Road disease inspection data analysis system based on Internet of Things and artificial intelligence

Abstract

The present invention relates to the technical field of road disease identification, and specifically to a road disease inspection data analysis system based on the Internet of Things and artificial intelligence; the system comprises: a central server, an inspection planning module and a disease identification module; by calling various types of disease sets in the central server, a plurality of inspection paths including main roads are automatically generated, the vehicle flow, road diseases and the like of each path are comprehensively analyzed to obtain a priority index, and the efficiency of inspection equipment is analyzed to obtain an inspection value, and path inspection tasks are allocated accordingly, thereby forming a dynamic planning of path inspection; by performing disease identification on collected road images, the disease type and disease degree in the road image can be accurately extracted, and a quantitative disease evaluation can be provided; then, de-duplication identification is performed on road images with road diseases, and the similarity between a target image and a comparison image can be accurately judged, repeated detection is avoided, and the efficiency and accuracy of disease detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of road hazard identification, and in particular to a road hazard inspection data analysis system based on the Internet of Things and artificial intelligence. Background Art

[0002] Road damage refers to various damages that occur to the road surface during long-term use due to various external factors (such as vehicle load, climate change, construction quality, material aging, etc.). These damages will affect the service life, driving safety and driving comfort of the road and need to be discovered and repaired in a timely manner;

[0003] Traditional traffic and municipal road maintenance inspection work is done by staff driving vehicles to inspect road surface defects and roadside assets according to target inspection roads. When defects are found, they pull over and complete the inspection task by measuring, taking photos, uploading, and filling out inspection forms. They then conduct unified sorting, statistics, and analysis, and arrange subsequent maintenance work. Obviously, this traditional inspection method has low inspection efficiency and low reliability of inspection results, and it is difficult to cope with the current situation of rapid traffic development. Summary of the invention

[0004] The purpose of the present invention is to provide a road disease inspection data analysis system based on the Internet of Things and artificial intelligence to solve the problems mentioned in the above background technology.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A road disease inspection data analysis system based on the Internet of Things and artificial intelligence includes: a central server, an inspection planning module and a disease identification module;

[0006] The central server exchanges data with the inspection equipment through the communication network. The various sensors carried on the inspection equipment conduct inspections along the corresponding paths to collect road images and their corresponding location information in real time, and upload them to the central server for storage through the communication network. In addition, the central server also has a number of disease type sets, each of which stores a number of disease images corresponding to the road disease type, as well as the location information, three characteristic elements, disease parameters, and disease degree values ​​corresponding to each disease image. The central server also communicates with the mobile terminals of each road maintenance personnel, and sends each road disease and its corresponding location information, disease type, etc. to the mobile terminals of the road maintenance personnel. When the maintenance completion is received from the mobile terminal of the road maintenance personnel, it is deleted from the corresponding disease type set.

[0007] The inspection planning module plans inspection routes based on various types of disease sets to guide inspection equipment to conduct inspections and collect road images;

[0008] The damage recognition module extracts features based on the road image and performs damage recognition to determine whether there are road damages, and then removes duplicates from the road image with road damages. The specific process of removing duplicates is as follows:

[0009] Step 1: randomly select one of the road images with road damage as the target image, use each damage image in the corresponding damage type set as the comparison image, retrieve the geometric contours of the target image and the comparison image and find the center point of the geometric contours, overlap the center points of the geometric contours of the target image and the comparison image, and rotate 360° until the overlap area is the largest, then stop rotating, calculate the intersection area and union area under the overlap form, and record them as Q1 and Q2 respectively;

[0010] Step 2: Match the disease parameters corresponding to the target image and the comparison image one by one to obtain several parameter pairs, and perform difference calculation on the parameters of each parameter pair to obtain the parameter difference of each parameter pair; perform mean calculation on the parameter difference of each parameter pair to obtain the mean value of the parameter difference between the target image and the comparison image, and record it as C;

[0011] Retrieve the corresponding position information of the target image and the comparison image, and calculate the spatial distance between the two based on it, and record it as H;

[0012] Step 3: retrieve the damage degree values ​​corresponding to the target image and the comparison image, and record them as S1 and S2 respectively;

