Remote sensing image road segmentation method, system, equipment and medium based on hybrid scale feature extraction and dynamic intelligent core feature enhancement

Abstract

The present invention relates to the field of remote sensing image processing technology, and discloses a method, system, device, and medium for remote sensing image road segmentation based on mixed-scale feature extraction and dynamic intelligent core feature enhancement. The method comprises extracting road information at different scales from a remote sensing image using a mixed-scale feature extraction unit to obtain a remote sensing image feature map; extracting multi-scale features and global context information from the remote sensing image feature map using a dynamic intelligent core feature enhancement unit to obtain a feature enhancement map; and performing road segmentation on the feature enhancement map based on Mamba enhancement and a diffusion model, outputting the road segmentation result. This method not only improves the accuracy of road segmentation but also effectively addresses the problems of insufficient multi-scale information processing and decreased segmentation accuracy in complex backgrounds, existing in the prior art. This provides a more efficient and accurate solution for intelligent transportation systems.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image processing, and more particularly to a remote sensing image road segmentation method, system, device and medium based on hybrid scale feature extraction and dynamic intelligent core feature enhancement. Background Art

[0002] With the acceleration of urbanization, the development of intelligent transportation systems has become a key priority for improving traffic efficiency and ensuring safety. Road segmentation, a fundamental task within these systems, directly impacts the accuracy of subsequent technologies such as road recognition and vehicle navigation. Accurately segmenting roads in complex environments, particularly those with blurred details or complex backgrounds, presents a significant challenge.

[0003] Currently, road segmentation methods based on deep learning have achieved certain results. Traditional road segmentation methods mainly rely on image processing techniques such as edge detection and color segmentation. These methods can achieve good results when the background is simple or the road is clear. However, in complex backgrounds, especially in harsh environments such as at night and in rainy and snowy weather, the performance of these methods is greatly reduced. To meet this challenge, methods based on deep neural networks (DNNs) have been gradually proposed and applied to road segmentation tasks. Convolutional neural networks (CNNs), as one of the important architectures, have performed well in many visual tasks due to their powerful feature extraction capabilities. However, existing deep learning methods generally have two problems: first, they cannot effectively process road information at different scales; second, in scenes with complex backgrounds or blurred road details, the segmentation accuracy of the model is significantly affected.

[0004] To address these two issues, some existing approaches attempt to enhance model robustness through multi-scale feature fusion. For example, some U-Net-based networks gradually fuse features from different levels to improve segmentation. However, these approaches still fall short in practical applications, particularly for accurately capturing road details. Other approaches employ attention mechanisms to enhance the representation of important features, but these methods often rely on large-scale data training and still perform poorly when dealing with complex backgrounds.

[0005] Therefore, how to provide a remote sensing image road segmentation method based on mixed-scale feature extraction and dynamic intelligent core feature enhancement is a problem that those skilled in the art urgently need to solve. Summary of the Invention

[0006] In light of this, the present invention provides a method, system, device, and medium for remote sensing image road segmentation based on hybrid-scale feature extraction and dynamic intelligent core feature enhancement. By improving the feature extraction and enhancement modules, the method aims to enhance the accuracy and robustness of road segmentation in remote sensing images, particularly for road recognition in complex backgrounds, while reducing the computational resource consumption of model training and deployment. Through multi-scale feature extraction and adaptive dynamic enhancement, the present invention effectively captures road information in images and is highly adaptable to diverse environments.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A remote sensing image road segmentation method based on hybrid scale feature extraction and dynamic intelligent core feature enhancement includes:

[0009] Extracting road information of different scales from remote sensing images based on a mixed-scale feature extraction unit to obtain a remote sensing image feature map;

[0010] The dynamic intelligent core feature enhancement unit is used to extract multi-scale features and global context information from the remote sensing image feature map to obtain a feature enhancement map.

[0011] Based on the Mamba enhancement and Diffusion model, the feature enhancement image is segmented and the road segmentation result is output.

