Novel image edge feature extraction method and device based on MDR-CNN and AEC technologies, and storage medium

Abstract

The invention discloses a novel image edge feature extraction method based on MDR-CNN and AEC technologies, relates to the technical field of image processing, and aims at effectively solving the problem of multi-scale edge detection in a complex scene and ensuring the comprehensiveness and accuracy of edge information by capturing various features in an image through convolution kernels with different sizes and different scales. And secondly, a residual module is introduced into the convolutional network, so that the problem of gradient disappearance is avoided to a certain extent, the feature expression capability is enhanced, and the network can learn image features in a deeper manner. After the construction of the multi-scale deep residual convolutional network is completed, the convolutional weight is automatically adjusted according to the local gray change and texture features of the image by using an adaptive mechanism, and the features of the edge region are significantly enhanced, so that the method can better adapt to different image contents and improve the robustness of edge detection.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and specifically to a novel image edge feature extraction method, apparatus, and storage medium based on MDR-CNN and AEEC technology. Background Technology

[0002] In the field of image processing, edge feature extraction is a crucial task. It plays a central role in numerous applications, such as object recognition, image segmentation, medical image analysis, and autonomous driving. In object recognition, edge feature extraction technology helps systems accurately identify target objects, playing a key role in industrial inspection, monitoring, and control. In image segmentation, edge feature extraction is an image processing technique that achieves efficient analysis by segmenting images, making image recognition technology more accurate, secure, and reliable.

[0003] However, existing edge feature extraction methods still face some technical challenges, especially when dealing with complex backgrounds, low-contrast images, blurred edges, and multi-scale scenes. Traditional edge detection methods often fail to effectively extract accurate edge information, are greatly affected by the usage scenario, and have poor image processing accuracy. Summary of the Invention

[0004] In order to overcome the shortcomings of the above technologies, the present invention provides an image edge feature extraction method, apparatus and storage medium that can capture image edge features of different sizes and characteristics, and at the same time significantly improve the accuracy and efficiency of edge extraction by enhancing the clarity of edge information and reducing the influence of noise and blur on edge extraction.

[0005] The technical solution adopted by this invention to overcome its technical problems is: A novel image edge feature extraction method based on MDR-CNN and AEEC technology includes: S1. Acquire Image ; S2. Transfer the image Preprocessing is performed to obtain the preprocessed image. ; S3. Process the preprocessed image Perform convolution operations to obtain feature maps. ; S4. Utilize the processed image and feature map The final fused feature map is obtained. ; S5. Utilize the final fused feature map Obtain the gradient matrix ; S6. Using the gradient matrix Obtain the weighted graph ; S7. Using a weighted graph With the final fused feature map Obtain the feature map after edge enhancement ; S8. Final fused feature map Obtain the feature map after edge enhancement Perform weighted fusion to obtain the final fused feature map. .

[0006] Furthermore, step S2 includes the following steps: S2-1. Transfer the image Perform Gaussian filtering to obtain the filtered image. ; S2-2. Filter the image. Normalization is performed to obtain the normalized image. .

[0007] Furthermore, step S3 includes the following steps: S3-1. The preprocessed image... The input is fed into the first convolutional layer, and the output is the feature vector. ; S3-2. The preprocessed image... The input is fed into the second convolutional layer, and the output is the feature vector. ; S3-3. The preprocessed image... The input is fed into the third convolutional layer, and the output is the feature vector. ; S3-4. Eigenvectors eigenvectors eigenvectors The feature map is obtained by performing a stitching operation. .

[0008] Furthermore, step S4 includes the following steps: S4-1. Process the image With feature map Perform an addition operation to obtain the feature map. ; S4-2. Feature Map The input is fed into the ReLU activation function, and the output is the final fused feature map. Furthermore, in step S5, the Sobel operator is used to compute the final fused feature map. The gradient magnitude of each pixel is used to obtain the gradient matrix. .

[0009] Furthermore, step S6 includes the following steps: S6-1. Gradient matrix The input is fed into a convolutional layer, and the output is a feature map. ; S6-2. Feature Map The input is fed into the Sigmoid function, and the output is the weight map. .

[0010] In step S7, the weight map is... With the final fused feature map Perform element-wise multiplication to obtain the feature map after edge enhancement. .

[0011] Furthermore, step S8 includes the following steps: S8-1. Feature map after edge enhancement With the final fused feature map Perform a stitching operation to obtain the feature map. ; S8-2. Feature Map The input is fed into a convolutional layer, and the output is a feature map. ; S8-3. Feature Map The input is fed into the Sigmoid function, and the output is the fused weight map. ; S8-4. Through formula The final fused feature map is calculated. .

