Optical remote sensing image salient target detection method and related device

Abstract

The invention discloses an optical remote sensing image salient target detection method and a related device. The method comprises the following steps: inputting a to-be-detected optical remote sensing image into a feature coding module to extract coding features of different levels; the coding features of different levels are input into an edge extraction module for edge extraction operation to obtain a salient target edge graph; and inputting the coding features of different levels and the salient target edge graph into a label generation module to carry out salient target detection operation to obtain a salient target graph. According to the invention, the feature coding module is used for learning features with higher discriminability so as to improve the saliency detectability, and the edge extraction module is used for effectively utilizing edge information so as to enhance the feature discrimination ability; and the problems of edge and feature proportional relation and information redundancy are effectively solved through the label generation module.

Description

Technical Field

[0001] The present invention relates to the field of computer vision processing technology, and in particular to a method for detecting salient targets in optical remote sensing images and related devices. Background Art

[0002] Salient object detection is a key task in computer vision, which aims to detect the most visually distinctive object from a given image. In recent years, with the development of deep neural networks, salient object detection in natural scene images has become relatively mature, but salient object detection in optical remote sensing images remains to be explored. Salient object detection in optical remote sensing images has strong practical value. This task can be applied to various fields of computer vision to play a preprocessing role, such as object segmentation, visual tracking, image retrieval, cropping, image quality assessment, etc.

[0003] In recent years, many technologies have been introduced into optical remote sensing image salient object detection, such as scale fusion, edge guidance, complementary loss, etc. For example, LVNet uses convolutional networks, feature pyramids and specific modules to suppress background and highlight targets; DAFNet uses global context-aware attention (GCA) to capture semantic contextual relationships; ERPNet, composed of an encoder and two parallel decoders, uses edge information to calibrate the decoding process. However, existing optical remote sensing image salient object detection methods have defects in dealing with details and edge blur caused by long imaging distances, complex backgrounds, and large changes in target scales, resulting in unsatisfactory detection results. Summary of the invention

[0004] The present invention provides a method and related device for detecting salient targets in optical remote sensing images, which are used to solve the defects of existing methods for detecting salient targets in optical remote sensing images in processing the details and edges blurred, the background complex, the large change of the target scale caused by the long imaging distance, etc., improve the effect of detecting salient targets in remote sensing images, and make it more advantageous when processing complex backgrounds and multi-scale objects.

[0005] The present invention provides a method for detecting salient targets in optical remote sensing images, the method comprising:

[0006] Input the optical remote sensing image to be detected into the feature encoding module to extract the encoding features at different levels;

[0007] Inputting the encoding features of different levels into an edge extraction module to perform edge extraction operations to obtain a salient object edge map;

[0008] The encoding features at different levels and the salient object edge map are input into a label generation module to perform a salient object detection operation to obtain a salient object map.

[0009] Furthermore, the feature encoding module is composed of five shared encoders; each of the shared encoders is composed of a basic convolution submodule and a CASA submodule;

[0010] Among them, the basic convolution submodule is constructed based on the ResNet-34 backbone network or the VGG-16 backbone network; the CASA submodule includes a channel attention structure and a spatial attention structure, which are used to extract content information and location information respectively.

[0011] Furthermore, the calculation process of the CASA submodule is as follows:

[0012]

[0013] Where: represents the encoding feature of the i-th level, where ; represents the spatial features obtained by spatial attention processing through the i-th level spatial attention structure, Represents the shallow features output by the basic convolutional submodule corresponding to the i-th level; represents the addition of elements, Represents element-wise multiplication; represents the channel features obtained by channel attention processing through the i-th level channel attention structure, represents the spatial weight of the spatial attention structure at the i-th level; Represents the channel weight of the i-th level channel attention structure.

[0014] Furthermore, the edge extraction module is composed of a significant convolution head and five serially connected edge decoders;

[0015] In the edge extraction module, five serially connected edge decoders are used to perform edge extraction operations on coding features at different levels to generate edge features at the first level; and the salient convolution head is used to perform convolution operations on the edge features at the first level to obtain a salient target edge map.

[0016] Furthermore, the edge extraction operation process of the edge extraction module is as follows:

[0017]

[0018] Where: represents the feature map of the i-th level, Indicates series connection, represents the encoding features of the fifth level, Represented by the fifth level edge decoder Encoding features for the fifth level The fifth level edge features are obtained by performing edge feature extraction operation;

[0019]

[0020] Where: represents the feature map of the i-th level, represents the encoding features of the i-th level, Represents the edge decoder at level i The encoding features of the i-th level and the feature map of level i+1 Perform edge feature extraction to obtain the edge features of the i-th level;

[0021]

[0022] Where: Represented by the first level edge decoder Encoding features of the first level and the second-level feature map Perform edge feature extraction to obtain the first level of edge features;

[0023]

[0024] Where: represents the salient object edge map, Represents a convolution operation.