[0012] Preferably, extracting road information of different scales from a remote sensing image based on a mixed-scale feature extraction unit to obtain a remote sensing image feature map specifically includes:

[0013] Step 1.1: Preprocess the remote sensing image;

[0014] Step 1.2: Input the preprocessed remote sensing image into the initial convolution layer to obtain the initial feature map f o ;

[0015] Step 1.3: Initial feature map f o After passing through the normalization layer, the upper branch and lower branch features are extracted respectively, and the features of the upper branch and the lower branch are spliced ​​to obtain the corresponding feature map f 1 and feature map f 2 ;

[0016] Step 1.4: Feature map f 1 The feature map f is obtained by sequentially passing through a 3×3 depth-separable convolution layer and a ReLU activation function. 1' ;

[0017] Feature map f 2 The feature map f is obtained by sequentially passing through a 5×5 depth-separable convolution layer and a ReLU activation function. 2′ ;

[0018] Step 1.5: Transform the feature map f 1' and feature map f 2' Splice to get the fusion feature map f + ;

[0019] Step 1.6: Fusion feature map f + After the windmill convolution and the initial feature map f o Add and get the remote sensing image feature map

[0020] Preferably, the dynamic intelligent core feature enhancement unit extracts multi-scale features and global context information from the remote sensing image feature map to obtain a feature enhancement map, specifically including:

[0021] Step 2.1: Linearly project the remote sensing image feature map and input it into the depth-separable convolution layer of different sizes to obtain two feature maps and feature maps

[0022] Step 2.2: Feature Map The weighted feature map A is obtained by sequentially passing through the average pooling layer, the dynamic deformable convolution layer and the sigmoid activation function. 5×5 ;

[0023] Feature Map The weighted feature map A is obtained by sequentially passing through the maximum pooling layer, the full-dimensional dynamic convolution layer and the sigmoid activation function. 7×7 ;

[0024] Step 2.3: Feature map and feature maps Splice and combine the spliced ​​feature maps with the weighted feature maps A 5×5 And weighted feature map A 7×7 Multiply and fuse the two corresponding weighted feature maps to obtain the feature map

[0025] Step 2.4: Feature map The weights are obtained by averaging the global pooling layer, convolution layer and sigmoid activation function respectively.

[0026] Step 2.5: Weight and feature maps Perform point-by-point multiplication and compare the features after point-by-point multiplication with the remote sensing image feature map The features in the graph are skipped to obtain the feature enhancement graph

[0027] Preferably, the feature enhancement map is segmented based on the Mamba enhancement and Diffusion model, and the road segmentation result is output, which specifically includes:

[0028] Step 3.1: After the feature enhancement map passes through the super-resolution module, the resolution of the feature map is expanded by upsampling, and the detail area is enhanced by Mamba enhancement;

[0029] Step 3.2: After processing in step 3.1, remove noise through the Diffusion model;

[0030] Step 3.3: After processing in step 3.2, the convolutional layer is used for classification and the road segmentation result is output.

[0031] Preferably, step 1.3 specifically includes:

[0032] When extracting features from the upper branch, the initial feature map f o After passing through the 1×1 convolution layer, 3×3 depth-separable convolution layer and ReLU activation function, the feature f is obtained. up ;

[0033] When extracting features from the lower branch, the initial feature map f o After passing through the 1×1 convolution layer, 5×5 depth-separable convolution layer and ReLU activation function, the feature f is obtained. down ;

[0034] The feature f up and feature f down Splicing is performed to obtain the corresponding feature map f 1 and feature map f 2 , the calculation formula is:

[0035] f 1 =Concat(f up , f down ), f 2 =Concat(f down , f up )

[0036] Among them, Concat represents the concatenation operation of feature maps.

[0037] Preferably, the model loss function is:

[0038] L total =L1+α·L2+β·L3

[0039]

[0040] Among them, α and β are hyperparameters used to balance the losses, L1 is the cross entropy loss, L2 is the Dice loss, L3 is the adaptive regularization loss, and y i is the actual label, is the predicted output of the model, λ is the regularization strength, ||·||2 represents the L2 norm, and i represents the pixel in the data sample.

[0041] Remote sensing image road segmentation system based on hybrid scale feature extraction and dynamic intelligent core feature enhancement, including:

[0042] Mixed-scale feature extraction unit: used to extract road information of different scales from remote sensing images and obtain remote sensing image feature maps;

[0043] Dynamic intelligent core feature enhancement unit: used to extract multi-scale features and global context information from remote sensing image feature maps to obtain feature enhancement maps;

[0044] Road segmentation unit: used to perform road segmentation on the feature enhancement map based on Mamba enhancement and Diffusion model, and output the road segmentation result.