[0012] On the other hand, the present invention also relates to a novel image edge feature extraction device based on MDR-CNN and AEEC technology, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; wherein: The memory is used to store computer programs; The processor is configured to execute by running programs stored in the memory: Get Image ; Image Preprocessing is performed to obtain the preprocessed image. ; Preprocessed image Perform convolution operations to obtain feature maps. ; Using the processed image and feature map The final fused feature map is obtained. ; Using the final fused feature map Obtain the gradient matrix ; Using the gradient matrix Obtain the weighted graph ; Using weighted graphs With the final fused feature map Obtain the feature map after edge enhancement ; For the final fused feature map Obtain the feature map after edge enhancement Perform weighted fusion to obtain the final fused feature map. .

[0013] On the other hand, the present invention also relates to a computer-readable storage medium on which a computer program is stored, the computer program being implemented when executed by a processor: Get Image ; Image Preprocessing is performed to obtain the preprocessed image. ; Preprocessed image Perform convolution operations to obtain feature maps. ; Using the processed image and feature map The final fused feature map is obtained. ; Using the final fused feature map Obtain the gradient matrix ; Using the gradient matrix Obtain the weighted graph ; Using weighted graphs With the final fused feature map Obtain the feature map after edge enhancement ; For the final fused feature map Obtain the feature map after edge enhancement Perform weighted fusion to obtain the final fused feature map. .

[0014] The beneficial effects of this invention are as follows: Traditional single-edge detection methods suffer from problems such as being greatly affected by the application scenario and having poor image processing accuracy. This invention has several significant advantages over traditional detection methods: First, by capturing various features in the image through convolutional kernels of different sizes and scales, it effectively solves the multi-scale edge detection problem in complex scenes, ensuring the comprehensiveness and accuracy of edge information. Second, by introducing a residual module into the convolutional network, it avoids the gradient vanishing problem to a certain extent, enhances feature representation capabilities, and enables the network to learn image features at a deeper level. After completing the construction of the aforementioned multi-scale deep residual convolutional network, this invention uses an adaptive mechanism to automatically adjust the convolutional weights based on the local grayscale changes and texture features of the image, significantly enhancing the features of the edge region, thereby better adapting to different image content and improving the robustness of edge detection. Finally, an adaptive fusion strategy integrates multi-scale features and edge enhancement features to generate the final edge feature map. Detailed Implementation

[0015] The preferred embodiments of the present invention will now be described in detail so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby providing a clearer and more explicit definition of the scope of protection of the present invention.

[0016] In the description of this invention, it should be noted that the embodiments described in this invention are only some embodiments of this invention, not all embodiments; based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0017] The terms "first," "second," etc., used in this specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] Example 1: A novel image edge feature extraction method based on MDR-CNN and AEEC technology includes: S1. Acquire Image .

[0019] S2. Transfer the image Preprocessing is performed to obtain the preprocessed image. .

[0020] S3. Process the preprocessed image Perform convolution operations to obtain feature maps. .

[0021] S4. Utilize the processed image and feature map The final fused feature map is obtained. .

[0022] S5. Utilize the final fused feature map Obtain the gradient matrix .

[0023] S6. Using the gradient matrix Obtain the weighted graph .

[0024] S7. Using a weighted graph With the final fused feature map Obtain the feature map after edge enhancement .

[0025] S8. Final fused feature map Obtain the feature map after edge enhancement Perform weighted fusion to obtain the final fused feature map. The final fused feature map By enhancing edges, this method achieves more accurate edge feature extraction for complex scenes, addressing the challenge of acquiring edge information at different scales and with different feature maps in low-contrast, high-noise environments. This novel edge detection method, capable of high efficiency and accuracy in complex backgrounds, low-contrast images, blurred edges, and multi-scale scenes, possesses significant research value and practical significance. On one hand, this type of method overcomes the shortcomings of traditional methods and models in dealing with low-contrast images and blurred edges, exhibiting greater robustness in high-noise and multi-scale environments. It also promotes the development of fields such as medical image analysis, autonomous driving, and remote sensing image processing. The construction of this edge extraction method represents a significant breakthrough in edge detection technology, providing a new solution for image processing while improving the accuracy and robustness of edge detection. This method can be applied to more complex scenes, expanding the application scope of image processing technology and possessing significant practical and research value.

[0026] This method is applicable to fields such as object recognition, image segmentation, medical image analysis, and autonomous driving. In medical image analysis, accurate edge detection is crucial for identifying lesion areas; in autonomous driving, clear edge information helps vehicles better perceive their surroundings and improves driving safety. With improved detection accuracy, this method has broad application prospects in these fields and can significantly improve system performance and reliability. Using the method described in this invention, edge features of images can be effectively extracted in all embodiments.