[0013] Step 4: Normalize the intersection area Q1, the union area Q2, the parameter difference mean C, the spatial spacing H, the target image damage level S1 and the comparison image damage level S2 and take their values, and calculate the values ​​in a formula to obtain the repetition index SQ. The specific calculation formula is:

[0014]

[0015] Among them, η1, η2, η3, and η4 are respectively set weight constants; thereby, the repetition index between the target image and each image in the corresponding disease type set can be obtained, and the repetition index is compared and analyzed with the set repetition threshold. When the repetition index is greater than the set repetition threshold, a repeated image is accumulated; when the accumulated number of repeated images is greater than or equal to one, it means that this road image has been collected in the previous inspection, and it will be deleted; when the accumulated number of repeated images is zero, the road image with road diseases is recorded as a disease image, and it and its corresponding location information, three characteristic elements, disease parameters and disease degree values ​​are stored in the corresponding disease type set of the central server.

[0016] Preferably, the specific process of disease identification is:

[0017] Preprocess the collected road images, analyze and associate the image location information, so that each road image accurately corresponds to its corresponding location information; extract the geometric features, texture features and color features of the road images to obtain the three feature elements of the road images, where the three features include geometric features, texture features and color features;

[0018] Based on the labeled sample data set, the convolutional neural network is trained to enable it to identify different types of road damage; the three characteristic elements are input into the trained convolutional neural network algorithm model to output the damage type and its corresponding damage parameter; the damage parameter in each damage type is set to correspond to a weighting coefficient denoted as Pi, and each disease parameter of the disease type is denoted as Wi, and the damage degree value S is obtained by weighted calculation according to the corresponding weighting coefficient; the specific calculation formula is:

[0019]

[0020] Wherein i=1,2,3...I, I is a positive integer, I is the total number of disease parameters in a certain disease type, and i is any parameter among the disease parameters; thus, disease identification can be performed on each road image to screen out road images with road diseases, and the disease type and disease degree value corresponding to each road image with road diseases can be output.

[0021] Preferably, the specific process of planning the inspection path based on each type of disease set is as follows:

[0022] Step 1: retrieve a number of damage images corresponding to the road damage types stored in each damage type set, as well as the location information and damage degree value corresponding to each damage image, mark the damage corresponding to each damage image at the corresponding position in the city map, and generate a number of paths according to the path algorithm, so that all road damages are dispersed to each path;

[0023] Step 2: Analyze the saturation degree of the traffic flow on the path to obtain the path saturation value, which is recorded as Bα;

[0024] Step 3: retrieve the number of road diseases on the path, as well as the disease type and disease severity value corresponding to each road disease; set a type base value corresponding to each type of disease; match each road disease on the path with all the set disease types to obtain the corresponding type base value, normalize the type base value and the disease severity value and take their values, perform weighted calculation on the values ​​to obtain the disease risk value of each road disease; sum up the disease risk values ​​of each road disease on the path to obtain the total risk value recorded as F;

[0025] Step 4: Obtain the path length, record it as G, normalize it with the road saturation value and the total risk value, and take its value. Formulate the value to obtain the priority index BF of the path. The specific calculation formula is:

[0026]

[0027] Among them, β1 and β2 are respectively set weight constants; thus, the priority indexes of several paths can be obtained, and the one with the largest priority index is selected as the target path; it is assumed that there are several idle inspection devices, each of which corresponds to an inspection value, and the inspection device with the largest inspection value is selected as the target inspection device, and the target path is assigned to the target inspection device, thereby changing the state of the target inspection device from the idle state to the inspection state; when the target inspection device completes the inspection task of this path, the historical inspection times of the inspection device increase by one, and the state of the target inspection device is changed from the inspection state to the idle state;

[0028] Step 5: Obtain the historical inspection times of each inspection device, and analyze the efficiency of the inspection device based on the times to obtain the inspection value of the inspection device; thereby, the inspection value of each inspection device can be obtained and updated to step 4;

[0029] Step 6: Whenever there is an inspection device in an idle state, repeat steps 2 to 5 until the inspection of all paths is completed.