[0025] Furthermore, the label generation module consists of five decoders with bottleneck structures, five edge embedding structures and a significant convolution head;

[0026] The salient object detection operation process of the label generation module is as follows:

[0027]

[0028] Where: Represents the scale characteristics of the fifth level, The edge capture operation representing the edge embedding structure, represents the decoding operation of the decoder with a bottleneck structure at the fifth level;

[0029]

[0030] Where: represents the scale feature of the i-th level, represents the decoding operation of the decoder with bottleneck structure at the i-th level;

[0031]

[0032] Where: represents a salient target map; Represents the scale characteristics of the first level.

[0033] Further, the edge embedding structure includes a residual connection structure and a pyramid attention structure;

[0034] In the residual connection structure, a smoothing operation is performed on the salient target edge map, and the smoothing operation result is downsampled to obtain downsampled features; the downsampled features are residually connected with decoded features obtained by decoding operations of decoders with bottleneck structures at different levels to obtain enhanced features at different levels; wherein the pyramid attention structure is used to extract scale information of the enhanced features at different levels;

[0035] The data processing process of the edge embedding structure is as follows:

[0036]

[0037] Where: represents the enhanced features of the i-th level, Indicates smoothing operation, represents the downsampling operation, represents the residual connection, where ;

[0038]

[0039] Where: represents the scale feature of the i-th level, Representing the pyramid scale operation of the pyramid attention structure.

[0040] The present invention also provides a system for detecting salient objects in optical remote sensing images, the system comprising:

[0041] The encoding unit is used to input the optical remote sensing image to be detected into the feature encoding module to extract the encoding features of different levels;

[0042] An edge extraction unit, used for inputting the coding features of different levels into an edge extraction module to perform edge extraction operations to obtain a salient object edge map;

[0043] The label generation unit is used to input the encoding features of different levels and the salient object edge map into the label generation module to perform a salient object detection operation to obtain a salient object map.

[0044] The present invention also provides a computer device, including a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the optical remote sensing image salient target detection method as described above.

[0045] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the optical remote sensing image salient target detection method as described above.

[0046] It can be seen from the above technical solutions that the present invention has the following advantages:

[0047] The present invention provides a method for detecting a significant target in an optical remote sensing image and a related device. The method comprises: inputting an optical remote sensing image to be detected into a feature encoding module to extract encoding features at different levels; inputting the encoding features at different levels into an edge extraction module to perform an edge extraction operation to obtain a significant target edge map; inputting the encoding features at different levels and the significant target edge map into a label generation module to perform a significant target detection operation to obtain a significant target map.

[0048] In the present invention, the feature encoding module introduces channel information and spatial information to make the learned features more discriminative; the edge extraction module effectively uses edge information to improve the ability to identify significant targets; the label generation module reduces information redundancy, aligns edges and features, and improves model performance and efficiency by introducing bottleneck structures and edge embedding structures. The present invention solves the defects of existing optical remote sensing image significant target detection methods in dealing with details and edge blurring caused by long imaging distances, complex backgrounds, and large changes in target scales, improves the effect of remote sensing image significant target detection, and has more advantages in dealing with complex backgrounds and multi-scale objects. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0050] Figure 1 A flowchart of a method for detecting salient objects in optical remote sensing images provided by an embodiment of the present invention;

[0051] Figure 2 A diagram of the CSENet network structure provided by an embodiment of the present invention;

[0052] Figure 3 A structural block diagram of a shared encoder provided by an embodiment of the present invention;

[0053] Figure 4 A schematic diagram of a bottleneck structure provided by an embodiment of the present invention;

[0054] Figure 5A schematic diagram of an edge embedding structure provided by an embodiment of the present invention;

[0055] Figure 6 A schematic diagram of a pyramid attention structure provided by an embodiment of the present invention;

[0056] Figure 7 A structural block diagram of a salient object detection system for optical remote sensing images provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The embodiments of the present invention provide a method and related devices for detecting salient targets in optical remote sensing images, which are used to solve the defects of existing methods for detecting salient targets in optical remote sensing images in processing the details and edges blurred, the background complex, the large change of the target scale caused by the long imaging distance, etc., and improve the effect of detecting salient targets in remote sensing images, so that it has more advantages in processing complex backgrounds and multi-scale objects.