[0045] A computer device includes: a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, a remote sensing image road segmentation method based on mixed-scale feature extraction and dynamic intelligent core feature enhancement is implemented.

[0046] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, a remote sensing image road segmentation method based on mixed-scale feature extraction and dynamic intelligent core feature enhancement is implemented.

[0047] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a method, system, device and medium for remote sensing image road segmentation based on hybrid scale feature extraction and dynamic intelligent core feature enhancement, which has the following advantages:

[0048] 1) The mixed-scale feature extraction module can extract road features at multiple scales by fusing multi-scale information, significantly improving the model's adaptability to different image scales. In practical applications, the size of images, the width, shape, and complexity of roads often vary, and traditional single-scale feature extraction methods are prone to losing important information. The mixed-scale feature extraction module uses convolution kernels of different scales (such as 1×1, 3×3, 5×5, etc.) to perform multi-scale feature extraction on the input image, further integrating feature information at these different scales to ensure accurate road segmentation regardless of the specific scale of the road in the image. The design of this module can better capture the image from large-scale road frameworks to small-scale road details, improving the model's performance on images of different sizes. It is of great significance, especially for road detection and segmentation tasks in complex scenes.

[0049] 2) Under changing environmental conditions, the background of an image can be very complex, and road details can become blurred or unclear due to a variety of factors. Traditional methods often have difficulty accurately distinguishing between the road and the background. The Dynamic Intelligent Core Feature Enhancement Module is designed to address this issue. It adaptively selects the most representative and discriminative features in the image, thereby enhancing the expressiveness of the features and thus improving segmentation accuracy. This module uses a self-attention mechanism or similar adaptive weighting method to dynamically adjust the contribution of different regions or features in the model based on their importance, ensuring that key features are strengthened in subsequent calculations. For example, in areas with complex backgrounds, it can automatically amplify road-related features while suppressing irrelevant background information. When road details are blurred, the module can automatically enhance the capture of road contours to prevent over-smoothing that leads to loss of details, thereby effectively improving segmentation accuracy, especially in cases with complex backgrounds or unclear road details. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0051] Figure 1 This is a flow chart of the remote sensing image road segmentation method based on hybrid scale feature extraction and dynamic intelligent core feature enhancement provided by the present invention.

[0052] Figure 2 This is a processing flow chart of the mixed-scale feature extraction unit provided by the present invention.

[0053] Figure 3This is a processing flow chart of the dynamic intelligent core feature enhancement unit provided by the present invention.

[0054] Figure 4 This is a structural block diagram of the remote sensing image road segmentation system based on hybrid scale feature extraction and dynamic intelligent core feature enhancement provided by the present invention. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0056] The embodiment of the present invention discloses a remote sensing image road segmentation method based on hybrid scale feature extraction and dynamic intelligence core feature enhancement. Figure 1 Shown, including:

[0057] Extracting road information of different scales from remote sensing images based on a mixed-scale feature extraction unit to obtain a remote sensing image feature map;

[0058] The dynamic intelligent core feature enhancement unit is used to extract multi-scale features and global context information from the remote sensing image feature map to obtain a feature enhancement map.

[0059] Based on the Mamba enhancement and Diffusion model, the feature enhancement image is segmented and the road segmentation result is output.

[0060] In this embodiment, the specific processing process of the mixed-scale feature extraction unit is as follows: Figure 2 shown.

[0061] Step 1.1: Preprocess the input remote sensing image. First, remote sensing images often contain noise and different lighting conditions, so preprocessing is required to improve the performance of the model. Normalization: Adjust the pixel value range to [0,1] to ensure the stability of the gradient during network training. Denoising: Remove noise from the remote sensing image through filtering operations. Image enhancement: Adjust the brightness, contrast, etc. of the image to increase the diversity of training samples. Preprocessed image It will be used as the input of the neural network and represented as a three-dimensional matrix: Where H is the height of the image, W is the width of the image, and C is the number of channels in the image.

[0062] Step 1.2: Feature extraction of initial convolutional layer. After preprocessing, the image The initial convolution layer of the network is used to extract preliminary feature information. The function of this convolution layer is to extract low-level features of the image (such as edges, textures, etc.) by learning the convolution kernel. The output of the initial convolution layer is the initial feature map f o .