[0030] Preferably, the specific process of analyzing the saturation degree of the traffic flow on the path is:

[0031] Select any one of the paths, divide the path into several sections, obtain the traffic flow of each section in real time, and compare and analyze it with the set traffic flow interval. When the traffic flow is greater than the maximum value in the set traffic flow interval, the section corresponding to the traffic flow is recorded as a highly saturated section; when the traffic flow is in the set traffic flow interval, the section corresponding to the traffic flow is recorded as a moderately saturated section; when the traffic flow is less than the minimum value in the set traffic flow interval, the section corresponding to the traffic flow is recorded as a low-saturated section; the number of highly saturated sections, moderately saturated sections and low-saturated sections in the path is counted respectively, and they are recorded as B1, B2 and B3 respectively; the traffic flows corresponding to the highly saturated sections, moderately saturated sections and low-saturated sections are summed up to obtain high saturated flow values, medium saturated flow values ​​and low saturated flow values, and they are recorded as B4, B5 and B6 respectively;

[0032] The number of highly saturated sections B1, the number of moderately saturated sections B2, the number of low saturated sections B3, the high saturated flow value B4, the medium saturated flow value B5 and the low saturated flow value B6 are normalized and their values ​​are taken, and the values ​​are calculated by formula to obtain the path saturation value Bα; the specific calculation formula is:

[0033]

[0034] Wherein α1, α2 and α3 are set weight constants respectively, and α1>α2>α3>1.

[0035] Preferably, the specific process of analyzing the efficiency of the inspection equipment is:

[0036] Obtain the historical inspection times of each inspection device and the path corresponding to each inspection, the total risk value F of the path, the path length G, the inspection start time (the time when the idle state is changed to the inspection state), and the inspection end time (the time when the inspection state is changed to the idle state); calculate the time difference between the inspection start time and the inspection end time to obtain the inspection duration, and thus the inspection duration corresponding to each inspection is recorded as T; normalize the total risk value F of the path, the path length G, and the inspection duration T and take their values, and perform formulaic calculation on the values ​​to obtain the efficiency value of each inspection; the specific calculation formula is:

[0037]

[0038] Among them, λ1 and λ2 are the set weight constants respectively;

[0039] The efficiency mean The interval distance A is normalized and its value is taken. The value is calculated and analyzed by formula to obtain the inspection value Aγ of the inspection equipment. The specific calculation formula is:

[0040]

[0041] Among them, γ1 and γ2 are the set weight constants respectively.

[0042] Beneficial effects of the present invention:

[0043] 1. By calling up various types of disease collections in the central server, several inspection routes including main roads are automatically generated to ensure that all road diseases are dispersed to different routes to achieve the purpose of comprehensive coverage; the traffic volume, road diseases, etc. of each route are comprehensively analyzed to obtain the priority index, and the efficiency of the inspection equipment is analyzed to obtain the inspection value, and the path inspection tasks are allocated based on this, forming a dynamic planning of path inspection, ensuring the flexibility and adaptability of inspection activities, and significantly improving the overall efficiency and safety of road inspection;

[0044] 2. By performing defect recognition on the collected road images, the type and degree of defects in the road images can be accurately extracted, providing a quantitative defect evaluation. Then, the road images with road defects are de-identified and the repetition index is calculated, which can accurately determine the similarity between the target image and the comparison image, avoid repeated detection, improve the efficiency and accuracy of defect detection, and provide intelligent and scientific decision-making support for road maintenance and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The present invention will be further described below in conjunction with the accompanying drawings.