[0058] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] See also Figure 1 and Figure 2 The first aspect of the present invention provides a method for detecting salient targets in optical remote sensing images, the method comprising:

[0060] Step 101: input the optical remote sensing image to be detected into a feature coding module to extract coding features at different levels.

[0061] Please note that Figure 2 , the present invention is applied to the CSENet network structure; Figure 3 As shown in the figure, the feature encoding module consists of five shared encoders; the optical remote sensing image to be detected can be a 256×256 RGB image, which is then input into the five shared encoders in sequence according to different levels to generate encoding features of different levels. .

[0062] Each shared encoder consists of a basic convolution submodule (Base-conv) and a CASA submodule; the basic convolution submodule can be built using popular neural networks as the backbone network, such as the ResNet-34 backbone or the VGG-16 backbone; the CASA submodule includes a channel attention structure (CA) and a spatial attention structure (SA) for extracting content information and location information, respectively.

[0063] 1) For the basic convolution submodule:

[0064] For the ResNet-34 backbone network, this implementation further modifies the structure of the ResNet-34 backbone network: Base-conv1 at the first level consists of a convolutional layer, a batch normalization layer, and a ReLU activation function. The convolutional layer has 3 input channels (IC), 64 output channels (OC), a 3×3 convolution kernel size (K), and 1 supplement (P); then the first layer (layer1) of ResNet34 is superimposed, and the output size is 256×256. At the second to fourth levels, namely Base-conv2, Base-conv3, and Base-conv4, the second, third, and fourth layers of the ResNet-34 model are directly used, respectively, with output sizes of 128×128, 64×64, and 32×32, respectively. The last Base-conv5 consists of a max pooling layer (kernel size = 2×2 and stride = 2) and three basic block modules; the max pooling layer is used for downsampling to reduce the output size to 16×16; the basic block consists of two 3×3 convolutional layers with 512 input channels and 512 output channels.

[0065] For the VGG-16 backbone network, this implementation removes the fully connected layer and makes some other structural adjustments. Specifically: First, Base-conv1 consists of two convolution operations, where the first convolution layer converts the 3-channel input image into 64 channels (kernel size = 3×3, stride = 1 and padding = 1), followed by a batch normalization layer and a ReLU activation function; the input and output channels of the second convolution layer are also 64, and the rest of the settings are the same as the first convolution layer; after this, batch normalization and ReLU activation are performed again. Then, Base-conv2 consists of a maximum pooling layer (kernel size = 2×2 and stride = 2) and two convolution operations. In the third to fifth layers, namely Base-conv3, Base-conv4 and Base-conv5, they are all composed of a maximum pooling layer (kernel size = 2×2 and stride = 2) and three convolution operations; these four convolution modules have a total of 11 convolutional layers (each convolutional layer is followed by a batch normalization layer and a ReLU activation function); except for the number of input and output channels, the settings of kernel size, stride and padding are the same as the first convolutional layer in Base-conv1.

[0066] It is understandable that the backbone network of the shared encoder can replace ResNet-34 and VGG-16 with other network structures with similar feature extraction capabilities, such as ResNet-50, etc., and be appropriately adjusted to adapt to the module connection and functional implementation of the present invention.

[0067] 2) For CASA submodule:

[0068] See also Figure 3 , the CASA submodule uses the shallow features output by the basic convolution submodule As input, the features enter the channel attention structure (CA) and the spatial attention structure (SA) in turn. Specifically: the shallow features Channel weights with channel attention structure Multiply element by element to obtain channel features ; Then, the spatial weight With channel characteristics Multiply element-wise to get spatial features ; Finally, and initial input Add element by element to get the final output ; Here The encoding feature , the above process is expressed by the formula as follows:

[0069]

[0070] Where: represents the addition of elements, Represents element-wise multiplication.

[0071] It should be noted that the channel attention structure (CA) consists of an adaptive max pooling layer and two convolutional layers. The first convolutional layer is followed by a ReLU activation function, and the second convolutional layer is followed by a Sigmoid activation function; CA mainly extracts content information by determining the importance of different channels. Usually, an image will generate a series of feature maps through a convolutional network. Each channel of the feature map contains different levels of image features, such as edges, textures, or colors. The channel attention mechanism weights the channels, allowing the model to emphasize or suppress information from specific channels.

