Self-adaptive feature selection-based camouflage target detection method and device

Abstract

The invention relates to a camouflage target detection method and device for adaptive feature selection, and the method comprises the steps: carrying out the frequency feature extraction processing of an input image, and obtaining a high-frequency feature group and a low-frequency feature group, the high-frequency feature group at least comprises two high-frequency feature patterns, and the low-frequency feature pattern comprises two low-frequency feature patterns; the low-frequency feature group at least comprises two low-frequency feature maps; determining adaptive features based on the high-frequency feature group and the low-frequency feature group by using an adaptive selection strategy; and detecting a camouflage target object in the input image by using the adaptive features. According to the embodiment of the invention, the detection accuracy of the camouflage target is improved by utilizing the self-adaptive correlation characteristics of high frequency and low frequency.

Description

Technical Field

[0001] This disclosure relates to image processing technology, and in particular to a method and apparatus for detecting camouflaged targets with adaptive feature selection, an electronic device, and a storage medium. Background Technology

[0002] Camouflaged object detection (COD) is an important research area in computer vision, aiming to automatically identify and locate target objects in complex natural scenes that are highly similar to the background in terms of color, texture, or structural features. These targets typically possess significant concealment and are widely used in applications such as natural ecological monitoring, military reconnaissance, medical image analysis, and industrial inspection, placing high demands on the feature representation and discrimination capabilities of detection algorithms. Most existing COD methods utilize frequency domain feature extraction to extract and enhance low-frequency and high-frequency features separately. However, these methods often fail to establish an adaptive correlation between high-frequency and low-frequency features, making it difficult to fully utilize the complementary information of different frequency domain features, thus potentially leading to missed or false detections in complex backgrounds. Summary of the Invention

[0003] This disclosure proposes an adaptive feature selection method, apparatus, electronic device, and storage medium for camouflaged target detection, which addresses the problem of low detection rate of camouflaged targets in the prior art.

[0004] According to a first aspect of this disclosure, an adaptive feature selection method for detecting camouflaged targets is provided, comprising: performing frequency feature extraction processing on an input image to obtain a high-frequency feature group and a low-frequency feature group, wherein the high-frequency feature group includes at least two high-frequency feature maps and the low-frequency feature group includes at least two low-frequency feature maps; determining adaptive features based on the high-frequency feature group and the low-frequency feature group using an adaptive selection strategy; and detecting camouflaged target objects in the input image using the adaptive features.

[0005] In some possible implementations, performing frequency feature extraction processing on the input image to obtain a high-frequency feature group and a low-frequency feature group includes: performing Fourier transform or wavelet transform on the input image to obtain the high-frequency feature group and the low-frequency feature group.

[0006] In some possible implementations, an adaptive selection strategy is used to determine adaptive features based on the high-frequency feature group and the low-frequency feature group, including: performing a first fusion process on the high-frequency feature map of the high-frequency feature group to obtain a first fused feature; obtaining at least two first connection features based on the different connection methods of each low-frequency feature in the low-frequency feature map; and determining the adaptive feature based on a first similarity between the first fused feature and the first connection feature. Alternatively, the method of using an adaptive selection strategy to determine adaptive features based on the high-frequency feature group and the low-frequency feature group includes: performing a second fusion process on the low-frequency feature map of the low-frequency feature group to obtain a second fusion feature; obtaining at least two second connection features based on the different connection methods of each high-frequency feature in the high-frequency feature map; and determining the adaptive feature based on the second similarity between the second fusion feature and the second connection feature.

[0007] In some possible implementations, performing a first fusion process on the high-frequency feature maps of the high-frequency feature group to obtain a first fused feature includes: performing connection fusion processes on the high-frequency feature maps of the high-frequency feature group in the height direction, width direction, and diagonal direction respectively to obtain high-frequency connection features; and performing continuous convolution and residual processing on the high-frequency connection features to obtain the first fused feature.

[0008] In some possible implementations, determining the adaptive feature based on the first similarity between the first fusion feature and the first connection feature includes: calculating the first similarity between the first fusion feature and each of the first connection feature maps; determining the first connection feature whose first similarity satisfies a condition greater than a first threshold as a selected low-frequency feature map; and determining the adaptive feature based on the first fusion feature and the selected low-frequency feature map.

[0009] In some possible implementations, determining the adaptive feature based on the second similarity between the second fusion feature and the second connection feature includes: calculating the second similarity between the second fusion feature and each of the second connection feature maps; determining the second connection feature whose second similarity satisfies a condition greater than a first threshold as a selected high-frequency feature map; and determining the adaptive feature based on the second fusion feature and the high-frequency and low-frequency feature maps.

[0010] In some possible implementations, detecting a disguised target object in the input image using the adaptive features includes: performing basic convolution processing on the adaptive features to obtain basic features; and performing at least one joint convolution and residual processing on the basic features to obtain target features. The location of the disguised target object is determined based on the target features.