[0046] Figure 1 is a schematic diagram of system module connection of the present invention;

[0047] Figure 2 It is a schematic diagram of the process of road disease identification and deduplication of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] See also Figure 1 As shown, the present invention is a road disease inspection data analysis system based on the Internet of Things and artificial intelligence, including: a central server, an inspection planning module and a disease identification module;

[0050] The central server exchanges data with inspection equipment (such as inspection vehicles or drones) through the communication network. Various sensors (such as high-precision GPS, high-definition cameras, etc.) carried on the inspection equipment conduct inspections along the corresponding paths to collect road images and their corresponding location information in real time (location information refers to the road name and location coordinates where the road image is located), and uploads them to the central server for storage through the communication network. In addition, the central server also has a number of disease type sets, each of which stores a number of disease images corresponding to the road disease type, as well as the location information, three characteristic elements, disease parameters, and disease severity values ​​corresponding to each disease image. The central server also communicates with the mobile terminals of each road maintenance personnel, and sends each road disease and its corresponding location information, disease type, etc. to the mobile terminals of the road maintenance personnel. When the maintenance completion is received from the mobile terminal of the road maintenance personnel, it is deleted from the corresponding disease type set.

[0051] The inspection planning module plans inspection routes based on the collection of various types of defects in the central server to guide the inspection equipment to conduct inspections and collect road images, thereby effectively improving inspection efficiency. Specifically:

[0052] Step 1: retrieve a number of damage images corresponding to the road damage types stored in each damage type set, as well as the location information and damage degree value corresponding to each damage image, mark the damage corresponding to each damage image at the corresponding position in the city map, and generate a number of paths according to the path algorithm. Usually, the several paths here include the main roads that need to be inspected in the entire city; thus, all road damages are dispersed to each path;

[0053] Step 2: Select any one of the paths, divide the path into several sections, obtain the traffic flow of each section in real time, and compare and analyze it with the set traffic flow interval. When the traffic flow is greater than the maximum value in the set traffic flow interval, the section corresponding to the traffic flow is recorded as a highly saturated section; when the traffic flow is in the set traffic flow interval, the section corresponding to the traffic flow is recorded as a moderately saturated section; when the traffic flow is less than the minimum value in the set traffic flow interval, the section corresponding to the traffic flow is recorded as a low-saturated section; the number of highly saturated sections, moderately saturated sections and low-saturated sections in the path is counted respectively. The number of highly saturated sections B1, the number of moderately saturated sections B2, the number of low-saturated sections B3, the high-saturated flow value B4, the medium-saturated flow value B5 and the low-saturated flow value B6 are normalized and their values ​​are taken, and the values ​​are calculated by formula to obtain the path saturation value Bα; the specific calculation formula is:

[0054]

[0055] Among them, α1, α2 and α3 are respectively set weight constants, and their sizes are custom settings, and they satisfy α1>α2>α3>1. It can be seen from the formula that when the traffic volume of each section in the path is larger, the path saturation value is larger, which means that it is less suitable to conduct inspection at this time, which will interfere with the normal operation of the vehicle.

[0056] Step 3: retrieve the number of road diseases on the path, as well as the disease type and disease severity value corresponding to each road disease; set a type base value corresponding to each type of disease. It should be noted that different types of diseases correspond to different type base values. The type base value corresponding to potholes is greater than the type base value corresponding to ruts, and the type base value corresponding to ruts is greater than the type base value corresponding to cracks; match each road disease on the path with all the set disease types to obtain the corresponding type base value, normalize the type base value and the disease severity value and take their values, perform weighted calculation on the values ​​to obtain the disease risk value of each road disease; sum up the disease risk values ​​of each road disease on the path to obtain the total risk value, recorded as F;

[0057] Step 4: Obtain the path length, record it as G, normalize it with the road saturation value Bα and the total risk value F, and take their values. Formulate the values ​​to obtain the path priority index BF. The specific calculation formula is:

[0058]

[0059] Among them, β1 and β2 are respectively set weight constants, and their sizes are custom settings. It can be seen from the formula that when the path length is shorter and the total risk value is larger, the priority index of the path is larger; when the path saturation value of the path is larger, the priority index of the path is smaller; thus, the priority indexes of several paths can be obtained, and the one with the largest priority index is selected as the target path; it is assumed that there are several idle inspection devices, each of which corresponds to an inspection value, and the inspection device with the largest inspection value is selected as the target inspection device, and the target path is assigned to the target inspection device, thereby changing the state of the target inspection device from the idle state to the inspection state; when the target inspection device completes the inspection task of this path, the historical inspection times of the inspection device increase by one, and the state of the target inspection device is changed from the inspection state to the idle state;