[0072] In the spatial attention structure (SA), the mean and maximum values ​​of the input feature maps in the channel dimension are first calculated, and then the two feature maps are concatenated to perform a convolution operation using a Sigmoid activation function. SA mainly extracts location information that is critical for positioning, and it generates a spatial attention map by assigning weights to different spatial locations, allowing the model to focus on specific areas and ignore irrelevant background information.

[0073] Step 102: Input the coded features at different levels into an edge extraction module to perform edge extraction operations to obtain a salient object edge map.

[0074] It should be noted that the edge extraction module consists of a significant convolution head and five cascaded edge decoders. Composition, through five cascaded edge decoders to encode features at different levels Perform edge extraction operations; during the processing, the edge extraction module processes in order from deep to shallow, that is, from the deepest level edge decoder To the shallowest edge decoder , get the first level edge features ; Finally, through the significant convolution head For the first level edge features Perform convolution operation to obtain the salient target edge map .

[0075] These five decoders are composed of three convolutional layers, each of which is followed by a batch normalization layer and a ReLU activation function. The parameter settings of their convolutional layers are shown in Table 1.

[0076] Table 1 Parameter settings of the convolutional layer in the edge decoder of the edge extraction module

[0077]

[0078] in, The fifth level encoding features As input, output the fifth level edge features ; Then connect and , and then an upsampling operation is performed to obtain a coding feature similar to the fourth level The fifth-level feature map of the same size , and and Series as The input is as follows:

[0079]

[0080] Where: Indicates series connection.

[0081] remove In addition, the other four decoders take the encoding features of the same level and the edge features of the deeper level as input, as shown below:

[0082]

[0083] Where: represents the feature map of the i-th level, Represents the edge decoder at level i This fusion method is expected to improve the accuracy and edge detection.

[0084] Finally, through the significant convolution head Get edge map , as shown below:

[0085]

[0086] in, It consists of four convolutional layers, all of which do not require batch normalization and the fourth convolutional layer has no activation function, while the other three convolutional layers are connected to the ReLU activation function; The specific parameter settings are shown in Table 2, where x in Table 2 is the number of input channels.

[0087] Table 2 Parameters of the convolutional layer

[0088]

[0089] Step 103: Input the encoding features of different levels and the salient object edge map into the label generation module to perform a salient object detection operation to obtain a salient object map.

[0090] Among them, the label generation module is mainly composed of five decoders with bottleneck structure , five edge embedding structures (EES) with pyramid attention structures (PA) and a salient convolutional head. In the process of salient object detection, the label generation module is also processed in a deep-to-shallow order, starting from the decoder Start with the first level decoder and move to the first level decoder. The decoded features at different levels generated by the decoder After that, decode the features First, the residual is connected to the output of the edge extraction module Here, It plays an auxiliary supervisory role and guides the edge embedding structure to better understand the shape and boundary of the object. Then, the connected features are passed through the pyramid attention module to further highlight the area containing the salient object and obtain the scale feature. Then, the scale feature is fed into the previous layer decoder together with the encoding feature of the previous layer to generate a new scale feature, and finally the first-level scale feature is obtained. . Using the saliency convolutional head For the first level scale features Perform convolution operation to generate a salient target map .

[0091] Specifically, first, the fifth-level decoder with a bottleneck structure Encoding features for the fifth level Perform decoding operation and output the fifth-level decoding features ; Then, the decoded features are fused through the edge embedding structure (EES) and the edge map of the salient object , generate the fifth level scale features , as shown below:

[0092]

[0093] in, An edge capture operation representing an edge embedding structure;

[0094] The other four decoders in the label generation module are encoded with the same level of features. and scale features from deeper levels As input, that is:

[0095]

[0096] Where: represents the scale feature of the i-th level, Represents the decoding operation of the decoder with a bottleneck structure at the i-th level.

[0097] Finally, through Perform convolution operation to generate salient target map ,Right now:

[0098]

[0099] It should be noted that the label generation module The structure is similar to that in the edge extraction module , but the label generation module The number of input channels is 128.

[0100] Structurally, The bottleneck structure has three convolutional layers. Figure 4 As shown in the figure, each convolutional layer is followed by a batch normalization layer and a ReLU activation function, and the number of channels from its input to output is 512, 128, 128, and 512, respectively. This not only reduces the number of parameters and computational complexity, but also filters out redundant features. The remaining four decoders are composed of upsampling, feature connection, channel dimension reduction, and bottleneck structure filtering.

