A Bridge Disease Detection and Maintenance Method and System Based on the Yolo Algorithm

Abstract

The present application relates to a bridge defect detection and maintenance method and system based on the Yolo algorithm, which detects the defect type of the target detection information based on a bridge multi-disease detection model and a bridge defect detection knowledge base, as well as real-time screening rules and dynamic thresholds; the real-time screening rules can be used to distinguish the defects of steel-concrete structure, concrete structure, and steel structure bridges, and according to the different types of bridges, the defects that do not exist in this type of bridge are pre-eliminated to achieve the purpose of reducing interference, thereby increasing the accuracy of recognition; the corresponding dynamic threshold is obtained by pre-judging the defect type, and then the dynamic threshold is used for identification to avoid the redundancy of detection targets and the occurrence of false detection; thereby combining expert knowledge and real-time screening rules and dynamic thresholds for double screening, so that the interference caused by the defect types of bridges made of different materials and the defect types in different locations can be avoided during the identification process.

Description

Technical Field

[0001] This application relates to the technical field of bridge operation and maintenance, and particularly to a bridge disease detection and maintenance method and system based on the Yolo algorithm. Background Art

[0002] Up to now, the bridges completed in the early stage will gradually enter the high-incidence period of diseases. The detection and maintenance business volume is huge and the situation is severe. On the one hand, for the identification, diagnosis, and maintenance treatment decision-making of various diseases of the current bridge structure, most are based on the personal experience of the inspectors. The diagnosis and maintenance decision-making experience is stored in their respective brains, and it is difficult to utilize and share the mature and professional bridge management and maintenance experience. On the other hand, although various technologies in the field of artificial intelligence, especially the application of computer vision in bridge intelligent management and maintenance, can help improve the automation and intelligence level of bridge management and maintenance, the current application of this technology in bridge management and maintenance is relatively rough.

[0003] In some related technologies, Yolo is a classic and mature algorithm for object detection and recognition in the field of computer vision. Many experts and scholars have done relevant research and application implementation on the bridge disease detection and recognition technology based on Yolo. However, the current research and application have the following problems:

[0004] (1) The link of using expert experience is only in the stage of bridge disease sample annotation, that is, the stage of making the dataset samples. After the dataset samples are made, the Yolo algorithm mines and learns the potential features of disease instances in the dataset samples to train a deep learning model suitable for bridge disease detection and recognition, and screens the recognized target results through an important parameter, the confidence score. However, there are often some errors in the screening results because the disease types that may occur at the same position of bridges made of different materials may be different, and the disease types that may occur in the components at different positions of bridges made of the same material may be different, causing certain interference.

[0005] (2) The setting of the confidence score is to set a unified threshold for the multi-disease model, and there is no application in the field of bridge disease detection to set thresholds for each disease type separately. The setting of the confidence score depends to a large extent on the tests and empirical values of the algorithm engineers and bridge engineers of this project.

[0006] (3) The cause analysis and maintenance countermeasures of bridge diseases rely heavily on the professional knowledge and experience reserve of the bridge inspection engineers of this project. It is difficult to ensure relatively appropriate cause analysis conclusions and relatively appropriate maintenance methods for the various and complexly characterized bridge diseases. Summary of the Invention

[0007] An embodiment of the present application provides a bridge disease detection and maintenance method and system based on the Yolo algorithm to solve the problem of recognition errors caused by interference brought about by different bridge materials and different positions of components to which diseases belong in the related art.

[0008] In a first aspect, a bridge disease detection and maintenance method based on the Yolo algorithm is provided, which includes:

[0009] Based on the bridge multi-disease detection model and the bridge disease detection knowledge base, obtain the real-time screening rules for target detection information; the target detection information includes video stream information and picture information;

[0010] Based on the bridge multi-disease detection model and the bridge disease detection knowledge base, and obtain the dynamic threshold for target detection information under the real-time screening rules;

[0011] Based on the bridge multi-disease detection model, the bridge disease detection knowledge base, and the real-time screening rules and dynamic threshold, identify the disease types of the target detection information.