[0011] According to a second aspect of this disclosure, a disguised target detection apparatus is provided, comprising: a frequency feature extraction module for performing frequency feature extraction processing on an input image to obtain a high-frequency feature group and a low-frequency feature group, wherein the high-frequency feature group includes at least two high-frequency feature maps and the low-frequency feature group includes at least two low-frequency feature maps; an adaptive module for determining adaptive features based on the high-frequency feature group and the low-frequency feature group using an adaptive selection strategy; and a detection module for detecting disguised target objects in the input image using the adaptive features.

[0012] According to a third aspect of this disclosure, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to perform the method described in any one of the first aspects.

[0013] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided that stores computer program instructions thereon, which, when executed by a processor, implement the method described in any one of the first aspects.

[0014] This disclosure provides an adaptive feature selection method for camouflaged target detection. This method extracts high-frequency and low-frequency features from an input image to obtain high-frequency and low-frequency feature groups. These are then combined with an adaptive selection strategy to obtain adaptive features, which are then used for camouflaged target detection. This disclosure further establishes an adaptive correlation between high-frequency and low-frequency features, fully utilizing the complementary information of different frequency domain features, thereby improving detection accuracy while reducing false negatives or negatives.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0016] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the specification, serve to illustrate the technical solutions of this disclosure.

[0018] Figure 1 This is a flowchart of the camouflage target detection method with adaptive feature selection in the embodiments of this disclosure; Figure 2 This is a block diagram of the adaptive feature extraction model according to an embodiment of the present disclosure; Figure 3 This is a schematic diagram of the channel compression and fusion structure; Figure 4This is a block diagram illustrating the combined convolution and residual processing of embodiments of this disclosure. Detailed Implementation

[0019] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0020] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0021] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0022] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0023] The method disclosed herein can be executed by an image processing device, such as a terminal device, a server, or other processing device. The terminal device can be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, in-vehicle device, wearable device, etc. In some possible implementations, the method can be implemented by a processor calling computer-readable instructions stored in memory.

[0024] It is understood that the various method embodiments mentioned above in this disclosure can be combined with each other to form combined embodiments without violating the principle and logic. Due to space limitations, this disclosure will not elaborate further.

[0025] Figure 1 This is a flowchart of the camouflaged target object detection method in the embodiments of this disclosure; as follows: Figure 1 As shown, the method for detecting camouflaged target objects includes: S10: Perform frequency feature extraction processing on the input image to obtain a high-frequency feature group and a low-frequency feature group, wherein the high-frequency feature group includes at least two high-frequency feature maps and the low-frequency feature group includes at least two low-frequency feature maps; In some possible implementations, a color image of the camouflaged object can be obtained using an image acquisition device, which is the image in the example of this disclosure. Specifically, the input image in the embodiment of this disclosure includes the camouflaged object to be detected by this disclosure. The camouflaged object can be any species, object, or medical detection target, for example, it can be a rabbit, but this is not a specific limitation of the embodiment of this disclosure.

[0026] In this embodiment, frequency feature extraction processing can be performed on the input image. By performing frequency domain analysis on the image, the image features are divided into a high-frequency feature group and a low-frequency feature group. The high-frequency feature group includes at least two high-frequency feature maps, and the low-frequency feature group includes at least two low-frequency feature maps. The high-frequency feature group is used to characterize high-frequency information related to edge variations and texture details in the image, while the low-frequency feature group is used to characterize the overall structure, regional distribution, and global semantic information in the image. By separating the frequency features of the input image, different frequency components can be modeled and utilized separately in subsequent processing, thereby providing basic feature input for subsequent differentiated fusion processing of high-frequency and low-frequency features.

[0027] S20: Using an adaptive selection strategy, determine adaptive features based on the high-frequency feature group and the low-frequency feature group; In this embodiment, the low-frequency feature maps in the low-frequency feature group can be fused to obtain fused features representing the overall structure and semantic information of the target. Simultaneously, multiple connection features are constructed based on different connection methods of the high-frequency feature maps in the high-frequency feature group to enhance the correlation expression of high-frequency features in different spatial directions. Then, the similarity between the low-frequency fused features and each of the high-frequency connection features is calculated. Based on the similarity magnitude, high-frequency connection features with high correlation to the low-frequency fused features are adaptively selected. The selected high-frequency connection features are then synergistically fused with the low-frequency fused features to obtain adaptive features that simultaneously consider both overall structural information and local detail information.

[0028] Alternatively, a fusion process can be performed on the high-frequency features in the high-frequency feature group, and then multiple connection features can be constructed based on different connection methods of the low-frequency feature maps in the low-frequency feature group. The similarity between the high-frequency fusion features and each of the low-frequency connection features can be calculated, adaptive feature association can be performed, and the above adaptive feature extraction process can be executed.