[0101] In the edge map of the salient object With decoding features It plays a vital role in data fusion, which helps the network capture better boundaries of objects in the area of ​​interest. Figure 5 As shown, It mainly consists of a residual connection structure (RC) and a pyramid attention structure (PA).

[0102] In the residual connection structure (RC), the edge map of the salient object conduct Function smoothing, downsampling ( ) The processing function smoothes the result of the operation and downsamples it to the same size as the corresponding decoded feature to obtain the downsampled feature; using the downsampled feature as auxiliary information, the residual connection is used to connect the downsampled feature and the decoded feature obtained by the decoding operation of the decoder with the bottleneck structure at different levels to obtain the enhanced features at different levels. , as shown below:

[0103]

[0104] Where: .

[0105] To effectively handle objects or regions of varying scales, features are enhanced Feed into the PA structure to obtain scale characteristics ,Right now:

[0106]

[0107] in, Representing the pyramid scale operation of the pyramid attention structure.

[0108] See also Figure 6 , the pyramid attention structure is set with single-scale branches and multi-scale branches, then the output of the pyramid attention structure Can be single-scale information or multi-scale information , depending on the specific selections in the structure, as follows:

[0109]

[0110] The single-scale branch will Introduce the spatial attention structure (SA) to obtain the spatial attention weight , and then obtain the single-scale output through element-by-element multiplication and element-by-element addition, that is:

[0111]

[0112] Similarly, the multi-scale branch also adopts as input, but using the cascaded pyramid attention ( ) to obtain information at various scales, such as Figure 6The branch is divided into three streams, and each stream operates in a similar way to the previous single-scale operation except for the size of the input and output. In addition, the output of the previous stream is also concatenated with the subsequent stream (C) after upsampling (US).

[0113] Specifically, the first stream uses the four-fold downsampled (DS4) features to generate new features Capture large objects or areas, i.e.:

[0114]

[0115] in .

[0116] Next, the second stream integrates the features downsampled twice (DS2) with the output of the previous stream, namely:

[0117]

[0118] in,

[0119]

[0120] The last stream further integrates the features of the original size to accommodate objects or regions with different scales, namely:

[0121]

[0122] in,

[0123]

[0124] It is understandable that by changing the number of streams and sampling multiples of the multi-scale branch, we can explore a better multi-scale feature extraction method while maintaining its ability to integrate features of different scales. In addition, in the decoder of the edge extraction module and the label generation module, the parameter settings of the convolution layer (such as the size and number of convolution kernels) and the activation function type can be adjusted. For example, try to use different nonlinear activation functions instead of ReLU to observe the impact on model performance.

[0125] The present invention provides a method for detecting salient objects in optical remote sensing images, which has the following advantages:

[0126] 1. Design a model that can effectively utilize edge information and enhance feature recognition capabilities to solve the problem of salient target detection in optical remote sensing images caused by imaging distance and accurately detect salient targets.

[0127] 2. Through the attention mechanism and specific module design, the model's ability to generalize across scenes and suppress irrelevant targets is improved to cope with complex backgrounds; at the same time, modules such as pyramid attention are used to effectively detect multi-scale objects and solve the problem of scale changes.

[0128] 3. Improve the encoder and decoder structures to retain position information, process edge and feature ratio relationships, and reduce information redundancy. In the encoder, CASA is used to enable the encoder to learn more discriminative features, which can better retain position information and extract effective features than existing encoders, thereby more accurately identifying and locating targets in salient target detection in optical remote sensing images. In decoding, the bottleneck structure of the label generation module and the PA unit effectively handle the edge and feature ratio relationships and information redundancy problems, improve model performance and efficiency, and have more advantages in processing complex backgrounds and multi-scale objects, which is a significant improvement over existing decoders.

[0129] See also Figure 7 The second aspect of the present invention provides a system for detecting salient objects in optical remote sensing images, the system comprising:

[0130] The encoding unit 201 is used to input the optical remote sensing image to be detected into the feature encoding module to extract the encoding features of different levels;

[0131] The edge extraction unit 202 is used to input the coding features of different levels into the edge extraction module to perform edge extraction operations to obtain a salient object edge map;

[0132] The label generation unit 203 is used to input the encoding features of different levels and the salient object edge map into the label generation module to perform a salient object detection operation to obtain a salient object map.

[0133] A third aspect of the present invention provides a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the optical remote sensing image salient target detection method as described above.

[0134] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the optical remote sensing image salient target detection method as described above.