[0012] In some embodiments, obtaining the real-time screening rules for target detection information based on the bridge multi-disease detection model and the bridge disease detection knowledge base includes the following steps:

[0013] Based on the bridge multi-disease detection model, judge the bridge material type and the component to which the disease belongs in the target detection information;

[0014] Derive the real-time screening rules according to the bridge material type and the component to which the disease belongs; the real-time screening rules are: exclude the disease types in the bridge disease detection knowledge base that are not relevant to the bridge material type and the component to which the disease belongs.

[0015] In some embodiments, obtaining the dynamic threshold for target detection information based on the bridge multi-disease detection model and the bridge disease detection knowledge base and under the real-time screening rules includes the following steps:

[0016] Based on the bridge multi-disease detection model, judge the corresponding bridge material type, the component to which the disease belongs, and the disease location in the target detection information;

[0017] According to the bridge material type, the component to which the disease belongs, and the disease location, and in combination with the bridge disease detection knowledge base, pre-judge the corresponding disease type;

[0018] Increase or decrease the fluctuation value on the basis of the reference confidence score to obtain the actual confidence score corresponding to the pre-judged disease type;

[0019] Take the actual confidence score as the dynamic threshold.

[0020] In some embodiments, constructing the bridge disease detection knowledge base includes the following steps:

[0021] Sort and classify the target data according to bridge material type, component to which the disease belongs, disease location, disease type, disease spatial distribution and trend, and disease pictures;

[0022] The target data includes data from existing bridge regular inspection platforms and network data.

[0023] In some embodiments, after detecting the disease type of the target detection information, the following steps are further included:

[0024] Generate an inspection and maintenance report according to the disease type of the target detection information, and in combination with the bridge disease detection knowledge base and the bridge disease maintenance knowledge base.

[0025] In some embodiments, sort and classify the expert maintenance knowledge data according to disease type, disease cause analysis, and disease maintenance countermeasures to construct the bridge disease maintenance knowledge base.

[0026] In some embodiments, generating a disease cause analysis report and maintenance countermeasures according to the disease type of the target detection information, and in combination with the bridge disease detection knowledge base and the bridge disease maintenance knowledge base includes the following steps:

[0027] Classify and sort the professional terms of all bridge diseases, and establish a bridge disease feature word library in combination with the bridge disease detection knowledge base and the bridge disease maintenance knowledge base;

[0028] Associate the bridge disease feature word library, the bridge disease detection knowledge base, and the bridge disease maintenance knowledge base;

[0029] Extract the feature words of the disease type of the target detection information to fuzzy match the inspection and maintenance report corresponding to the extracted feature words; the inspection and maintenance report includes bridge material type, disease location, component to which the disease belongs, disease type, cause analysis report, and best maintenance countermeasures.

[0030] In some embodiments, based on the bridge disease detection knowledge base, combine expert knowledge data to label and produce a multi-disease sample data set applicable to the Yolo algorithm, and update and iteratively train the bridge multi-disease detection model.

[0031] In a second aspect, a bridge disease detection and maintenance system based on the Yolo algorithm is provided, which includes:

[0032] The first module is used to obtain real-time screening rules for target detection information based on the bridge multi-disease detection model and the bridge disease detection knowledge base;

[0033] A second module, which is used to obtain a dynamic threshold for target detection information based on a bridge multi-disease detection model and a bridge disease detection knowledge base and under the real-time screening rule;

[0034] A third module, which is used to detect the disease type of the target detection information based on a bridge multi-disease detection model, a bridge disease detection knowledge base, the above real-time screening rule, and the dynamic threshold.

[0035] In a third aspect, a computer-readable storage medium is provided. A bridge disease detection and maintenance program based on the Yolo algorithm is stored on the computer-readable storage medium. When the bridge disease detection and maintenance program based on the Yolo algorithm is executed by a processor, the steps of the bridge disease detection and maintenance method based on the Yolo algorithm are implemented.