[0029] S30: Detect the disguised target object in the input image using the adaptive features.

[0030] In this embodiment of the disclosure, the adaptive features can be subjected to basic convolution processing to extract basic spatial features and generate basic features; then, at least one joint convolution and residual processing is performed on the basic features to enhance the feature representation ability and obtain more discriminative target features; finally, based on the target features, analysis and reasoning are performed to determine the specific location of the disguised target object in the input image, thereby achieving accurate detection and localization of the target.

[0031] Based on the above configuration, this embodiment integrates the high-frequency and low-frequency feature groups of the input image into the same camouflage target detection framework for processing. By fusing and modeling the high-frequency features, the edge details and local differences of the camouflage target in the image can be effectively represented. By fusing and modeling the low-frequency features, the overall structure and semantic information of the camouflage target can be stably depicted. Furthermore, by further fusing the high-frequency and low-frequency fusion features, the synergistic expression of detail and structural information is achieved, ultimately used for the detection of camouflage targets in the input image. Through this method, the ability to perceive subtle differences between camouflage targets and complex backgrounds can be effectively enhanced, thereby improving the detection accuracy and robustness of camouflage targets.

[0032] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings. First, an input image can be obtained in the embodiments of this disclosure. The input image is a color image including a disguised target object. In the embodiments of this disclosure, the method of obtaining the input image may include at least one of the following methods: A) Directly use an image acquisition device to acquire color images including camouflaged target objects; wherein, the image acquisition device can be a color camera, video recording device or other device with image acquisition capabilities, and the specific type of image acquisition device is not limited in this disclosure embodiment.

[0033] B) Transmitting and receiving color images including camouflaged target objects via electronic devices; in this embodiment of the disclosure, color images transmitted by other electronic devices can be received via wired and / or wireless communication, and the communication method is not a specific limitation of this disclosure.

[0034] C) Reading a color image including a disguised target object from a storage medium or database; in this embodiment of the disclosure, a pre-stored color image can be read from a local storage device or server according to a received read instruction, and this disclosure does not specifically limit this.

[0035] It should be noted that the input images in the embodiments of this disclosure may come from the same scene or different scenes, and the type, shape and background environment of the disguised target object in the image can be set according to the actual application requirements. Those skilled in the art can select the corresponding image source according to the specific application scenario, and no specific limitation is made here.

[0036] After obtaining a color image including a disguised target object, disguised target object detection processing can be performed on the color image. Specifically, in this embodiment of the disclosure, the color image can be used as input and fed into a pre-built disguised target object detection model for processing, so as to realize the identification and location of the disguised target object in the image; The camouflaged target detection model in this embodiment is constructed based on frequency feature analysis, including multi-path feature processing for high-frequency and low-frequency feature groups. The detection model first performs frequency feature extraction on the input image, dividing the image features into high-frequency and low-frequency feature groups. Then, it fuses and models the high-frequency and low-frequency feature groups to obtain adaptive feature representations that simultaneously characterize local detail information and overall structural semantic information. Finally, it uses the adaptive features for convolution and residual processing to generate discriminative features for target detection, and determines the position of the camouflaged target in the input image based on these features. Through the above processing, this embodiment can achieve efficient and accurate detection of camouflaged target objects while maintaining the diversity of the input image and simultaneously utilizing the texture detail information and global structural information in the image.

[0037] In some possible implementations, given an input image, frequency feature extraction is performed on the input image to obtain a high-frequency feature group and a low-frequency feature group. The high-frequency feature group includes at least two high-frequency feature maps, and the low-frequency feature group includes at least two low-frequency feature maps.

[0038] In some possible implementations, a Fourier transform or wavelet transform is performed on the input image to obtain the high-frequency feature group and the low-frequency feature group. In this embodiment, the input image can be a color image or a grayscale image, and includes a camouflaged target object to be detected. To fully characterize the local details and overall structural information of the image, a Fourier transform or wavelet transform is performed on the input image to obtain the high-frequency feature group and the low-frequency feature group. The high-frequency feature group includes at least two high-frequency feature maps to characterize high-frequency information related to edge changes and texture details in the image; the low-frequency feature group includes at least two low-frequency feature maps to characterize the overall structure, regional distribution, and global semantic information of the image. Specifically, through Fourier transform, the image can be mapped from the spatial domain to the frequency domain, and high-frequency and low-frequency information can be divided according to the amplitude of the frequency components; or, through wavelet transform, the image can be decomposed into frequency sub-bands of different scales and directions, thereby obtaining multi-scale, high-resolution high- and low-frequency features. This step effectively separates the frequency features of the input image, enabling high-frequency and low-frequency features to be modeled and utilized separately in subsequent adaptive selection and fusion processing stages, providing rich basic feature inputs for camouflaged target detection.