[0063] Step 1.3: Dual-branch network feature extraction. Next, the initial feature map f extracted by the initial convolutional layer is o The feature maps are passed to the normalization layer. This layer normalizes the features to zero mean and unit variance, further improving the network's convergence speed and robustness. After the normalization layer, the feature maps enter the network's dual-branch structure, undergoing feature extraction in both the upper and lower branches. These two branches are designed to extract features at different scales, enhancing the network's ability to learn road structures in remote sensing images.

[0064] Upper branch: The upper branch first adjusts the number of channels of the input features through a 1×1 convolution layer, then extracts local features through a 3×3 depth-separable convolution layer, and obtains the feature f through the ReLU activation function. up .

[0065] Lower branch: The lower branch adjusts the number of channels through a 1×1 convolution layer, and then uses a 5×5 depth-separable convolution layer to extract features in a wider range. The feature f is also obtained through the ReLU activation function. down .

[0066] Next, the features of the upper and lower branches are spliced. First, cross splicing is performed to obtain two new feature maps f 1 and f 2 .

[0067] f 1 =Concat(f up , f down ), f 2 =Concat(f down , f up )

[0068] Here, Concat represents the concatenation operation of feature maps.

[0069] Step 1.4: Further feature extraction. 1 and f 2 They are sent to the upper branch and the lower branch respectively for further feature extraction.

[0070] Upper branch: feature map f 1 After processing, a 3×3 depth separable convolution layer and ReLU activation function are used to obtain a new feature map f 1′ .

[0071] Lower branch: for feature f 2 After processing, a 5×5 depth separable convolution layer and ReLU activation function are used to obtain a new feature map f 2′ .

[0072] Step 1.5: Feature fusion and output. 1′ and f 2’ Splice and get the final fusion feature map f + This fusion feature map f + It includes road information from different scales and enhances the expressive power of the road segmentation task.

[0073] Step 1.6: Fuse feature map f + After the pinwheel-shaped convolution (PSConv) operation, it is combined with the initial feature map f o Add together to get the final remote sensing image feature map This step obtains richer and more accurate road segmentation information by combining local and global features, laying the foundation for subsequent segmentation tasks.

[0074] The hybrid-scale feature extraction unit of this invention captures road information at different scales through a multi-scale feature extraction approach. By combining convolution kernels of different sizes, it extracts both large-scale framework features and small-scale detail features within the image. This ensures that the model maintains good performance when processing images of varying sizes, avoiding the neglect of detail information at certain scales by traditional methods. Through this process, the network can effectively extract road features from remote sensing images, providing reliable input for road segmentation tasks.

[0075] In this embodiment, the specific processing process of the dynamic intelligence core feature enhancement unit is as follows: Figure 3 This step aims to construct a dynamic intelligent core feature enhancement unit, which effectively extracts multi-scale features and global context information to enhance the performance of road segmentation in remote sensing images.

[0076] Step 2.1: Feature input and linear projection. Remote sensing image feature map After linear projection, the number of channels is reduced to a smaller dimension. This step is accomplished through a 1×1 convolutional layer, reducing computational complexity and preparing for subsequent feature extraction. This operation reduces the number of channels in the feature map, lowering computational complexity while preserving the integrity of feature information, helping to improve computational efficiency. This optimization allows subsequent convolution operations to be performed at a lower computational cost without sacrificing segmentation performance.

[0077] In order to effectively extract features of different scales, after linear projection, it is input into two depth-wise separable convolutional layers of different sizes: 5×5 and 7×7. These two convolutional layers generate two feature maps respectively: and Depthwise separable convolutions significantly reduce computational complexity while simultaneously capturing features at different scales, adapting to the varying sizes of objects such as roads in remote sensing images. By using multiple large-scale convolution kernels, the model can gradually expand its receptive field, better capturing long-range dependencies and understanding global context.

[0078] Step 2.2: Dynamic feature selection and enhancement. Next, and For dynamic selection and enhancement. First, the dimensionality is reduced by the average pooling layer, then input into the dynamic deformable convolution layer for adaptive feature selection, and finally the weighted feature map A is obtained by the sigmoid activation function. 5×5 :

[0079]

[0080] Here, σ represents the sigmoid activation function, DCNv4 represents the dynamic deformable convolution, and AVP represents the average pooling layer. This dynamic adjustment mechanism adaptively highlights the most important channel and spatial features based on the global information of the input features, thereby improving the robustness and adaptability of the model and enabling better separation of features from complex backgrounds.