[0027] In this embodiment, step S2 includes the following steps: S2-1. Transfer the image Perform Gaussian filtering to obtain the filtered image. .

[0028] S2-2. Filter the image. Normalization is performed to obtain the normalized image. .

[0029] Gaussian filtering is used to preprocess the input image, convolution operation is used to smooth the image and reduce noise interference to edge detection, and the image is normalized to map pixel values ​​to a uniform range to ensure the stability and accuracy of subsequent edge extraction.

[0030] In this embodiment, step S3 includes the following steps: S3-1. The preprocessed image... The input is fed into the first convolutional layer, and the output is the feature vector. .

[0031] S3-2. The preprocessed image... The input is fed into the second convolutional layer, and the output is the feature vector. .

[0032] S3-3. The preprocessed image... The input is fed into the third convolutional layer, and the output is the feature vector. .

[0033] Convolutional layers with kernels of varying sizes are used to extract features from images, capturing edge information at different levels and scales, thus efficiently capturing both local and global features. The outputs of each layer are connected through a residual structure, which preserves the original feature information through skip connections, thereby reducing information loss and enhancing the network's ability to express features.

[0034] S3-4. Eigenvectors eigenvectors eigenvectors The feature map is obtained by performing a stitching operation. .

[0035] In this embodiment, step S4 includes the following steps: S4-1. Process the image With feature map Perform an addition operation to obtain the feature map. .

[0036] S4-2. Feature Map The input is fed into the ReLU activation function, and the output is the final fused feature map. In this embodiment, in step S5, the Sobel operator is used to calculate the final fused feature map. The gradient magnitude of each pixel is used to obtain the gradient matrix. .

[0037] In this embodiment, step S6 includes the following steps: S6-1. Gradient matrix The input is fed into a convolutional layer, and the output is a feature map. .

[0038] S6-2. Feature Map The input is fed into the Sigmoid function, and the output is the weight map. .

[0039] In this embodiment, step S7 involves weighting the graph. With the final fused feature map Perform element-wise multiplication to obtain the feature map after edge enhancement. .

[0040] The system automatically enhances the edge response based on the feature intensity of the edge region and dynamically adjusts the convolution kernel weights, thereby enhancing the response of the edge region, highlighting the features of the edge region, and improving the accuracy of edge detection.

[0041] In this embodiment, step S8 includes the following steps: S8-1. Feature map after edge enhancement With the final fused feature map Perform a stitching operation to obtain the feature map. .

[0042] S8-2. Feature Map The input is fed into a convolutional layer, and the output is a feature map. .

[0043] S8-3. Feature Map The input is fed into the Sigmoid function, and the output is the fused weight map. .

[0044] S8-4. Through formula The final fused feature map is calculated. .

[0045] By fusing multi-scale convolutional features and edge enhancement features using a weighted average method, not only are the detailed information of the multi-scale features preserved, but the response of the edge region is also enhanced through an adaptive adjustment mechanism, thereby obtaining the final edge feature map and achieving efficient and accurate edge extraction.

[0046] Example 2 A novel image edge feature extraction device based on MDR-CNN and AEEC technologies includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus; wherein: The memory is used to store computer programs; The processor is configured to execute by running programs stored in the memory: Get Image . Image Preprocessing is performed to obtain the preprocessed image. .

[0047] Preprocessed image Perform convolution operations to obtain feature maps. .

[0048] Using the processed image and feature map The final fused feature map is obtained. .

[0049] Using the final fused feature map Obtain the gradient matrix .

[0050] Using the gradient matrix Obtain the weighted graph .

[0051] Using weighted graphs With the final fused feature map Obtain the feature map after edge enhancement .

[0052] For the final fused feature map Obtain the feature map after edge enhancement Perform weighted fusion to obtain the final fused feature map. .

[0053] Example 3: A computer-readable storage medium having a computer program stored thereon, the computer program being implemented when executed by a processor: Get Image .

[0054] Image Preprocessing is performed to obtain the preprocessed image. .

[0055] Preprocessed image Perform convolution operations to obtain feature maps. .

[0056] Using the processed image and feature map The final fused feature map is obtained. .

[0057] Using the final fused feature map Obtain the gradient matrix .

[0058] Using the gradient matrix Obtain the weighted graph .

[0059] Using weighted graphs With the final fused feature map Obtain the feature map after edge enhancement .

[0060] For the final fused feature map Obtain the feature map after edge enhancement Perform weighted fusion to obtain the final fused feature map. .

[0061] It should be understood that in the various embodiments of this document, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this document.

[0062] It should also be understood that, in the embodiments herein, the term "and / or" is merely a description of the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, the character " / " in this document generally indicates that the preceding and following associated objects have an "or" relationship.