[0036] The beneficial effects brought by the technical solution provided in this application include:

[0037] The embodiments of this application provide a bridge disease detection and maintenance method and system based on the Yolo algorithm. Through the real-time screening rule, the diseases of steel-concrete structures, concrete structures, and steel structures can be distinguished. According to the different types of bridges, the diseases that do not exist in this type of bridge can be excluded in advance to achieve the purpose of reducing interference, thereby increasing the recognition accuracy; by pre-judging the disease type to obtain the corresponding dynamic threshold, and then using the dynamic threshold for recognition can avoid the occurrence of redundant detection targets and misdetection; thus, combining expert knowledge, as well as real-time screening rules and dynamic thresholds for double screening, so as to avoid the interference caused by the disease types of different material bridges and the disease types of different positions during the recognition process; at the same time, with the support of expert knowledge, a relatively high level of cause analysis and maintenance suggestions can be realized.

[0038] Description of the Drawings

[0039] In order to more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0040] Figure 1 It is a general flowchart of bridge disease detection and maintenance based on the Yolo algorithm provided by the embodiments of this application;

[0041] Figure 2 It is a schematic diagram of a bridge disease detection and maintenance system based on the Yolo algorithm provided by the embodiments of this application. Detailed Embodiments

[0042] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are some, but not all, of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts fall within the scope of protection of this application.

[0043] The embodiments of this application provide a method for detecting and maintaining bridge diseases based on the Yolo algorithm to solve the problem of recognition errors caused by interference due to different bridge materials and different positions of components to which diseases belong in related technologies.

[0044] A method for detecting bridge diseases based on the Yolo algorithm and combined with expert knowledge, characterized in that it includes:

[0045] Step 100: Based on the bridge multi-disease detection model and the bridge disease detection knowledge base, obtain real-time screening rules for target detection information; the target detection information includes video stream information and picture information;

[0046] Step 101: Based on the bridge multi-disease detection model and the bridge disease detection knowledge base, and obtain dynamic thresholds for target detection information under the real-time screening rules;

[0047] Step 102: Based on the bridge multi-disease detection model, the bridge disease detection knowledge base, and the above real-time screening rules and dynamic thresholds, detect the disease types of the target detection information.

[0048] Through the real-time screening rules, diseases of steel-concrete structures, concrete structures, and steel structure bridges can be distinguished. According to the different bridge types, diseases that do not exist in this type of bridge can be excluded in advance to achieve the purpose of reducing interference, thereby increasing the recognition accuracy; by judging the pre-judged disease types to obtain corresponding dynamic thresholds, and then using the dynamic thresholds for recognition can avoid the occurrence of redundant detection targets and misdetection situations; thus, combined with expert knowledge, as well as real-time screening rules and dynamic thresholds for double screening, to avoid interference caused by disease types of bridges with different materials and disease types at different positions during the recognition process.

[0049] In some preferred embodiments, step 100 specifically includes the following steps:

[0050] Based on the bridge multi-disease detection model, judge the bridge material type and the component to which the disease belongs in the target detection information;

[0051] Derive real-time screening rules based on the bridge material type and the component to which the disease belongs; the real-time screening rules are: exclude the disease types in the bridge disease detection knowledge base that are not relevant to the bridge material type and the component to which the disease belongs; thus, the interference items of diseases with similar characteristics can be excluded in advance.

[0052] It has the following advantages:

[0053] For example, cracks are very common bridge diseases, such as cracks at the bottom of the beams and side webs of the upper structure of concrete structures, cracks in the over-welded holes of the diagonal web members of steel box girders, cracks in the piers and abutments of the lower structure, and cracks in the bearings. Cracks have highly similar characteristics. It is difficult for a simple disease detection method based on Yolo to distinguish different crack diseases. Therefore, screening rules need to be used for screening to exclude the interference items of diseases with similar characteristics. The following takes crack diseases as an example to illustrate the detection method of this patent:

[0054] Use the video stream transmitted by the camera to detect crack diseases in real time.