[0039] Given low-frequency and high-frequency feature groups, an adaptive selection strategy can be further utilized to determine adaptive features based on the high-frequency and low-frequency feature groups. In this embodiment, feature fusion processing can be performed on one of the feature groups (high-frequency and low-frequency) to obtain the corresponding fused feature. Different connection methods can be applied to the other feature group to obtain the corresponding connection feature. The adaptive feature is then determined based on the similarity between the connection feature and the fused feature.

[0040] In some possible implementations, an adaptive selection strategy is used to determine adaptive features based on the high-frequency feature group and the low-frequency feature group, including: performing a first fusion process on the high-frequency feature map of the high-frequency feature group to obtain a first fused feature; obtaining at least two first connection features based on the different connection methods of each low-frequency feature in the low-frequency feature map; and determining the adaptive feature based on a first similarity between the first fused feature and the first connection feature.

[0041] In this embodiment of the disclosure, the first fusion process performed on the high-frequency feature map of the high-frequency feature group to obtain the first fused feature includes: performing connection fusion processing on the high-frequency feature map of the high-frequency feature group in the height direction, width direction and diagonal direction respectively to obtain high-frequency connection features; and performing continuous convolution and residual processing on the high-frequency connection features to obtain the first fused feature.

[0042] By performing connection and fusion processing on the multiple high-frequency feature maps contained in the high-frequency feature group in the height, width, and diagonal directions, multi-directional texture detail information can be aggregated to obtain high-frequency connection features. Since high-frequency features contain more detailed relationships, this embodiment introduces more spatial connection methods in high-frequency feature fusion, further enriching the high-frequency feature information. Specifically, connection processing can be performed on the high-frequency feature maps of the high-frequency feature group in the height, width, and diagonal directions to obtain a first high-frequency connection feature, a second high-frequency connection feature, and a third high-frequency connection feature; channel direction connection processing is then performed on the first, second, and third high-frequency connection features to obtain the high-frequency connection features.

[0043] Having obtained high-frequency connectivity features, at least two convolutional layers can be applied to these features. The residual process is then performed by summing the convolutional features and the high-frequency connectivity features obtained from the convolutional processing, yielding the first fused feature. Based on this configuration, the fusion of high-frequency information can be achieved.

[0044] In other embodiments, embodiments of this disclosure perform a first fusion process on the high-frequency feature group to obtain a first fused feature, including: performing connection fusion processing on the high-frequency feature map of the high-frequency feature group in the height direction, width direction, and diagonal direction respectively to obtain high-frequency connection features; performing attention processing on the high-frequency connection features to obtain high-frequency attention features; and performing feature transformation processing on the high-frequency attention features to obtain the first fused feature.

[0045] The method for obtaining high-frequency connectivity features is as described in the above embodiments and will not be repeated here. After obtaining the high-frequency connectivity features, attention processing can be performed on the high-frequency connectivity features to obtain high-frequency attention features. Specifically, firstly, recombination processing is performed on the high-frequency connectivity features to obtain recombination features; then, three branch processing is performed on the recombination features to obtain first branch features, second branch features, and third branch features; the product of the second branch features and the third branch features is used to obtain the first branch product features; and the product of the first branch features and the first branch product features is used to obtain the high-frequency attention features.

[0046] The reorganization process for the high-frequency connectivity features includes: performing block processing on the high-frequency connectivity features in the width and height directions to obtain multiple block features; and performing connection processing on the block features in the height direction to obtain the reorganized features. The number of blocks can be set according to requirements, such as 2×2, or other numbers; this disclosure does not specifically limit this. This block processing strategy aims to achieve local attention enhancement: compared to global attention, which easily allows the camouflaged region to be submerged by background information, local attention allows the model to focus more on subtle differences within small regions, amplifying detailed textures and making it easier to detect edge changes of camouflaged objects; at the same time, block operations can reduce the computational complexity of the attention mechanism. Subsequently, the high-frequency connectivity features are input into the attention mechanism module for attention processing; during this process, the feature data undergoes dimensional reorganization, capturing short-distance dependencies and local statistical distribution characteristics between features, ultimately obtaining the first fusion feature for subsequent fusion calculations.

[0047] Upon obtaining the reconstructed features, multi-branch processing is performed on them. Within the attention processing unit, this embodiment improves the classic query, key, and value interaction logic by performing multi-level transformations on the features through statistical branches to enhance discriminability. Specifically, after performing reconstruction processing on the input feature stream, this embodiment generates Query, Key, and Value through three parallel branches. Specifically, Query is the first branch, Key is the second branch, and Value is the third branch.

