search unit includes: Perform feature extraction on the input optical remote sensing image to obtain a feature map with a preset resolution; The texture enhancement mechanism is used to capture the fine-grained texture with amplified context clues in the feature map, the neighbor connection decoder is used to extract the camouflaged target location information, and the fused feature map is obtained by fusing adjacent features. Based on the pixels contained in the feature map, the global spatial context of the feature map is aggregated using a self-attention mechanism to obtain images containing specific disguised targets.

6. The specific camouflage target recognition method based on optical remote sensing images according to claim 4 is characterized in that: The process of aggregating the global spatial context of the feature map using the self-attention mechanism is expressed as: Among them, P i and Q i represents the feature map corresponding to feature map i, N i is the number of pixels contained in feature map i, j,m,n∈{1,N i } represents the matrix index of each pixel, w k and w v is the linear transformation matrix used to project the feature map, a i j For P i and Q i Weight matrices of the same shape.

7. The specific camouflage target recognition method based on optical remote sensing images according to claim 1 is characterized in that: The process of the recognition unit confirming the identity of all the camouflaged targets found includes: All the searched camouflaged target images are inverted and cascaded grouped and embedded to obtain a combined feature map; The combined feature map is input into an adaptive detection head, which is used to perform decoupled classification and locate the output camouflaged target. The adaptive detection head adopts a decoupled method and uses parallel convolution branches to perform camouflaged target classification and regression frame positioning.

8. A specific camouflage target recognition system based on optical remote sensing images, characterized in that: It includes: sample collection module, model training module and target recognition module, among which, The sample collection module is used to mark specific targets in the collected optical remote sensing image data set, rank the camouflage effects of specific targets using the target camouflage evaluation system, and construct an optical remote sensing image sample set based on the ranking results; A model training module is used to train a search and recognition network using an optical remote sensing image sample set to obtain a target recognition module, wherein the search and recognition network includes: a search unit for searching for camouflaged targets in the optical remote sensing image and a recognition unit for confirming the identities of all searched camouflaged targets, and the search unit generates global features for confirming the identities of camouflaged targets using a texture enhancement mechanism and an attention mechanism; The target recognition module is used to input the optical remote sensing image to be tested into the target recognition model, and use the target recognition model to obtain a specific camouflaged target in the optical remote sensing image to be tested.

9. An electronic device, characterized in that: include: at least one processor, and a memory coupled to the at least one processor; The memory stores a computer program, and the computer program can be executed by the at least one processor to implement the method according to any one of claims 1 to 7.

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, the method according to any one of claims 1 to 7 can be implemented.

Images