Road crack detection method and system based on multi-scale feature fusion and superposition U-Net, and medium

Through the road crack detection method of multi-scale feature fusion and superposition U-Net, the attention mechanism and segmentation network are used for semantic segmentation, which solves the high-cost road crack detection problem in the existing technology and realizes efficient and low-cost road crack detection.

CN120656052APending Publication Date: 2025-09-16HANGZHOU HUICUI INTELLIGENT TECH CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202510448023.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-04-10
Publication Date
2025-09-16

AI Technical Summary

Technical Problem

Existing road crack detection technology is costly and difficult to achieve fast and high-precision detection.

Method used

A road crack detection method based on multi-scale feature fusion and superposition U-Net is adopted. The region of interest is analyzed through the attention mechanism, and semantic segmentation is performed in combination with the segmentation network to accurately analyze the characteristics of road defects.

Benefits of technology

It improves the defect recognition capability and efficiency, reduces the detection cost, and realizes efficient road crack detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120656052A_ABST
    Figure CN120656052A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a road crack detection method and system based on multi-scale feature fusion and superposition U-Net, and a medium, and the method comprises the steps: obtaining a road detection image, and carrying out the preprocessing of the road detection image, and obtaining a preprocessed image; analyzing the preprocessed image based on an attention mechanism, generating a region of interest, and obtaining a defect image based on the region of interest; performing semantic extraction on the defect image based on the segmentation network to generate semantic information, and performing semantic segmentation on the semantic information to obtain defect features; analyzing a road crack state based on the defect characteristics to obtain road crack distribution information, and evaluating the road crack distribution information to obtain road crack category information; according to the method, the region of interest is accurately analyzed through an attention mechanism, the defect image is efficiently acquired, semantic segmentation is performed on the defect image through the segmentation network, road defect features are accurately analyzed, and the defect recognition capability and recognition efficiency are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of road detection technology, and more specifically, to a road crack detection method, system, and medium based on multi-scale feature fusion and superposition U-Net. Background Art

[0002] In an era of rapid development in information technology, automation and informatization have become key drivers across all industries. In the field of road engineering, with the rapid development of road construction, road maintenance issues are becoming a growing concern. However, timely detection and remediation of road defects at an early stage could significantly reduce road maintenance costs. Obtaining fast, highly accurate, and cost-effective road crack detection technology at an early stage has become a major challenge facing road maintenance departments. Existing surface detection methods, such as ground-penetrating radar (GPR), laser profiling, and 3D reconstruction, while effective, are prohibitively expensive. Summary of the Invention

[0003] The purpose of the embodiments of the present application is to provide a road crack detection method, system and medium based on multi-scale feature fusion and superposition U-Net, which accurately analyzes the region of interest through the attention mechanism, efficiently obtains defect images, and semantically segments the defect images through the segmentation network, accurately analyzes road defect characteristics, and improves defect recognition capability and efficiency.

[0004] The present application also provides a road crack detection method based on multi-scale feature fusion and superposition U-Net, including:

[0005] Acquire a road detection image, and preprocess the road detection image to obtain a preprocessed image;

[0006] Analyze the pre-processed image based on the attention mechanism, generate the region of interest, and obtain the defect image based on the region of interest;

[0007] Based on the segmentation network, semantic extraction is performed on the defect image to generate semantic information, and the semantic information is segmented to obtain defect features;

[0008] Analyze the road crack status based on defect characteristics to obtain road crack distribution information, evaluate the road crack distribution information, and obtain road crack category information;

[0009] A matching maintenance strategy is established based on the road crack category information and transmitted to the terminal.

[0010] Optionally, in the road crack detection method based on multi-scale feature fusion and superposition U-Net described in the embodiment of the present application, obtaining a road detection image and preprocessing the road detection image to obtain a preprocessed image specifically include:

[0011] Acquire a road detection image, and perform translation processing on the road detection image to obtain a translated image;

[0012] Flip the translated image horizontally and vertically to obtain a flipped image;

[0013] The flipped images are cropped and spliced, and gray-valued to obtain grayscale images;

[0014] The grayscale image is enhanced to obtain a preprocessed image.