[0055] Step 10001: The detection part of the video stream is the 'upper structure', and the prediction categories of cracks in the 'piers and abutments' and 'bearings' in the multi-disease detection model based on Yolo are turned off;

[0056] Step 10002: If the video stream detects a 'concrete structure bridge', turn off the prediction category of cracks in the 'upper structure - steel structure bridge' in the multi-disease detection model based on Yolo;

[0057] Step 10003: If the video stream detects a'steel structure bridge', turn off the prediction category of cracks in the 'upper structure - concrete structure bridge' in the multi-disease detection model based on Yolo;

[0058] Step 10004: The detection part of the video stream is the 'piers and abutments', and turn off the prediction categories of cracks in the 'upper structure' and 'bearings' in the multi-disease detection model based on Yolo;

[0059] Step 10005: The detection part of the video stream is the 'bearing', and turn off the prediction categories of cracks in the 'upper structure' and 'piers and abutments' in the multi-disease detection model based on Yolo.

[0060] The above real-time screening rules are illustrated by the above embodiments and will change accordingly according to different video information. In some preferred embodiments, step 101 specifically includes the following steps:

[0061] Based on the bridge multi-disease detection model, judge the corresponding bridge material type, the component to which the disease belongs, and the disease location in the target detection information;

[0062] Pre-judge the corresponding disease type according to the bridge material type, the component to which the disease belongs, and the disease location, and in combination with the bridge disease detection knowledge base;

[0063] Increase or decrease the fluctuation value based on the reference confidence score to obtain the actual confidence score corresponding to the pre-judged disease type; after the Yolo model training is completed, a curve will be generated, which shows the decimal of all classes at, and this decimal is the reference confidence score.

[0064] Take the actual confidence score as the dynamic threshold.

[0065] The above steps improve the disease recognition rate by appropriately adjusting the confidence score of the corresponding type of disease at this position. The following specific embodiments are used for illustration.

[0066] For the upper structure part of the concrete structure bridge, some cracks are narrow in width and have a small contrast with the surrounding environment, resulting in the difficulty of detecting them by the disease detection algorithm based on the video stream under the setting of a unified confidence threshold. And setting a unified low confidence score will lead to redundant detection targets and misdetection. In order to detect all corresponding diseases as much as possible and minimize the misdetection situation, the multi-disease detection model based on Yolo should combine the bridge detection knowledge base to pre-determine the possible disease types of the current bridge, current component and location, and improve the accuracy of disease detection and recognition by appropriately adjusting the confidence score of the corresponding type of disease at this position. This method can be used for setting the confidence scores of various types of diseases of the bridge. The following takes the cracks in the upper structure of the concrete structure bridge as an example to illustrate the method of this patent for adjusting the confidence score to improve the accuracy of disease detection and recognition:

[0067] For example: Real-time detection of cracks in the upper structure of a concrete structure bridge using the video stream transmitted by the camera,

[0068] Step 10101: The video stream detection part is the "mid-span". Since "transverse cracks" are likely to appear at the mid-span and the harm of this crack is relatively high, the confidence score of "transverse cracks" can be appropriately reduced at this time; since "longitudinal cracks" are difficult to appear at the mid-span and the harm of this crack is relatively low, the confidence score of "longitudinal cracks" can be appropriately increased at this time

[0069] Step 10102: The video stream detection part is the "web". Since "diagonal cracks" are likely to appear in the web and the harm of this type of crack is relatively high, the confidence score of "web flexure-shear diagonal cracks" can be appropriately reduced at this time.

[0070] The recognition rate of diseases is improved by setting corresponding confidence scores at different positions. Among them, on the basis of the standard confidence score, the fluctuation value is increased or decreased, and the fluctuation value also changes continuously according to the pre-judged disease type. Of course, for each pre-judged disease type, we can preset it according to the experience of experts in advance.

[0071] In some preferred embodiments, constructing the bridge disease detection knowledge base includes the following steps: sorting and classifying the target data according to bridge material type, component to which the disease belongs, disease location, disease type, disease spatial distribution and trend, and disease pictures; the target data includes data from existing bridge regular inspection platforms and network data.