[0081] for The dimensionality is reduced through the maximum pooling layer, and then input into the full-dimensional dynamic convolution layer, and finally the weighted feature map A is obtained through the sigmoid activation function. 7×7 :

[0082]

[0083] ODConv represents a full-dimensional dynamic convolution layer, and MAP represents a maximum pooling layer. This dynamic selection mechanism can guide feature selection and weighting based on global information, further improving segmentation accuracy and ensuring that important features are not overlooked, especially when processing remote sensing images with multi-scale features.

[0084] Step 2.3: Multi-scale feature fusion. Feature maps from different scales and By splicing feature maps of different scales, the model can retain the key information captured from each scale and ensure that this information can be effectively integrated in subsequent steps. 5×5 and A 7×7Multiply them together to get the weighted feature map. Finally, the two feature maps are added and fused to get the feature map This dynamic fusion mechanism ensures that information from different scales can be fully combined while retaining the most critical features, significantly improving the detection ability of targets such as roads in remote sensing images.

[0085] Step 2.4: Global pooling and feature enhancement. Feature map The global information is compressed through the average global pooling layer, and then the final weight is obtained through the convolution layer and sigmoid activation function.

[0086]

[0087] Among them, GAP represents global average pooling. Weight The feature map is weighted to highlight the part that is most helpful for segmentation.

[0088] Step 2.5: Weight and This integration of global information not only improves the segmentation accuracy, but also enhances the generalization ability of the model under different backgrounds. Perform jump connections and finally obtain the feature enhancement graph

[0089]

[0090] in, represents point-by-point multiplication, Represents the additive fusion operation. Through skip connections, the model can utilize the comprehensive information of original features and enhanced features to further improve the accuracy and effect of road segmentation in remote sensing images.

[0091] The dynamic intelligent core feature enhancement unit in this invention adaptively adjusts feature weights to ensure that the most critical features are enhanced in complex backgrounds and blurred road scenes. During the segmentation process, this module intelligently identifies the most important features and dynamically adjusts their impact on the segmentation result, effectively improving segmentation accuracy. This is particularly important in complex backgrounds or unclear road details, preventing detail loss caused by over-smoothing of the model.

[0092] In this embodiment, the final segmentation result and training loss function are obtained based on the Mamba enhancement and Diffusion model.

[0093] Step 3.1: Super-resolution and Mamba enhancement. In order to improve the accuracy of remote sensing image segmentation, especially the recovery and enhancement of detail areas, we will The input is fed into a super-resolution module, which helps restore the image's detailed features by increasing its spatial resolution. The output of the super-resolution module expands the resolution of the feature map through upsampling, refining the image's details, which is particularly important for edge and road segmentation. In this process, we incorporate the Mamba model (a diffusion-based generative model). The Mamba enhancement module can further enhance detailed areas through generative adversarial learning, particularly repairing blurred or unclear areas. The introduction of the Mamba model further optimizes image details and enhances high-frequency information, making segmentation results more accurate.

[0094] Step 3.2: Diffusion Model and Global Context Modeling. To further enhance global context modeling capabilities, we introduced a diffusion model. This model uses a step-by-step guided noise removal approach to recover image details from noise and simulate the image's underlying structure, capturing more global information. The output generated by the diffusion model helps the model understand the global context in remote sensing images and enhances the distinction between background and foreground. This model helps handle complex backgrounds and subtle differences between roads, enhancing the model's adaptability to diverse environments.

[0095] Step 3.3: Generate the final segmentation result. After the super-resolution, Mamba enhancement and Diffusion model processing, the feature map will be input into a convolution layer for final classification to generate the category label of each pixel. Indicates the category to which each pixel belongs.

[0096] To train the model, a composite loss function was designed that combines cross-entropy loss, Dice loss, and a new adaptive regularization loss to ensure the model has good robustness and adaptability when processing remote sensing images. Cross-entropy loss: Cross-entropy loss is used to measure the difference between the model output and the true label, and is particularly suitable for pixel-level classification tasks. In road segmentation in remote sensing images, cross-entropy loss can help the model effectively distinguish between road and background:

[0097]

[0098] Among them, y i is the actual label, Is the predicted output of the model. Dice loss: Dice loss is used to measure the overlap between the predicted result and the actual label, which is particularly suitable for processing unbalanced data (such as road segmentation tasks). It can improve the segmentation accuracy of the model for small objects (such as small roads):

[0099]

[0100] Adaptive regularization loss: Adaptive regularization loss is used to enhance the generalization ability of the model and ensure that the model can maintain high segmentation performance in complex scenarios. This loss function helps the model adaptively select useful features by dynamically adjusting parameters:

[0101]

[0102] Among them, λ is the regularization strength, ||·||2 represents the L2 norm, which is used to measure the difference between features. Final loss function: The weighted sum of the above losses is used to obtain the final training loss function:

[0103] L total =L1+α·L2+β·L3

[0104] Among them, α and β are hyperparameters used to balance various losses.