[0015] Optionally, in the road crack detection method based on multi-scale feature fusion and superposition U-Net described in an embodiment of the present application, the preprocessed image is analyzed based on the attention mechanism to generate a region of interest, and a defect image is obtained based on the region of interest, specifically including:

[0016] Acquire a preprocessed image, and segment the preprocessed image into regions to obtain several sub-regions;

[0017] Analyze each sub-region based on the attention mechanism to obtain the sub-region most likely to have defects;

[0018] Assign corresponding weight coefficients to the sub-regions most likely to have defects;

[0019] The sub-regions are weighted based on the weight coefficients to obtain the region of interest.

[0020] Optionally, in the road crack detection method based on multi-scale feature fusion and superposition U-Net described in the embodiment of the present application, semantic extraction of the defect image is performed based on the segmentation network to generate semantic information, and the semantic information is semantically segmented to obtain defect features, specifically including:

[0021] Constructing a segmentation network that integrates multi-scale information, dilated spatial pyramid pooling, and a U-Net model;

[0022] The encoder and decoder framework based on the U-Net model extracts semantic features from defect images through encoder convolution and pooling operations;

[0023] Classify the semantic features to obtain semantic features of multiple categories, and fuse the semantic features of multiple categories to obtain fused features;

[0024] Semantic information is generated based on the fusion features, and the semantic information is segmented according to the segmentation rules to obtain defect features.

[0025] Optionally, in the road crack detection method based on multi-scale feature fusion and superposition U-Net described in the embodiment of the present application, analyzing the road crack status based on defect features to obtain road crack distribution information, and evaluating the road crack distribution information to obtain road crack category information specifically includes:

[0026] Obtaining defect characteristics, analyzing the defect characteristics, and obtaining a road crack status, wherein the road crack status includes a crack location, a crack shape, a crack width, a crack depth, and a crack length;

[0027] Analyze road crack distribution information based on crack location, crack shape, crack width, crack depth and crack length;

[0028] analyzing road crack divergence state information according to road crack distribution information, and evaluating road cracks based on the road crack divergence state information to obtain evaluation information;

[0029] The types of road cracks are analyzed based on the evaluation information to obtain road crack category information.

[0030] Optionally, in the road crack detection method based on multi-scale feature fusion and superposition U-Net described in the embodiment of the present application, establishing a matching maintenance strategy based on road crack category information and transmitting the maintenance strategy to the terminal specifically includes:

[0031] Acquiring road crack category information, wherein the road crack category information includes longitudinal cracks, transverse cracks, diagonal cracks, and crocodile cracks;

[0032] Analyze road crack category information based on the maintenance expert database and generate maintenance strategies;

[0033] Analyze the maintenance matching degree based on the maintenance strategy and compare the maintenance matching degree with the set matching degree threshold;

[0034] If the repair matching degree is greater than or equal to the set matching degree threshold, the road cracks are repaired based on the current repair strategy;

[0035] If the maintenance matching degree is less than the set matching degree threshold, the maintenance strategy is adjusted, and the maintenance strategy includes repairing and re-repairing.

[0036] In a second aspect, an embodiment of the present application provides a road crack detection system based on multi-scale feature fusion and superposition U-Net. The system includes: a memory and a processor. The memory includes a program for a road crack detection method based on multi-scale feature fusion and superposition U-Net. When the program for a road crack detection method based on multi-scale feature fusion and superposition U-Net is executed by the processor, the following steps are implemented:

[0037] Acquire a road detection image, and preprocess the road detection image to obtain a preprocessed image;

[0038] Analyze the pre-processed image based on the attention mechanism, generate the region of interest, and obtain the defect image based on the region of interest;

[0039] Based on the segmentation network, semantic extraction is performed on the defect image to generate semantic information, and the semantic information is segmented to obtain defect features;

[0040] Analyze the road crack status based on defect characteristics to obtain road crack distribution information, evaluate the road crack distribution information, and obtain road crack category information;

[0041] A matching maintenance strategy is established based on the road crack category information and transmitted to the terminal.

[0042] Optionally, in the road crack detection system based on multi-scale feature fusion and superposition U-Net described in an embodiment of the present application, obtaining a road detection image and preprocessing the road detection image to obtain a preprocessed image specifically include:

[0043] Acquire a road detection image, and perform translation processing on the road detection image to obtain a translated image;

[0044] Flip the translated image horizontally and vertically to obtain a flipped image;

[0045] The flipped images are cropped and spliced, and gray-valued to obtain grayscale images;

[0046] The grayscale image is enhanced to obtain a preprocessed image.