[0135] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0136] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0137] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0138] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0139] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0140] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting salient objects in optical remote sensing images, characterized in that: The method comprises: Input the optical remote sensing image to be detected into the feature encoding module to extract the encoding features at different levels; Inputting the encoding features of different levels into an edge extraction module to perform edge extraction operations to obtain a salient object edge map; The encoding features at different levels and the salient object edge map are input into a label generation module to perform a salient object detection operation to obtain a salient object map.

2. The optical remote sensing image salient object detection method according to claim 1, characterized in that: The feature encoding module is composed of five shared encoders; each of the shared encoders is composed of a basic convolution submodule and a CASA submodule; Among them, the basic convolution submodule is constructed based on the ResNet-34 backbone network or the VGG-16 backbone network; the CASA submodule includes a channel attention structure and a spatial attention structure, which are used to extract content information and location information respectively.

3. The optical remote sensing image salient object detection method according to claim 2, characterized in that: The calculation process of the CASA submodule is as follows: Where: represents the encoding feature of the i-th level, where ; represents the spatial features obtained by spatial attention processing through the i-th level spatial attention structure, Represents the shallow features output by the basic convolutional submodule corresponding to the i-th level; represents the addition of elements, Represents element-wise multiplication; represents the channel features obtained by channel attention processing through the i-th level channel attention structure, represents the spatial weight of the spatial attention structure at the i-th level; Represents the channel weight of the i-th level channel attention structure.

4. The optical remote sensing image salient object detection method according to claim 1, characterized in that: The edge extraction module consists of a significant convolution head and five edge decoders connected in series; In the edge extraction module, five serially connected edge decoders are used to perform edge extraction operations on the coding features of different levels to generate edge features of the first level; and the salient convolution head is used to perform convolution operations on the edge features of the first level to obtain a salient target edge map.

5. The optical remote sensing image salient object detection method according to claim 4, characterized in that: The edge extraction operation process of the edge extraction module is as follows: Where: represents the feature map of the i-th level, Indicates series connection, represents the encoding features of the fifth level, Represented by the fifth level edge decoder Encoding features for the fifth level The fifth level edge features are obtained by performing edge feature extraction operation; Where: represents the feature map of the i-th level, represents the encoding features of the i-th level, Represents the edge decoder at level i The encoding features of the i-th level and the feature map of level i+1 Perform edge feature extraction to obtain the edge features of the i-th level; Where: Represented by the first level edge decoder Encoding features of the first level and the feature map of the second level Perform edge feature extraction to obtain the first level of edge features; Where: represents the salient object edge map, Represents a convolution operation.

6. The optical remote sensing image salient object detection method according to claim 1, characterized in that: The label generation module consists of five decoders with bottleneck structures, five edge embedding structures and a significant convolution head; The salient object detection operation process of the label generation module is as follows: Where: Represents the scale characteristics of the fifth level, The edge capture operation representing the edge embedding structure, represents the decoding operation of the decoder with a bottleneck structure at the fifth level; Where: represents the scale feature of the i-th level, represents the decoding operation of the decoder with bottleneck structure at the i-th level; Where: represents a salient target map; Represents the scale characteristics of the first level.

7. The optical remote sensing image salient object detection method according to claim 6, characterized in that: The edge embedding structure includes a residual connection structure and a pyramid attention structure; In the residual connection structure, a smoothing operation is performed on the salient target edge map, and the smoothing operation result is downsampled to obtain downsampled features; the downsampled features are residually connected with decoded features obtained by decoding operations of decoders with bottleneck structures at different levels to obtain enhanced features at different levels; wherein the pyramid attention structure is used to extract scale information of the enhanced features at different levels; The data processing process of the edge embedding structure is as follows: Where: represents the enhanced features of the i-th level, represents smoothing operation, represents the downsampling operation, represents the residual connection, where ; Where: represents the scale feature of the i-th level, Representing the pyramid scale operation of the pyramid attention structure.

8. A system for detecting salient objects in optical remote sensing images, characterized in that: The system comprises: The encoding unit is used to input the optical remote sensing image to be detected into the feature encoding module to extract the encoding features of different levels; An edge extraction unit, used for inputting the coding features of different levels into an edge extraction module to perform edge extraction operations to obtain a salient object edge map; The label generation unit is used to input the encoding features of different levels and the salient target edge map into the label generation module to perform a salient target detection operation to obtain a salient target map.

9. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the optical remote sensing image salient object detection method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the optical remote sensing image salient object detection method as described in any one of claims 1 to 7 are implemented.

Images