[0063] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this document.

[0064] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0065] In the embodiments provided herein, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or units, or they may be electrical, mechanical, or other forms of connection.

[0066] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments described herein, depending on actual needs.

[0067] Furthermore, the functional units in the various embodiments of this document can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0068] If the integrated unit is implemented as 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 this paper, 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. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this paper. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0069] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A novel image edge feature extraction method based on MDR-CNN and AEEC technology, characterized in that, include: S1. Acquire Image ; S2. Transfer the image Preprocessing is performed to obtain the preprocessed image. ; S3. Process the preprocessed image Perform convolution operations to obtain feature maps. ; S4. Utilize the processed image and feature map The final fused feature map is obtained. ; S5. Utilize the final fused feature map Obtain the gradient matrix ; S6. Using the gradient matrix Obtain the weighted graph ; S7. Using a weighted graph With the final fused feature map Obtain the feature map after edge enhancement ; S8. Final fused feature map Obtain the feature map after edge enhancement Perform weighted fusion to obtain the final fused feature map. .

2. The novel image edge feature extraction method based on MDR-CNN and AEEC technology according to claim 1, characterized in that, Step S2 includes the following steps: S2-1. Transfer the image Perform Gaussian filtering to obtain the filtered image. ; S2-2. Filter the image. Normalization is performed to obtain the normalized image. .

3. The novel image edge feature extraction method based on MDR-CNN and AEEC technology according to claim 1, characterized in that, Step S3 includes the following steps: S3-1. The preprocessed image... The input is fed into the first convolutional layer, and the output is the feature vector. ; S3-2. The preprocessed image... The input is fed into the second convolutional layer, and the output is the feature vector. ; S3-3. The preprocessed image... The input is fed into the third convolutional layer, and the output is the feature vector. ; S3-4. Eigenvectors eigenvectors eigenvectors The feature map is obtained by performing a stitching operation. .

4. The novel image edge feature extraction method based on MDR-CNN and AEEC technology according to claim 1, characterized in that, Step S4 includes the following steps: S4-1. Process the image With feature map Perform an addition operation to obtain the feature map. ; S4-2. Feature Map The input is fed into the ReLU activation function, and the output is the final fused feature map. .

5. The novel image edge feature extraction method based on MDR-CNN and AEEC technology according to claim 1, characterized in that: In step S5, the Sobel operator is used to compute the final fused feature map. The gradient magnitude of each pixel is used to obtain the gradient matrix. .

6. The novel image edge feature extraction method based on MDR-CNN and AEEC technology according to claim 1, characterized in that, Step S6 includes the following steps: S6-1. Gradient matrix The input is fed into a convolutional layer, and the output is a feature map. ; S6-2. Feature Map The input is fed into the Sigmoid function, and the output is the weight map. .

7. The novel image edge feature extraction method based on MDR-CNN and AEEC technology according to claim 1, characterized in that: In step S7, the weight map is... With the final fused feature map Perform element-wise multiplication to obtain the feature map after edge enhancement. .

8. The novel image edge feature extraction method based on MDR-CNN and AEEC technology according to claim 1, characterized in that, Step S8 includes the following steps: S8-1. Feature map after edge enhancement With the final fused feature map Perform a stitching operation to obtain the feature map. ; S8-2. Feature Map The input is fed into a convolutional layer, and the output is a feature map. ; S8-3. Feature Map The input is fed into the Sigmoid function, and the output is the fused weight map. ; S8-4. Through formula The final fused feature map is calculated. .

9. A novel image edge feature extraction device based on MDR-CNN and AEEC technology, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; wherein: The memory is used to store computer programs; The processor is configured to execute by running programs stored in the memory: Get Image ; Image Preprocessing is performed to obtain the preprocessed image. ; Preprocessed image Perform convolution operations to obtain feature maps. ; Using the processed image and feature map The final fused feature map is obtained. ; Using the final fused feature map Obtain the gradient matrix ; Using the gradient matrix Obtain the weighted graph ; Using weighted graphs With the final fused feature map Obtain the feature map after edge enhancement ; For the final fused feature map Obtain the feature map after edge enhancement Perform weighted fusion to obtain the final fused feature map. .

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is implemented when executed by a processor: Get Image ; Image Preprocessing is performed to obtain the preprocessed image. ; Preprocessed image Perform convolution operations to obtain feature maps. ; Using the processed image and feature map The final fused feature map is obtained. ; Using the final fused feature map Obtain the gradient matrix ; Using the gradient matrix Obtain the weighted graph ; Using weighted graphs With the final fused feature map Obtain the feature map after edge enhancement ; For the final fused feature map Obtain the feature map after edge enhancement Perform weighted fusion to obtain the final fused feature map. .

Images