[0047] Optionally, in the road crack detection system based on multi-scale feature fusion and superposition U-Net described in an embodiment of the present application, the pre-processed image is analyzed based on the attention mechanism to generate a region of interest, and a defect image is obtained based on the region of interest, specifically including:

[0048] Acquire a preprocessed image, and segment the preprocessed image into regions to obtain several sub-regions;

[0049] Analyze each sub-region based on the attention mechanism to obtain the sub-region most likely to have defects;

[0050] Assign corresponding weight coefficients to the sub-regions most likely to have defects;

[0051] The sub-regions are weighted based on the weight coefficients to obtain the region of interest.

[0052] In a third aspect, an embodiment of the present application also provides a computer-readable storage medium, which includes a road crack detection method program based on multi-scale feature fusion and superposition U-Net. When the road crack detection method program based on multi-scale feature fusion and superposition U-Net is executed by a processor, the steps of the road crack detection method based on multi-scale feature fusion and superposition U-Net as described in any one of the above items are implemented.

[0053] From the above, it can be seen that the embodiment of the present application provides a road crack detection method, system and medium based on multi-scale feature fusion and superposition U-Net, which obtains a road detection image, preprocesses the road detection image, and obtains a preprocessed image; analyzes the preprocessed image based on the attention mechanism to generate a region of interest, and obtains a defect image based on the region of interest; semantically extracts the defect image based on the segmentation network to generate semantic information, and semantically segments the semantic information to obtain defect features; analyzes the road crack status based on the defect features to obtain road crack distribution information, evaluates the road crack distribution information, and obtains road crack category information; establishes a matching maintenance strategy based on the road crack category information, and transmits the maintenance strategy to the terminal; accurately analyzes the region of interest through the attention mechanism, efficiently obtains the defect image, and semantically segments the defect image through the segmentation network to accurately analyze the road defect features, thereby improving the defect recognition capability and recognition efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 Flowchart of the road crack detection method based on multi-scale feature fusion and superposition U-Net provided in an embodiment of the present application;

[0056] Figure 2 A flowchart of a road detection image preprocessing method for a road crack detection method based on multi-scale feature fusion and superposition U-Net provided in an embodiment of the present application;

[0057] Figure 3 Flowchart of the defect image analysis method for the road crack detection method based on multi-scale feature fusion and superposition U-Net provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of the present application.

[0059] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0060] Please refer to Figure 1 , Figure 1 This is a flowchart of a road crack detection method based on multi-scale feature fusion and superposition of U-Net in some embodiments of the present application. The road crack detection method based on multi-scale feature fusion and superposition of U-Net is used in a terminal device and includes the following steps:

[0061] S101, acquiring a road detection image, and preprocessing the road detection image to obtain a preprocessed image;

[0062] S102, analyzing the preprocessed image based on the attention mechanism, generating a region of interest, and obtaining a defect image based on the region of interest;

[0063] S103, performing semantic extraction on the defect image based on the segmentation network to generate semantic information, and performing semantic segmentation on the semantic information to obtain defect features;

[0064] S104, analyzing the road crack status based on the defect characteristics to obtain road crack distribution information, and evaluating the road crack distribution information to obtain road crack category information;

[0065] S105: Establish a matching maintenance strategy based on the road crack category information, and transmit the maintenance strategy to the terminal.

[0066] It should be noted that the defect areas in the image are enhanced through the attention mechanism, thereby improving the efficiency of defect detection, and the defect image is semantically segmented through the segmentation network to accurately analyze the defect features and improve the accuracy of defect detection.

[0067] Please refer to Figure 2 , Figure 2 This is a flowchart of a road detection image preprocessing method based on a road crack detection method based on multi-scale feature fusion and superposition U-Net in some embodiments of the present application. According to an embodiment of the present invention, obtaining a road detection image and preprocessing the road detection image to obtain a preprocessed image specifically includes:

[0068] S201, acquiring a road detection image, and performing translation processing on the road detection image to obtain a translated image;

[0069] S202, flipping the translated image horizontally and vertically to obtain a flipped image;

[0070] S203, cropping and splicing the flipped image, and performing gray value processing to obtain a grayscale image;

[0071] S204: Perform enhancement processing on the grayscale image to obtain a pre-processed image.