[0072] Sort and classify the expert maintenance knowledge data according to disease type, disease cause analysis and disease maintenance countermeasures to construct the bridge disease maintenance knowledge base.

[0073] Based on the bridge disease detection knowledge base, combined with expert knowledge data, mark and produce a multi-disease sample data set suitable for the Yolo algorithm, and update and iteratively train the bridge multi-disease detection model. How to train the model is not the focus of this application. Instead, it is the establishment of the bridge disease detection knowledge base and how to cooperate with the bridge multi-disease detection model to screen and exclude interference items and improve the accuracy.

[0074] You can refer to the following table to understand the bridge disease detection knowledge base

[0075]

[0076]

[0077]

[0078] In some preferred embodiments, step 102, after detecting the disease type of the target detection information, further includes the following steps:

[0079] Generate an inspection and maintenance report according to the disease type of the target detection information, in combination with the bridge disease detection knowledge base and the bridge disease maintenance knowledge base. This step includes:

[0080] Step 10200: Classify and sort the professional terms of all bridge diseases, and establish a bridge disease feature word bank in combination with the bridge disease detection knowledge base and the bridge disease maintenance knowledge base;

[0081] Step 10201: Associate the bridge disease feature word bank, the bridge disease detection knowledge base and the bridge disease maintenance knowledge base;

[0082] Step 10202: Extract feature words for the disease types in the target detection information to fuzzily match the inspection and maintenance reports corresponding to the extracted feature words; the inspection and maintenance reports include bridge material types, disease locations, components to which the diseases belong, disease types, cause analysis reports, and optimal maintenance countermeasures.

[0083] The bridge disease maintenance knowledge base can refer to the following table:

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090] Specifically:

[0091] Through the analysis and classification of professional words related to bridge diseases, a bridge disease feature word library is established on the basis of the bridge disease detection knowledge base and the bridge disease maintenance knowledge base, and the accuracy of the feature word library is reviewed by senior industry experts to ensure the logical rationality and professionalism of the data.

[0092] After associating the complete bridge disease detection knowledge base, bridge disease maintenance knowledge base, and bridge disease feature word library, it is necessary to extract the feature words of the diseases for fuzzy matching of the diseases.

[0093] After identifying the disease type and storing the information, based on the associated bridge disease detection knowledge base, bridge disease maintenance knowledge base, and bridge disease feature word library, the editing clustering algorithm can be used to fuzzily match the disease causes and recommended treatment plans.

[0094] That is, through the fuzzy matching of disease feature words, the user can input a small number of disease feature words at the user end to complete the disease cause analysis and maintenance countermeasures, and finally form a complete description of "this disease is the 'disease name' located in the 'bridge material type', 'bridge component', and 'location'. The possible cause of the disease is 'disease cause', and the recommended maintenance measure is'maintenance countermeasure'" to form an inspection and maintenance report.

[0095] It should be understood from the above description that taking typical highway bridge diseases and steel structure cable-stayed bridges as examples, this intelligent bridge disease detection method based on Yolo combined with expert knowledge is illustrated. The confidence score currently highly depends on the testing and experience of algorithm R & D personnel. To make more full use of the expert knowledge of bridge diseases, based on bridge detection knowledge such as bridge material type, subordinate sub-components, disease spatial distribution and trend characteristics, the disease type is predicted in advance to improve the accuracy of disease identification; at the same time, after completing disease detection, by associating with the bridge maintenance knowledge base, the cause analysis of bridge diseases and recommended treatment strategies are realized.

[0096] According to the bridge material type, diseases of steel-concrete structures, concrete structures, and steel structure bridges can be distinguished. Although it is a unified multi-disease recognition model, according to the different bridge types, diseases that do not exist in this type of bridge can be excluded in advance. By reducing the prediction options, the purpose of reducing interference is achieved, thereby increasing the recognition accuracy.

[0097] According to the different bridge components where the diseases are located, based on expert experience, the diseases that are likely to occur in this component can be judged. For example, transverse cracks are likely to occur at the bottom of concrete beams. At this time, the confidence score can be reduced; cracks: both bearing cracks and concrete cracks are cracks, but because the crack type can be judged in advance according to the component where the disease is located, misjudgment can be avoided.