[0048] The first and second branch processing methods respectively include feature centering, feature squarening, channel normalization, and height*width normalization. The purpose of feature centering is to highlight the "subtle differences" between the camouflage region and the background by eliminating the mean background. In this processing, the third branch is not centered to maintain the integrity of its original expression, while the second branch and its related paths are centered, which amplifies feature fluctuations and compresses flat background areas, thereby making subtle textures and edge changes clearer. In this embodiment, a feature square transformation is performed on the centered features. The square operation can achieve a contrast stretching effect of "larger values ​​are larger, smaller values ​​are smaller," so that the model pays more attention to texture boundaries, detail changes, and the outline of the camouflage, enhancing the response of significant features while suppressing weak background features, resulting in stronger recognition contrast. To further improve the robustness of the model in complex backgrounds, this embodiment performs double normalization processing on the channel dimension (C) and spatial dimension (H×W) after feature squarening. Among them, normalization along the channel dimension is used to constrain the activation amplitude of each semantic channel (such as texture, color, and edge channels), suppressing the contributions of different channels to the same order of magnitude, and preventing certain high-energy channels or noise channels from dominating the correlation calculation; normalization along the spatial dimension balances the energy distribution at each location in the image, making the contribution of each pixel location comparable, and effectively avoiding strong texture backgrounds or large areas from masking small camouflaged targets.

[0049] Furthermore, the product of the second branch features and the third branch features is used to obtain the first branch product features; the product of the first branch features and the first branch product features is used to obtain the high-frequency attention features. Given the high-frequency attention features, feature transformation processing is performed on them to obtain the first fused features. The feature transformation includes transforming the feature dimensions; through block processing in the width and height directions, the features are restored to the dimensions of the high-frequency features within the high-frequency feature group, thus obtaining the first fused features.

[0050] Furthermore, at least two first connection features are obtained based on different connection methods of each low-frequency feature in the low-frequency feature map. Specifically, connection processing in the height and width directions can be performed on the low-frequency feature maps in the low-frequency feature group to obtain a first connection feature in the height direction and a first connection feature in the width direction. When different first connection features are obtained, an adaptive feature is determined based on the similarity between the first connection feature and the first fusion feature. The step of determining the adaptive feature based on the first similarity between the first fusion feature and the first connection feature includes: calculating the first similarity between the first fusion feature and each of the first connection feature maps; determining the first connection feature whose first similarity satisfies a condition greater than a first threshold as a selected low-frequency feature map; and determining the adaptive feature based on the first fusion feature and the selected low-frequency feature map.

[0051] In some possible implementations, the first similarity can be calculated using cosine similarity; in other implementations, Euclidean distance or other methods can be used. This disclosure does not specifically limit the method. If the first similarity is greater than a first threshold, the corresponding first connection feature is determined as a low-frequency feature map. Then, the first fusion feature and the selected low-frequency feature map can be used to determine the adaptive feature.

[0052] The adaptive feature determination using the first fused feature and the selected low-frequency feature map includes: performing feature alignment processing on the first fused feature and the selected low-frequency feature map to obtain aligned spatial features; performing channel compression and weighted attention prediction processing on the aligned spatial features to obtain weighted attention for the corresponding spatial features; performing product processing on the weighted attention and the aligned spatial features to obtain weighted spatial features; and performing summation processing on each weighted spatial feature to obtain the adaptive feature.

[0053] Figure 2 This is a block diagram of the adaptive feature extraction model according to an embodiment of the present disclosure. The embodiments of the present disclosure first perform feature alignment processing on the first fused feature and the selected low-frequency feature map. This process ensures that all features can be processed and fused at the same spatial scale, avoiding potential mismatches between different scales. The alignment process includes performing basic convolution processing and channel integration processing on the first fused feature and the selected low-frequency feature map. The basic convolution processing includes convolution, batch normalization, and activation processing (such as ReLU). It is important to note that because the first fused feature and the selected low-frequency feature map have different scales, the convolution parameters for each spatial feature in the basic convolution are different, thus guiding low-level features to focus on local patterns, while high-level features integrate broader contextual information. This asymmetric design ensures that the effective receptive field of each layer matches its semantic role, rather than treating all layers as interchangeable. The channel integration processing uses 1×1 convolution to project the refined features onto a unified embedding space. Through this configuration, the scale of the dominant fused representation is adaptively selected at each spatial location.

[0054] Next, channel compression and weighted attention prediction are performed on the alignment space features. Figure 3 This is a schematic diagram of the channel compression and fusion structure. Channel compression is performed using 1×1 convolutions, which reduces the number of channels in the feature map and effectively extracts important information from each channel; meanwhile, the weighted attention prediction process dynamically predicts weights for each spatial location, reflecting the importance of each spatial location.