[0072] It should be noted that by translating, flipping, cropping and splicing the road detection image, the defects in the road detection image are optimized, the road cracks are displayed more clearly, and the defect detection accuracy is improved.

[0073] Please refer to Figure 3 , Figure 3 This is a flowchart of a defect image analysis method for a road crack detection method based on multi-scale feature fusion and superposition U-Net in some embodiments of the present application. According to an embodiment of the present invention, a pre-processed image is analyzed based on an attention mechanism to generate a region of interest, and a defect image is obtained based on the region of interest, specifically including:

[0074] S301, obtaining a pre-processed image, and performing region segmentation on the pre-processed image to obtain a plurality of sub-regions;

[0075] S302, analyzing each sub-region based on the attention mechanism to obtain the sub-region most likely to have defects;

[0076] S303, assigning corresponding weight coefficients to the sub-regions most likely to have defects;

[0077] S304: weighting the sub-regions based on the weight coefficients to obtain a region of interest.

[0078] It should be noted that the weight coefficients of the sub-regions are assigned through the attention mechanism, so as to accurately obtain the region of interest and improve the efficiency of defect detection.

[0079] According to an embodiment of the present invention, semantic extraction is performed on the defect image based on a segmentation network to generate semantic information, and the semantic information is semantically segmented to obtain defect features, specifically including:

[0080] Build a segmentation network that integrates multi-scale information, dilated spatial pyramid pooling, and the U-Net model;

[0081] The encoder and decoder framework based on the U-Net model extracts semantic features from defect images through encoder convolution and pooling operations;

[0082] Classify the semantic features to obtain semantic features of multiple categories, and fuse the semantic features of multiple categories to obtain fused features;

[0083] Semantic information is generated based on the fusion features, and the semantic information is segmented according to the segmentation rules to obtain defect features.

[0084] It should be noted that the semantic features are extracted and classified through the segmentation network, and the extracted features are then fused and analyzed to accurately analyze road defects.

[0085] According to an embodiment of the present invention, the road crack status is analyzed based on defect characteristics to obtain road crack distribution information, and the road crack distribution information is evaluated to obtain road crack category information, specifically including:

[0086] Obtain defect characteristics, analyze the defect characteristics, and obtain road crack status, which includes crack location, crack shape, crack width, crack depth, and crack length;

[0087] Analyze road crack distribution information based on crack location, crack shape, crack width, crack depth and crack length;

[0088] analyzing road crack divergence state information according to road crack distribution information, and evaluating road cracks based on the road crack divergence state information to obtain evaluation information;

[0089] The types of road cracks are analyzed based on the evaluation information to obtain road crack category information.

[0090] It should be noted that the crack position and shape are analyzed according to the defect characteristics, and the crack divergence state is evaluated, so as to analyze the safety impact of the cracks and classify the cracks, providing an effective basis for subsequent crack repair.

[0091] According to an embodiment of the present invention, establishing a matching maintenance strategy based on road crack category information and transmitting the maintenance strategy to a terminal specifically includes:

[0092] Obtain road crack category information, including longitudinal cracks, transverse cracks, diagonal cracks, and crocodile cracks;

[0093] Analyze road crack category information based on the maintenance expert database and generate maintenance strategies;

[0094] Analyze the maintenance matching degree based on the maintenance strategy and compare the maintenance matching degree with the set matching degree threshold;

[0095] If the repair matching degree is greater than or equal to the set matching degree threshold, the road cracks are repaired based on the current repair strategy;

[0096] If the maintenance matching degree is less than the set matching degree threshold, the maintenance strategy is adjusted, and the maintenance strategy includes repairing and re-repairing.

[0097] It should be noted that by analyzing the types of cracks and establishing maintenance strategies, and by analyzing the repair effects of the maintenance strategies on the cracks, the maintenance strategies are continuously adjusted to ensure that the cracks can be repaired accurately and improve road safety.