[0060] Step 5: Obtain the historical inspection times of each inspection device and the path corresponding to each inspection, the total risk value F of the path, the path length G, the inspection start time (the time when the idle state is changed to the inspection state), and the inspection end time (the time when the inspection state is changed to the idle state); calculate the time difference between the inspection start time and the inspection end time to obtain the inspection duration, and thus the inspection duration corresponding to each inspection is recorded as T; normalize the total risk value F of the path, the path length G, and the inspection duration T and take their values, and perform formulaic calculation on the values ​​to obtain the efficiency value of each inspection; the specific calculation formula is:

[0061]

[0062] Among them, λ1 and λ2 are the set weight constants. It can be seen from the formula that when the length of the path is longer, the risk value is greater, and the inspection time is shorter, it means that the inspection efficiency is higher, and the efficiency value is greater; the efficiency value of each inspection is averaged to obtain the efficiency mean of the inspection equipment, which is recorded as , obtain the interval distance between each idle inspection device and the target path as A; the efficiency mean The interval distance A is normalized and its value is taken. The value is calculated and analyzed by formula to obtain the inspection value Aγ of the inspection equipment. The specific calculation formula is:

[0063]

[0064] Where γ1 and γ2 are respectively set weight constants, from which the inspection value of each inspection device can be obtained and updated to step 4;

[0065] Step 6: Whenever there is an inspection device in an idle state, repeat steps 2 to 5 until the inspection of all paths is completed;

[0066] By retrieving various types of disease collections from the central server, several inspection routes including main roads are automatically generated to ensure that all road diseases are dispersed to different paths to achieve the goal of comprehensive coverage; a comprehensive analysis of the traffic volume, road diseases, etc. of each path is conducted to obtain a priority index, and the efficiency of the inspection equipment is analyzed to obtain the inspection value, and the path inspection tasks are allocated based on this, forming a dynamic planning of path inspections, ensuring the flexibility and adaptability of inspection activities, and significantly improving the overall efficiency and safety of road inspections.

[0067] like Figure 2 As shown, the damage recognition module extracts and analyzes features based on the road image to determine whether there are road damages and the corresponding damage types and damage levels, thereby identifying various road damages on each path and performing deduplication detection on them and the road damages stored in the server to prevent duplicate marking of the damages; specifically:

[0068] Defect identification: The collected road images are preprocessed, such as denoising, graying, contrast enhancement, etc., to improve image quality. Common methods include Gaussian filtering, histogram equalization, etc., to enhance the details of the road images and facilitate the subsequent extraction and analysis of defect features; the location information of the image is parsed and associated, including the road name, GPS coordinates, timestamp, etc., to ensure that each road image accurately corresponds to its corresponding location information; the road images are respectively subjected to geometric feature extraction (usually using edge detection, morphological operation and other algorithms), texture feature extraction (usually using local binary pattern (LBP) or gray level co-occurrence matrix (GLCM) and other algorithms) and color feature extraction (usually using color histogram (Color Histogram) or color space conversion (Color Space Conversion)) to obtain the three characteristic elements of the road image, where the three characteristics must include geometric features, texture features and color features;

[0069] Based on the labeled sample data set, the deep learning algorithm training model such as convolutional neural network (CNN) is used to realize the recognition of different types of diseases (such as cracks, potholes, rutting, settlement, etc.); the three characteristic elements are input into the trained deep learning algorithm model such as convolutional neural network (CNN) to output the disease type and its corresponding disease parameters. Different disease types correspond to different disease parameters. For example, the disease parameters corresponding to cracks are the number of cracks and the crack area corresponding to each crack (the product of crack length and width); the disease parameters corresponding to potholes are pothole diameter, depth, and pothole edge damage; the disease parameters corresponding to settlement are settlement area and depth; set the disease parameters in each disease type to correspond to a weighting coefficient denoted as Pi, and denoted as Wi. The disease parameters of each disease type are weighted according to the corresponding weighting coefficient to obtain the disease degree value S; the specific calculation formula is:

[0070]