[0105] The embodiment of the present invention provides a remote sensing image road segmentation system based on hybrid scale feature extraction and dynamic intelligence core feature enhancement, such as Figure 4 Shown, including:

[0106] Mixed-scale feature extraction unit: used to extract road information of different scales from remote sensing images and obtain remote sensing image feature maps;

[0107] Dynamic intelligent core feature enhancement unit: used to extract multi-scale features and global context information from remote sensing image feature maps to obtain feature enhancement maps;

[0108] Road segmentation unit: used to perform road segmentation on the feature enhancement map based on Mamba enhancement and Diffusion model, and output the road segmentation result.

[0109] The specific implementation process and method of each unit of the system of the present invention are consistent with each other and will not be repeated here.

[0110] This embodiment provides a computer device including a memory and a processor. The memory stores a computer program that can be run on the processor. When the processor executes the computer program, a remote sensing image road segmentation method based on mixed-scale feature extraction and dynamic intelligent core feature enhancement is implemented.

[0111] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a remote sensing image road segmentation method based on hybrid-scale feature extraction and dynamic intelligent core feature enhancement is implemented.

[0112] Those skilled in the art will understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks or optical disks, and other media that can store program codes.

[0113] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0114] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A remote sensing image road segmentation method based on hybrid scale feature extraction and dynamic intelligent core feature enhancement is characterized by: include: Extracting road information of different scales from remote sensing images based on a mixed-scale feature extraction unit to obtain a remote sensing image feature map; The dynamic intelligent core feature enhancement unit is used to extract multi-scale features and global context information from the remote sensing image feature map to obtain a feature enhancement map. Perform road segmentation on the feature enhancement image based on Mamba enhancement and Diffusion model, and output the road segmentation result; The dynamic intelligent core feature enhancement unit extracts multi-scale features and global context information from the remote sensing image feature map to obtain a feature enhancement map, which specifically includes: Step 2.1: Linearly project the remote sensing image feature map and input it into the depth-separable convolution layer of different sizes to obtain two feature maps and feature maps Step 2.2: Feature Map The weighted feature map A is obtained by sequentially passing through the average pooling layer, the dynamic deformable convolution layer and the sigmoid activation function. 5×5 , Among them, σ represents the sigmoid activation function, DCNv4 represents dynamic deformable convolution, and AVP represents the average pooling layer; Feature Map The weighted feature map A is obtained by sequentially passing through the maximum pooling layer, the full-dimensional dynamic convolution layer and the sigmoid activation function. 7×7 , Among them, ODConv represents the full-dimensional dynamic convolution layer, and MAP represents the maximum pooling layer; Step 2.3: Feature map and feature maps Splice and combine the spliced ​​feature maps with the weighted feature maps A 5×5 And weighted feature map A 7×7 Multiply and fuse the two corresponding weighted feature maps to obtain the feature map Step 2.4: Feature map The weights are obtained by averaging the global pooling layer, convolution layer and sigmoid activation function respectively. Step 2.5: Weight and feature maps Perform point-by-point multiplication and compare the features after point-by-point multiplication with the remote sensing image feature map The features in the graph are skipped to obtain the feature enhancement graph .

2. The method for remote sensing image road segmentation based on hybrid scale feature extraction and dynamic intelligence core feature enhancement according to claim 1, characterized in that: The mixed-scale feature extraction unit is used to extract road information of different scales from the remote sensing image to obtain a remote sensing image feature map, specifically including: Step 1.1: Preprocess the remote sensing image; Step 1.2: Input the preprocessed remote sensing image into the initial convolution layer to obtain the initial feature map f o ; Step 1.3: Initial feature map f o After passing through the normalization layer, the upper branch and lower branch features are extracted respectively, and the features of the upper branch and the lower branch are spliced ​​to obtain the corresponding feature map f 1 and feature map f 2 ; Step 1.4: Feature map f 1 The feature map f is obtained by sequentially passing through a 3×3 depth-separable convolution layer and a ReLU activation function. 1′ ; Feature map f 2 The feature map f is obtained by sequentially passing through a 5×5 depth-separable convolution layer and a ReLU activation function. 2’ ; Step 1.5: Transform the feature map f 1’ and feature map f 2’ Splice to get the fusion feature map f + ; Step 1.6: Fusion feature map f + After the windmill convolution and the initial feature map f o Add and get the remote sensing image feature map .