[0098] In a second aspect, an embodiment of the present application provides a road crack detection system based on multi-scale feature fusion and superposition U-Net. The system includes: a memory and a processor. The memory includes a program for a road crack detection method based on multi-scale feature fusion and superposition U-Net. When the program for the road crack detection method based on multi-scale feature fusion and superposition U-Net is executed by the processor, the following steps are implemented:

[0099] Acquire a road detection image, and preprocess the road detection image to obtain a preprocessed image;

[0100] Analyze the pre-processed image based on the attention mechanism, generate the region of interest, and obtain the defect image based on the region of interest;

[0101] Based on the segmentation network, semantic extraction is performed on the defect image to generate semantic information, and the semantic information is segmented to obtain defect features;

[0102] Analyze the road crack status based on defect characteristics to obtain road crack distribution information, evaluate the road crack distribution information, and obtain road crack category information;

[0103] A matching maintenance strategy is established based on the road crack category information and transmitted to the terminal.

[0104] It should be noted that the defect areas in the image are enhanced through the attention mechanism, thereby improving the efficiency of defect detection, and the defect image is semantically segmented through the segmentation network to accurately analyze the defect features and improve the accuracy of defect detection.

[0105] According to an embodiment of the present invention, obtaining a road detection image and preprocessing the road detection image to obtain a preprocessed image specifically includes:

[0106] Acquire a road detection image, and perform translation processing on the road detection image to obtain a translated image;

[0107] Flip the translated image horizontally and vertically to obtain a flipped image;

[0108] The flipped images are cropped and spliced, and gray-valued to obtain grayscale images;

[0109] The grayscale image is enhanced to obtain a preprocessed image.

[0110] It should be noted that by translating, flipping, cropping and splicing the road detection image, the defects in the road detection image are optimized, the road cracks are displayed more clearly, and the defect detection accuracy is improved.

[0111] According to an embodiment of the present invention, the preprocessed image is analyzed based on the attention mechanism to generate a region of interest, and a defect image is obtained based on the region of interest, specifically including:

[0112] Acquire a preprocessed image, and segment the preprocessed image into regions to obtain several sub-regions;

[0113] Analyze each sub-region based on the attention mechanism to obtain the sub-region most likely to have defects;

[0114] Assign corresponding weight coefficients to the sub-regions most likely to have defects;

[0115] The sub-regions are weighted based on the weight coefficients to obtain the region of interest.

[0116] It should be noted that the weight coefficients of the sub-regions are assigned through the attention mechanism, so as to accurately obtain the region of interest and improve the efficiency of defect detection.

[0117] According to an embodiment of the present invention, semantic extraction is performed on the defect image based on a segmentation network to generate semantic information, and the semantic information is semantically segmented to obtain defect features, specifically including:

[0118] Build a segmentation network that integrates multi-scale information, dilated spatial pyramid pooling, and the U-Net model;

[0119] The encoder and decoder framework based on the U-Net model extracts semantic features from defect images through encoder convolution and pooling operations;

[0120] Classify the semantic features to obtain semantic features of multiple categories, and fuse the semantic features of multiple categories to obtain fused features;

[0121] Semantic information is generated based on the fusion features, and the semantic information is segmented according to the segmentation rules to obtain defect features.

[0122] It should be noted that the semantic features are extracted and classified through the segmentation network, and the extracted features are then fused and analyzed to accurately analyze road defects.

[0123] According to an embodiment of the present invention, the road crack status is analyzed based on defect characteristics to obtain road crack distribution information, and the road crack distribution information is evaluated to obtain road crack category information, specifically including:

[0124] Obtain defect characteristics, analyze the defect characteristics, and obtain road crack status, which includes crack location, crack shape, crack width, crack depth, and crack length;

[0125] Analyze road crack distribution information based on crack location, crack shape, crack width, crack depth and crack length;

[0126] analyzing road crack divergence state information according to road crack distribution information, and evaluating road cracks based on the road crack divergence state information to obtain evaluation information;

[0127] The types of road cracks are analyzed based on the evaluation information to obtain road crack category information.

[0128] It should be noted that the crack position and shape are analyzed according to the defect characteristics, and the crack divergence state is evaluated, so as to analyze the safety impact of the cracks and classify the cracks, providing an effective basis for subsequent crack repair.