[0055] Specifically, for channel compression, since the original channel dimensions differ across layers, the refined features are projected onto a unified embedding space using a 1×1 convolution to obtain the channel-compressed features. For weighted attention prediction, a 1×1 convolution is performed on the channel-compressed features to obtain the weighted attention coefficients.

[0056] After aligning spatial features to a uniform spatial resolution, the most suitable scale is dynamically selected based on specific needs, thereby achieving location-adaptive feature fusion rather than averaging across all scales. For each scale layer, we parameterize the "1×1 compression" and "1×1 convolution" modules as two learnable mappings. Based on the aligned feature maps, the "compression" layer independently applies a 1×1 convolution (including BN and ReLU) to each spatial location.

[0057] In this process, the weighted attention prediction uses a 1×1 convolution to represent the weight strength that should be assigned to the feature scale at position (H,W). This weight is then broadcast and multiplied by the alignment spatial features to produce a scale-weighted response, resulting in weighted spatial features. Subsequently, each weighted spatial feature is summed to obtain adaptive features.

[0058] Using the above method, a set of weights spanning four scales is learned for each spatial location, so that the fused representation is dominated by the most informative scale, while naturally suppressing the less relevant responses.

[0059] In this embodiment, the correlation between high-frequency features and low-frequency structural information is evaluated by calculating a first similarity between the first fused feature and each first connection feature. Based on the magnitude of the similarity, first connection features with high correlation to the low-frequency fused features are adaptively selected and synergistically fused with the low-frequency fused features to obtain adaptive features that simultaneously consider both overall structural information and local detail information. This adaptive feature can serve as the basic input for subsequent camouflaged target detection to improve the accuracy and robustness of target detection.

[0060] In addition, the method of using an adaptive selection strategy to determine adaptive features based on the high-frequency feature group and the low-frequency feature group may further include: performing a second fusion process on the low-frequency feature map of the low-frequency feature group to obtain a second fusion feature; obtaining at least two second connection features based on the different connection methods of each high-frequency feature in the high-frequency feature map; and determining the adaptive feature based on the second similarity between the second fusion feature and the second connection feature.

[0061] Correspondingly, embodiments of this disclosure may also first fuse low-frequency features and then correlate them with high-frequency features to obtain adaptive features. Specifically, the low-frequency feature maps of the low-frequency feature group undergo a second fusion process to obtain a second fused feature, including: performing connection processing in the height and width directions on the low-frequency feature maps in the low-frequency feature group respectively to obtain a first low-frequency connection feature and a second low-frequency connection feature; performing feature encoding processing on the first low-frequency connection feature and the second low-frequency connection feature respectively to obtain a first encoded feature and a second encoded feature; and fusing the first encoded feature and the second encoded feature to obtain the second fused feature.

[0062] This embodiment of the disclosure can first perform connection processing on the low-frequency feature maps within the low-frequency features, respectively, in the height and width directions. Specifically, the low-frequency features include at least two low-frequency feature maps, and these feature maps are connected to effectively fuse the low-frequency feature information. In the height direction, the rows of each low-frequency feature map are first concatenated to form a first low-frequency connection feature; in the width direction, the columns of each low-frequency feature map are further concatenated to obtain a second low-frequency connection feature. Through this connection processing, the low-frequency feature maps can be expanded in spatial dimensions, thereby retaining more image information and enhancing the effect of subsequent processing.

[0063] Subsequently, feature encoding processing can be performed on the first low-frequency connection feature and the second low-frequency connection feature respectively to obtain the first encoded feature and the second encoded feature. The feature encoding processing includes: performing a two-branch process on the input connection feature; wherein the first branch process includes: performing a linear transformation and activation processing on the connection feature; the second branch process includes: performing a linear transformation, feature dimensionality reduction, activation processing, and state space transformation processing on the connection feature; performing a product process and a linear transformation on the two features obtained from the two-branch process to obtain the encoded feature corresponding to the connection feature. The input connection feature is the first low-frequency connection feature and the second low-frequency connection feature. For ease of description, the following explanation uses the input connection feature as an example and will not elaborate further.

[0064] In this embodiment, after completing the connection processing of the low-frequency feature map, the obtained connection features are flattened to convert the feature map into one-dimensional data (B, H×W×C) for subsequent linear transformation and feature encoding. Next, feature encoding processing is performed on the first and second low-frequency connection features. Feature encoding includes performing a two-branch process on each connection feature. In the first branch, a linear transformation is first performed on the first low-frequency connection feature to obtain feature F, and then a nonlinear activation function is applied to this feature. For example, with Sigmoid, the activated feature A is the first encoded feature:

[0065] Where W and b are the weights and biases, respectively, and are learnable parameters.