3. The method for remote sensing image road segmentation based on hybrid scale feature extraction and dynamic intelligence core feature enhancement according to claim 1, characterized in that: Based on the Mamba enhancement and Diffusion model, the feature enhancement image is segmented and the road segmentation results are output, including: Step 3.1: After the feature enhancement map passes through the super-resolution module, the resolution of the feature map is expanded by upsampling, and the detail area is enhanced by Mamba enhancement; Step 3.2: After processing in step 3.1, remove noise through the Diffusion model; Step 3.3: After processing in step 3.2, the convolutional layer is used for classification and the road segmentation result is output.

4. The method for remote sensing image road segmentation based on hybrid scale feature extraction and dynamic intelligence core feature enhancement according to claim 2, characterized in that: Step 1.3 specifically includes: When extracting features from the upper branch, the initial feature map f o After passing through the 1×1 convolution layer, 3×3 depth-separable convolution layer and ReLU activation function, the feature f is obtained. up ; When extracting features from the lower branch, the initial feature map f o After passing through the 1×1 convolution layer, 5×5 depth-separable convolution layer and ReLU activation function, the feature f is obtained. down ; The feature f up and feature f down Splicing is performed to obtain the corresponding feature map f 1 and feature map f 2 , the calculation formula is: f 1 =Concat(f up ,f down ),f 2 =Concat(f down ,f up ) Among them, Concat represents the concatenation operation of feature maps.

5. The method for remote sensing image road segmentation based on hybrid scale feature extraction and dynamic intelligence core feature enhancement according to claim 1, characterized in that: The model loss function is: L total =L1+α·L2+β·L3 Among them, α and β are hyperparameters used to balance the losses, L1 is the cross entropy loss, L2 is the Dice loss, L3 is the adaptive regularization loss, and y i is the actual label, is the predicted output of the model, λ is the regularization strength, ||·||2 represents the L2 norm, and i represents the pixel in the data sample.

6. A remote sensing image road segmentation system based on hybrid scale feature extraction and dynamic intelligent core feature enhancement is characterized by: include: Mixed-scale feature extraction unit: used to extract road information of different scales from remote sensing images and obtain remote sensing image feature maps; Dynamic intelligent core feature enhancement unit: used to extract multi-scale features and global context information from remote sensing image feature maps to obtain feature enhancement maps; Specifically include: Step 2.1: Linearly project the remote sensing image feature map and input it into the depth-separable convolution layer of different sizes to obtain two feature maps and feature maps Step 2.2: Feature Map The weighted feature map A is obtained by sequentially passing through the average pooling layer, the dynamic deformable convolution layer and the sigmoid activation function. 5×5 , Among them, σ represents the sigmoid activation function, DCNv4 represents dynamic deformable convolution, and AVP represents the average pooling layer; Feature Map The weighted feature map A is obtained by sequentially passing through the maximum pooling layer, the full-dimensional dynamic convolution layer and the sigmoid activation function. 7×7 , Among them, ODConv represents the full-dimensional dynamic convolution layer, and MAP represents the maximum pooling layer; Step 2.3: Feature map and feature maps Splice and combine the spliced ​​feature maps with the weighted feature maps A 5×5 And weighted feature map A 7×7 Multiply and fuse the two corresponding weighted feature maps to obtain the feature map Step 2.4: Feature map The weights are obtained by averaging the global pooling layer, convolution layer and sigmoid activation function respectively. Step 2.5: Weight and feature maps Perform point-by-point multiplication and compare the features after point-by-point multiplication with the remote sensing image feature map The features in the graph are skipped to obtain the feature enhancement graph Road segmentation unit: used to perform road segmentation on the feature enhancement map based on Mamba enhancement and Diffusion model, and output the road segmentation result.

7. A computer device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that The storage medium stores a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 5.

Images