[0129] According to an embodiment of the present invention, establishing a matching maintenance strategy based on road crack category information and transmitting the maintenance strategy to a terminal specifically includes:

[0130] Obtain road crack category information, including longitudinal cracks, transverse cracks, diagonal cracks, and crocodile cracks;

[0131] Analyze road crack category information based on the maintenance expert database and generate maintenance strategies;

[0132] Analyze the maintenance matching degree based on the maintenance strategy and compare the maintenance matching degree with the set matching degree threshold;

[0133] If the repair matching degree is greater than or equal to the set matching degree threshold, the road cracks are repaired based on the current repair strategy;

[0134] If the maintenance matching degree is less than the set matching degree threshold, the maintenance strategy is adjusted, and the maintenance strategy includes repairing and re-repairing.

[0135] It should be noted that by analyzing the types of cracks and establishing maintenance strategies, and by analyzing the repair effects of the maintenance strategies on the cracks, the maintenance strategies are continuously adjusted to ensure that the cracks can be repaired accurately and improve road safety.

[0136] A third aspect of the present invention provides a computer-readable storage medium, which includes a road crack detection method program based on multi-scale feature fusion and superposition U-Net. When the road crack detection method program based on multi-scale feature fusion and superposition U-Net is executed by a processor, the steps of the road crack detection method based on multi-scale feature fusion and superposition U-Net as described above are implemented.

[0137] The present invention discloses a road crack detection method, system and medium based on multi-scale feature fusion and superposition U-Net. The method comprises the following steps: obtaining a road detection image, preprocessing the road detection image to obtain a preprocessed image; analyzing the preprocessed image based on an attention mechanism to generate a region of interest, and obtaining a defect image based on the region of interest; semantically extracting the defect image based on a segmentation network to generate semantic information, and semantically segmenting the semantic information to obtain defect features; analyzing the road crack state based on the defect features to obtain road crack distribution information, and evaluating the road crack distribution information to obtain road crack category information; establishing a matching maintenance strategy based on the road crack category information, and transmitting the maintenance strategy to a terminal; accurately analyzing the region of interest through an attention mechanism to efficiently obtain a defect image, and semantically segmenting the defect image through a segmentation network to accurately analyze road defect features, thereby improving defect recognition capability and efficiency.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0139] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0140] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0141] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0142] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention, or the part that contributes to the existing technology, can be embodied in the form of a software product. The software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program code, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

Claims

1. A road crack detection method based on multi-scale feature fusion and superposition U-Net, characterized in that: include: Acquire a road detection image, and preprocess the road detection image to obtain a preprocessed image; Analyze the pre-processed image based on the attention mechanism, generate the region of interest, and obtain the defect image based on the region of interest; Based on the segmentation network, semantic extraction is performed on the defect image to generate semantic information, and the semantic information is segmented to obtain defect features; Analyze the road crack status based on defect characteristics to obtain road crack distribution information, evaluate the road crack distribution information, and obtain road crack category information; A matching maintenance strategy is established based on the road crack category information and transmitted to the terminal.

2. The road crack detection method based on multi-scale feature fusion and superposition U-Net according to claim 1 is characterized in that: Acquire a road detection image and preprocess the road detection image to obtain a preprocessed image, specifically including: Acquire a road detection image, and perform translation processing on the road detection image to obtain a translated image; Flip the translated image horizontally and vertically to obtain a flipped image; The flipped images are cropped and spliced, and gray-valued to obtain grayscale images; The grayscale image is enhanced to obtain a preprocessed image.

3. The road crack detection method based on multi-scale feature fusion and superposition U-Net according to claim 2 is characterized in that: The pre-processed image is analyzed based on the attention mechanism to generate a region of interest, and a defect image is obtained based on the region of interest, specifically including: Acquire a preprocessed image, and segment the preprocessed image into regions to obtain several sub-regions; Analyze each sub-region based on the attention mechanism to obtain the sub-region most likely to have defects; Assign corresponding weight coefficients to the sub-regions most likely to have defects; The sub-regions are weighted based on the weight coefficients to obtain the region of interest.