[0066] Furthermore, in the second branch, the first low-frequency connection feature L is first linearly transformed to obtain the linearly transformed feature F. Then, a feature dimensionality reduction convolution operation is applied, using a convolution kernel size of 1×1, or alternatively 3×3, which is not specifically limited in this disclosure. Next, a nonlinear activation function is used to obtain nonlinear features, followed by a state space transformation to obtain the second encoded feature. The state space transformation process uses an SSM model, and the state equation and output equation used are set by those skilled in the art according to their needs, and are not specifically limited here. Through state space transformation, accurate representation of features in long-distance dimensions can be achieved, thereby improving model performance. Finally, the final state space transformed feature is obtained, which is the second encoded feature.

[0067] The linear transformations of the first and second branches use different weight parameters, which are learned during training. Further, a linear transformation is performed after multiplying the first and second encoded features to obtain the final first branch output features. Similarly, the second low-frequency connection features undergo the same encoding process to obtain the second branch output features. After feature encoding is complete, especially after multiple transformations (e.g., state-space transformation), the one-dimensional first branch output features and the first branch output features need to be restored to dimensional feature maps with the original input image, and then directly concatenated to obtain the final second fused feature.

[0068] This concatenation method preserves the independent information of each encoded feature and combines them to form a richer feature representation. In some optional implementations, the second fused feature can be further optimized, for example, by performing CBR (convolution, batch normalization, and ReLU activation) processing on the second fused feature. The CBR module extracts local features through convolution operations, improves training stability through batch normalization, and provides nonlinear transformation through activation functions. To further enhance feature representation capabilities and avoid the gradient vanishing problem, residual connections are introduced into feature processing. In residual connections, low-frequency features are directly added to the features processed by CBR to obtain the final optimized second fused feature. These steps, through multi-level feature processing, nonlinear transformation, and information fusion, significantly improve the robustness and accuracy of camouflaged target detection.

[0069] Furthermore, embodiments of this disclosure can obtain second connectivity features based on high-frequency feature groups. Connectivity processing can be performed on the high-frequency feature maps of the high-frequency feature groups in the height, width, and diagonal directions respectively to obtain the corresponding second connectivity features. After obtaining the second fusion feature, an adaptive feature is further determined based on the second fusion feature and the second connectivity feature. Determining the adaptive feature based on the second similarity between the second fusion feature and the second connectivity feature includes: calculating the second similarity between the second fusion feature and each of the second connectivity feature maps; determining the second connectivity feature whose second similarity satisfies a condition greater than a first threshold as a selected high-frequency feature map; and determining the adaptive feature based on the second fusion feature and the high-frequency and low-frequency feature maps.

[0070] In some possible implementations, the second similarity can be calculated using cosine similarity; in other embodiments, Euclidean distance or other methods can be used. This disclosure does not specifically limit this approach. If the second similarity is greater than a second threshold, the corresponding second connectivity feature is determined as a high-frequency feature map. Then, the second fusion feature and the selected high-frequency feature map can be used to determine adaptive features.

[0071] The adaptive feature determination using the second fusion feature and the selected high-frequency feature map includes: performing feature alignment processing on the second fusion feature and the selected high-frequency feature map to obtain aligned spatial features; performing channel compression and weighted attention prediction processing on the aligned spatial features to obtain weighted attention for the corresponding spatial features; performing product processing on the weighted attention and the aligned spatial features to obtain weighted spatial features; and performing summation processing on each weighted spatial feature to obtain the adaptive feature.

[0072] In this embodiment of the disclosure, the process of obtaining adaptive features can be described with reference to the above embodiments, and will not be repeated here.

[0073] Given adaptive features, these features can be used to detect disguised target objects in the input image. Detecting disguised target objects in the input image using adaptive features includes: directly performing activation processing on the adaptive features to obtain the location information of the disguised target object, wherein the positions of the activated features with feature values ​​greater than 0.5 are the locations of the disguised target object. Alternatively, the location of the disguised target object can be obtained by performing convolution processing on the adaptive features followed by activation processing. This disclosure does not specifically limit this approach.

[0074] In some preferred embodiments, detecting a disguised target object in the input image using the adaptive features includes: performing basic convolution processing on the adaptive features to obtain basic features; performing at least one joint convolution and residual processing on the basic features to obtain target features; and determining the location of the disguised target object based on the target features.

[0075] Figure 4 This is a flowchart illustrating the joint convolution and residual processing of an embodiment of this disclosure. The adaptive features are subjected to a 1x1 convolution to obtain basic convolutional features. These basic convolutional features are then subjected to two joint convolutions and residual processing. The joint convolutions include depth-direction convolutions and planar convolutions in the height and width directions, yielding Y1 and Y2 respectively. Residual processing is then performed on the obtained joint convolutional features by summing the input and output of the joint convolutions to obtain the corresponding residual features. The features obtained after the second residual processing are the target features. After layer normalization, the target features are summed with the basic convolutional features to obtain the final output features. The target location is determined using the positions in the output features where the threshold is greater than 0.5.