[0071] Where i=1,2,3...I, I is a positive integer, I is the total number of disease parameters in a certain disease type, and i is any parameter among the disease parameters; thus, the disease recognition can be performed on each road image to screen out the road images with road diseases, and the disease type and disease degree value corresponding to each road image with road diseases can be output;

[0072] De-duplication identification: randomly select one of the road images with road diseases as the target image, and use each disease image in the corresponding disease type set as the comparison image; retrieve the geometric contours of the target image and the comparison image, use the centroid calculation method to obtain the center point of the geometric contour, overlap the center points of the geometric contours of the target image and the comparison image, and rotate 360° until the overlap area is the largest, then stop rotating, calculate the intersection area and union area in the overlap form (the overlap form with the largest overlap area), and record them as Q1 and Q2 respectively. It should be noted that the intersection area refers to the overlapping area of ​​the geometric contours of the target image and the comparison image in the overlap form, and the union area refers to the total area of ​​the geometric contours of the target image and the comparison image in the overlap form;

[0073] The disease parameters corresponding to the target image and the comparison image are matched one by one to obtain several parameter pairs, and the parameters of each parameter pair are calculated by difference to obtain the parameter difference of each parameter pair; for example, the diameter of the pit of the target image and the diameter of the pit of the comparison image are a pair of parameters, and the difference between the two is calculated to obtain the difference of the pit diameter; the parameter difference of each parameter pair is calculated by mean to obtain the mean of the parameter difference between the target image and the comparison image, and it is recorded as C;

[0074] Retrieve the corresponding position information of the target image and the comparison image, and calculate the spatial distance between the two and record it as H;

[0075] Retrieve the disease degree values ​​corresponding to the target image and the comparison image, and record them as S1 and S2 respectively;

[0076] The intersection area Q1, the union area Q2, the parameter difference mean C, the spatial spacing H, the target image damage level S1 and the comparison image damage level S2 are normalized and their values ​​are taken. The values ​​are calculated by formula to obtain the repetition index SQ. The specific calculation formula is:

[0077]

[0078] Among them, η1, η2, η3, and η4 are respectively set weight constants; it should be noted that when the intersection area between the target image and the comparison image is closer to the union area, it means that the geometric contours of the target image and the comparison image are closer, and the repetition index is larger; when the parameter difference mean is smaller, it means that the disease parameters between the target image and the comparison image are closer, and the repetition index is larger; when the disease degree values ​​corresponding to the target image and the comparison image are closer, it means that the disease degrees of the target image and the comparison image are closer, and the repetition index is larger; thus, the repetition index between the target image and each image in the corresponding disease type set can be obtained, and the repetition index and The set repetition threshold is used for comparative analysis. When the repetition index is greater than the set repetition threshold, a repeated image is accumulated; when the accumulated number of repeated images is greater than or equal to one, it means that this road image has been collected in the previous inspection, and it will be deleted; when the accumulated number of repeated images is zero, the road image with road damage is recorded as a damaged image, and it and its corresponding location information, three characteristic elements, damage parameters and damage degree values ​​are stored in the corresponding damage type set of the central server; based on the dynamic update function of the inspection data, the real-time and accuracy of the damage data in the central server is guaranteed;

[0079] By performing defect recognition on the collected road images, the type and degree of defects in the road images can be accurately extracted, providing a quantitative defect evaluation. Then, by performing de-duplication recognition on road images with road defects and calculating the repetition index, the similarity between the target image and the comparison image can be accurately determined, duplicate detection can be avoided, the efficiency and accuracy of defect detection can be improved, and intelligent and scientific decision-making support can be provided for road maintenance and management.