[0098] A bridge disease detection and maintenance system based on the Yolo algorithm is also proposed, which includes:

[0099] The first module is used to obtain real-time screening rules for target detection information based on the bridge multi-disease detection model and the bridge disease detection knowledge base;

[0100] The second module is used to obtain dynamic thresholds for target detection information based on the bridge multi-disease detection model and the bridge disease detection knowledge base and under the real-time screening rules;

[0101] The third module is used to detect the disease type of the target detection information based on the bridge multi-disease detection model, the bridge disease detection knowledge base, and the above real-time screening rules and dynamic thresholds. The structural block diagram of the bridge disease detection and maintenance system based on the Yolo algorithm can be referred to Figure 2 as shown Figure 2 which shows the flow of the entire data.

[0102] A computer-readable storage medium is also proposed. A bridge disease detection and maintenance program based on the Yolo algorithm is stored on the computer-readable storage medium. When the bridge disease detection and maintenance program based on the Yolo algorithm is executed by a processor, the steps of the bridge disease detection and maintenance method based on the Yolo algorithm are realized.

[0103] In the embodiments of the present application, the bridge disease detection and maintenance device based on the Yolo algorithm may include a processor, a memory, a communication interface, and a communication bus.

[0104] Among them, the communication bus can be of any type and is used to interconnect the processor, the memory, and the communication interface.

[0105] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces, etc., which are used to interconnect the components inside the bridge disease detection and maintenance device based on the Yolo algorithm, as well as interfaces for interconnecting the bridge disease detection and maintenance device based on the Yolo algorithm with other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, a fiber optic interface, an ATM interface, etc.; the user device can be a display, a keyboard, etc.

[0106] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical memory, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0107] The processor can be a general-purpose processor, which can call the bridge disease detection and maintenance program based on the Yolo algorithm stored in the memory and execute the bridge disease detection and maintenance method provided in the embodiments of the present application. For example, the general-purpose processor can be a central processing unit (CPU). Among them, the method executed when the bridge disease detection and maintenance program based on the Yolo algorithm is called can refer to the various embodiments of the bridge disease detection and maintenance method based on the Yolo algorithm in the present application, which will not be elaborated here.

[0108] In the description of the present application, it should be noted that the orientation or positional relationship indicated by terms such as "upper" and "lower" is based on the orientation or positional relationship shown in the drawings. This is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present application. Unless otherwise clearly specified and defined, the terms "installed", "connected", and "coupled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0109] It should be noted that in the present application, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of another identical element in the process, method, article or device including the said element.

[0110] The above are only specific embodiments of the present application, enabling those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features claimed herein.

Claims

1. A bridge disease detection and maintenance method based on Yolo algorithm, characterized in that: It includes: Based on the bridge multi-disease detection model and bridge disease detection knowledge base, obtain real-time screening rules for target detection information; The target detection information includes video stream information and image information; Based on the bridge multi-disease detection model and the bridge disease detection knowledge base, a dynamic threshold value of target detection information is obtained under the real-time screening rule; Based on the bridge multi-disease detection model and the bridge disease detection knowledge base, and under the real-time screening rule, a dynamic threshold value of target detection information is obtained, including the following steps: Based on the bridge multi-defect detection model, the bridge material type, the component to which the defect belongs and the defect location corresponding to the target detection information are judged; based on the bridge material type, the component to which the defect belongs and the defect location, and in combination with the bridge defect detection knowledge base, the corresponding defect type is pre-judged; based on the reference confidence score, the fluctuation value is increased or decreased to obtain the actual confidence score corresponding to the pre-judged defect type; and the actual confidence score is used as the dynamic threshold; Based on the bridge multi-disease detection model and the bridge disease detection knowledge base, as well as the real-time screening rules and dynamic thresholds, the disease type of the target detection information is identified.