[0076] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0077] In addition, this disclosure also provides a disguised object detection device, electronic device, computer-readable storage medium, and program, all of which can be used to implement any of the disguised object detection methods provided in this disclosure. The corresponding technical solutions and descriptions are described in the corresponding records in the method section and will not be repeated here.

[0078] The disguised object detection device includes: a frequency feature extraction module, used to perform frequency feature extraction processing on the input image to obtain a high-frequency feature group and a low-frequency feature group, wherein the high-frequency feature group includes at least two high-frequency feature maps and the low-frequency feature group includes at least two low-frequency feature maps; an adaptive module, which uses an adaptive selection strategy to determine adaptive features based on the high-frequency feature group and the low-frequency feature group; and a detection module, which uses the adaptive features to detect disguised target objects in the input image.

[0079] In some embodiments, the apparatus provided in this disclosure may have functions or include modules that can be used to perform the methods described in the above method embodiments. Specific implementations can be referred to the descriptions in the above method embodiments, and for brevity, will not be repeated here. This disclosure also proposes a computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the above methods. The computer-readable storage medium may be a non-volatile computer-readable storage medium.

[0080] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A method for detecting camouflaged targets using adaptive feature selection, characterized in that, include: Frequency feature extraction processing is performed on the input image to obtain a high-frequency feature group and a low-frequency feature group, wherein the high-frequency feature group includes at least two high-frequency feature maps and the low-frequency feature group includes at least two low-frequency feature maps; An adaptive selection strategy is used to determine adaptive features based on the high-frequency feature group and the low-frequency feature group; The adaptive features are used to detect disguised target objects in the input image.

2. The method according to claim 1, characterized in that, The process of performing frequency feature extraction on the input image to obtain high-frequency feature groups and low-frequency feature groups includes: Perform a Fourier transform or wavelet transform on the input image to obtain the high-frequency feature group and the low-frequency feature group.

3. The method according to claim 1, characterized in that, Using an adaptive selection strategy, adaptive features are determined based on the high-frequency feature group and the low-frequency feature group, including: Perform a first fusion process on the high-frequency feature map of the high-frequency feature group to obtain a first fused feature; At least two first connection features are obtained based on the different connection methods of each low-frequency feature in the low-frequency feature map; The adaptive feature is determined based on the first similarity between the first fusion feature and the first connection feature; Alternatively, the method of using an adaptive selection strategy to determine adaptive features based on the high-frequency feature group and the low-frequency feature group includes: Perform a second fusion process on the low-frequency feature map of the low-frequency feature group to obtain a second fused feature; At least two second connection features are obtained based on the different connection methods of each high-frequency feature in the high-frequency feature map; The adaptive feature is determined based on the second similarity between the second fusion feature and the second connection feature.

4. The method according to claim 3, characterized in that, The first fusion process is performed on the high-frequency feature map of the high-frequency feature group to obtain the first fused feature, including: The high-frequency feature maps of the high-frequency feature group are subjected to connection fusion processing in the height direction, width direction, and diagonal direction, respectively, to obtain high-frequency connection features; The high-frequency connection features are subjected to continuous convolution and residual processing to obtain the first fused feature.

5. The method according to claim 3, characterized in that, Determining the adaptive feature based on the first similarity between the first fused feature and the first connection feature includes: Calculate the first similarity between the first fused feature and each of the first connection feature maps; The first connection feature whose first similarity satisfies the condition of being greater than the first threshold is determined as the selected low-frequency feature map; Adaptive features are determined based on the first fused features and the selected low-frequency feature map.

6. The method according to claim 3, characterized in that, The step of determining the adaptive feature based on the second similarity between the second fusion feature and the second connection feature includes: Calculate the second similarity between the second fusion feature and each of the second connection feature maps; The second connection feature whose second similarity satisfies the condition of being greater than the first threshold is determined as the selected high-frequency feature map; Adaptive features are determined based on the second fusion feature and the high-frequency low-frequency feature map.

7. The method according to claim 1, characterized in that, The step of detecting disguised target objects in the input image using the adaptive features includes: Perform basic convolution processing on the adaptive features to obtain basic features; Perform at least one joint convolution and residual processing on the basic features to obtain the target features; The location of the disguised target object is determined based on the target features.

8. A camouflaged target detection device, characterized in that, include: The frequency feature extraction module is used to perform frequency feature extraction processing on the input image to obtain a high-frequency feature group and a low-frequency feature group, wherein the high-frequency feature group includes at least two high-frequency feature maps and the low-frequency feature group includes at least two low-frequency feature maps. The adaptive module uses an adaptive selection strategy to determine adaptive features based on the high-frequency feature group and the low-frequency feature group; The detection module uses the adaptive features to detect disguised target objects in the input image.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images