4. The road crack detection method based on multi-scale feature fusion and superposition U-Net according to claim 3 is characterized in that: Based on the segmentation network, semantic extraction is performed on the defect image to generate semantic information, and the semantic information is segmented to obtain defect features, including: Constructing a segmentation network that integrates multi-scale information, dilated spatial pyramid pooling, and a U-Net model; The encoder and decoder framework based on the U-Net model extracts semantic features from defect images through encoder convolution and pooling operations; Classify the semantic features to obtain semantic features of multiple categories, and fuse the semantic features of multiple categories to obtain fused features; Semantic information is generated based on the fusion features, and the semantic information is segmented according to the segmentation rules to obtain defect features.

5. The road crack detection method based on multi-scale feature fusion and superposition U-Net according to claim 4 is characterized in that: Analyze the road crack status based on defect characteristics to obtain road crack distribution information, evaluate the road crack distribution information, and obtain road crack category information, including: Obtaining defect characteristics, analyzing the defect characteristics, and obtaining a road crack status, wherein the road crack status includes a crack location, a crack shape, a crack width, a crack depth, and a crack length; Analyze road crack distribution information based on crack location, crack shape, crack width, crack depth and crack length; analyzing road crack divergence state information according to road crack distribution information, and evaluating road cracks based on the road crack divergence state information to obtain evaluation information; The types of road cracks are analyzed based on the evaluation information to obtain road crack category information.

6. The road crack detection method based on multi-scale feature fusion and superposition U-Net according to claim 5 is characterized in that: Establish a matching repair strategy based on road crack category information and transmit the repair strategy to the terminal, including: Acquiring road crack category information, wherein the road crack category information includes longitudinal cracks, transverse cracks, diagonal cracks, and crocodile cracks; Analyze road crack category information based on the maintenance expert database and generate maintenance strategies; Analyze the maintenance matching degree based on the maintenance strategy and compare the maintenance matching degree with the set matching degree threshold; If the repair matching degree is greater than or equal to the set matching degree threshold, the road cracks are repaired based on the current repair strategy; If the maintenance matching degree is less than the set matching degree threshold, the maintenance strategy is adjusted, and the maintenance strategy includes repairing and re-repairing.

7. A road crack detection system based on multi-scale feature fusion and superposition U-Net, characterized by: The system includes: a memory and a processor, wherein the memory includes a program for a road crack detection method based on multi-scale feature fusion and superposition of U-Net, and when the program for the road crack detection method based on multi-scale feature fusion and superposition of U-Net is executed by the processor, the following steps are implemented: Acquire a road detection image, and preprocess the road detection image to obtain a preprocessed image; Analyze the pre-processed image based on the attention mechanism, generate the region of interest, and obtain the defect image based on the region of interest; Based on the segmentation network, semantic extraction is performed on the defect image to generate semantic information, and the semantic information is segmented to obtain defect features; Analyze the road crack status based on defect characteristics to obtain road crack distribution information, evaluate the road crack distribution information, and obtain road crack category information; A matching maintenance strategy is established based on the road crack category information and transmitted to the terminal.

8. The road crack detection system based on multi-scale feature fusion and superposition U-Net according to claim 7 is characterized in that: Acquire a road detection image and preprocess the road detection image to obtain a preprocessed image, specifically including: Acquire a road detection image, and perform translation processing on the road detection image to obtain a translated image; Flip the translated image horizontally and vertically to obtain a flipped image; The flipped images are cropped and spliced, and gray-valued to obtain grayscale images; The grayscale image is enhanced to obtain a preprocessed image.

9. The road crack detection system based on multi-scale feature fusion and superposition U-Net according to claim 8 is characterized in that: The pre-processed image is analyzed based on the attention mechanism to generate a region of interest, and a defect image is obtained based on the region of interest, specifically including: Acquire a preprocessed image, and segment the preprocessed image into regions to obtain several sub-regions; Analyze each sub-region based on the attention mechanism to obtain the sub-region most likely to have defects; Assign corresponding weight coefficients to the sub-regions most likely to have defects; The sub-regions are weighted based on the weight coefficients to obtain the region of interest.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a road crack detection method program based on multi-scale feature fusion and superposition U-Net. When the road crack detection method program based on multi-scale feature fusion and superposition U-Net is executed by a processor, the steps of the road crack detection method based on multi-scale feature fusion and superposition U-Net are implemented as described in any one of claims 1 to 6.