2. The bridge disease detection and maintenance method based on Yolo algorithm as claimed in claim 1 is characterized in that: Based on the bridge multi-disease detection model and the bridge disease detection knowledge base, real-time screening rules for target detection information are obtained, including the following steps: Based on the bridge multi-disease detection model, the bridge material type and the component to which the disease belongs in the target detection information are determined; A real-time screening rule is derived according to the bridge material type and the component to which the disease belongs; the real-time screening rule is: excluding the disease types that are not related to the bridge material type and the component to which the disease belongs in the bridge disease detection knowledge base.

3. The bridge disease detection and maintenance method based on Yolo algorithm as claimed in claim 1 is characterized in that: Constructing the bridge disease detection knowledge base includes the following steps: The target data is sorted and classified according to the bridge material type, the component to which the defect belongs, the defect location, the defect type, the defect spatial distribution and direction, and the defect picture; The target data includes data from existing bridge inspection platforms and network data.

4. The bridge disease detection and maintenance method based on Yolo algorithm as claimed in claim 1 is characterized by: After the disease type of the target detection information is detected, the following steps are also included: According to the defect type of the target detection information, an inspection and maintenance report is generated in combination with the bridge defect detection knowledge base and the bridge defect maintenance knowledge base.

5. The bridge disease detection and maintenance method based on Yolo algorithm as claimed in claim 4 is characterized by: The expert maintenance knowledge data is sorted and classified according to the types of defects, defect cause analysis and defect maintenance countermeasures to construct the bridge defect maintenance knowledge base.

6. The bridge disease detection and maintenance method based on Yolo algorithm as claimed in claim 5 is characterized in that: According to the damage type of the target detection information, and in combination with the bridge damage detection knowledge base and the bridge damage maintenance knowledge base, a damage cause analysis report and maintenance countermeasures are generated, including the following steps: Classify and organize all professional terms for bridge diseases, and establish a bridge disease characteristic vocabulary based on the bridge disease detection knowledge base and bridge disease maintenance knowledge base; Associating the bridge disease feature word library, the bridge disease detection knowledge base and the bridge disease maintenance knowledge base; Feature words are extracted from the defect type of the target detection information, and the maintenance report corresponding to the extracted feature words is fuzzy matched; the maintenance report includes the bridge material type, defect location, component to which the defect belongs, defect type, cause analysis report and optimal maintenance countermeasures.

7. The bridge disease detection and maintenance method based on Yolo algorithm as claimed in claim 1 is characterized in that: Based on the bridge defect detection knowledge base, a multi-disease sample data set suitable for the Yolo algorithm is produced by combining expert knowledge data labeling, and the bridge multi-disease detection model is updated and iteratively trained.

8. A bridge disease detection and maintenance system based on Yolo algorithm, characterized in that: It includes: The first module is used to obtain real-time screening rules for target detection information based on a bridge multi-disease detection model and a bridge disease detection knowledge base; The second module is used to obtain a dynamic threshold value of target detection information based on the bridge multi-disease detection model and the bridge disease detection knowledge base and under the real-time screening rule; Based on the bridge multi-disease detection model and the bridge disease detection knowledge base, and under the real-time screening rule, a dynamic threshold value for target detection information is obtained, including the following steps: based on the bridge multi-disease detection model, the bridge material type, the component to which the disease belongs, and the disease location corresponding to the target detection information are judged; based on the bridge material type, the component to which the disease belongs, and the disease location, and in combination with the bridge disease detection knowledge base, the corresponding disease type is pre-judged; based on the reference confidence score, the fluctuation value is increased or decreased to obtain the actual confidence score corresponding to the pre-judged disease type; and the actual confidence score is used as the dynamic threshold value; The third module is used to detect the disease type of the target detection information based on the bridge multi-disease detection model and the bridge disease detection knowledge base, as well as the above real-time screening rules and dynamic thresholds.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a bridge defect detection and maintenance program based on the Yolo algorithm, wherein when the bridge defect detection and maintenance program based on the Yolo algorithm is executed by a processor, the steps of the bridge defect detection and maintenance method based on the Yolo algorithm as described in any one of claims 1 to 7 are implemented.

Images