[0080] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A road disease inspection data analysis system based on the Internet of Things and artificial intelligence, including a central server, characterized in that: Also includes: Inspection planning module and disease identification module; The inspection planning module plans inspection routes based on various types of disease sets to guide inspection equipment to conduct inspections and collect road images; The disease recognition module extracts features from road images and performs disease recognition to determine whether there are road diseases, and then removes duplicates from road images with road diseases; the specific process of de-duplicate recognition is as follows: any one of the road images with road diseases is selected as the target image, each disease image in the corresponding disease type set is used as a comparison image, the geometric contours of the target image and the comparison image are retrieved and the center point of the geometric contour is found, the center point of the geometric contour of the target image and the comparison image is overlapped, and the rotation is stopped by 360° until the overlap area is the largest, and the intersection area and union area under the overlap form are calculated; the disease images corresponding to the target image and the comparison image are overlapped, and the intersection area and the ... The damage parameters are matched one by one to obtain multiple parameter pairs, and the parameters of each parameter pair are calculated to obtain the parameter difference of each parameter pair; the parameter difference of each parameter pair is calculated to obtain the mean value of the parameter difference between the target image and the comparison image; the position information corresponding to the target image and the comparison image is retrieved, and the spatial spacing between the two is calculated accordingly; the disease degree values ​​corresponding to the target image and the comparison image are retrieved; the intersection area Q1, the union area Q2, the parameter difference mean C, the spatial spacing H, the disease degree value S1 of the target image and the disease degree value S2 of the comparison image are normalized and their values ​​are taken, and the values ​​are calculated formulaically to obtain the repetition index SQ. The specific calculation formula is: , Where η1, η2, η3, and η4 are respectively set weight constants; when the repetition index is greater than the set repetition threshold, a repeated image is accumulated; when the accumulated number of repeated images is greater than or equal to one, it is deleted; when the accumulated number of repeated images is zero, the road image with road damage is recorded as a damaged image, and it is stored in the corresponding damage type set of the central server together with the corresponding location information, three characteristic elements, damage parameters, and damage degree values; The specific process of disease identification is as follows: Preprocess the collected road images, analyze and associate the image location information, so that each road image accurately corresponds to its corresponding location information; extract the geometric features, texture features and color features of the road images to obtain the three feature elements of the road images, where the three features include geometric features, texture features and color features; Based on the labeled sample data set, the convolutional neural network is trained to enable it to identify different types of road diseases; the three characteristic elements are input into the trained convolutional neural network algorithm model to output the disease type and its corresponding disease parameter; the disease parameters in each disease type are set to correspond to a weighting coefficient, and the disease parameters of the disease type are weighted and calculated according to the corresponding weighting coefficient to obtain the disease degree value; thereby, each road image is identified for disease to screen out road images with road diseases, and the disease type and disease degree value corresponding to each road image with road diseases are output.

2. The road disease inspection data analysis system based on the Internet of Things and artificial intelligence according to claim 1 is characterized in that: The specific process of inspection route planning based on various types of disease sets is as follows: Step 1: retrieve multiple damage images corresponding to road damage types, as well as the location information and damage degree values ​​corresponding to each damage image, in each damage type set, mark the damage corresponding to each damage image at the corresponding position in the city map, and generate multiple paths according to the path algorithm, so that all road damages are dispersed to each path; Step 2: Analyze the saturation degree of the traffic flow on the path to obtain the path saturation value; Step 3: retrieve the number of road diseases on the path, as well as the disease type and disease severity value corresponding to each road disease; set a type base value corresponding to each type of disease; match each road disease on the path with all the set disease types to obtain the corresponding type base value, normalize the type base value and the disease severity value and take their values, perform weighted calculation on the values ​​to obtain the disease risk value of each road disease; sum up the disease risk values ​​of each road disease on the path to obtain the total risk value; Step 4: Obtain the length of the path, normalize it with the road saturation value and the total risk value, and take the numerical value, and calculate the numerical value in a formula to obtain the priority index of the path; thereby obtaining the priority indexes of multiple paths, and selecting the one with the largest priority index as the target path; setting multiple idle inspection devices, each of which corresponds to an inspection value, selecting the inspection device with the largest inspection value as the target inspection device, and assigning the target path to the target inspection device, thereby changing the state of the target inspection device from the idle state to the inspection state; When the target inspection device completes the inspection task of this path, the historical inspection times of the inspection device increase by one, and the state of the target inspection device is changed from the inspection state to the idle state; Step 5: Obtain the historical inspection times of each inspection device, and analyze the efficiency of the inspection device based on the historical inspection times to obtain the inspection value of the inspection device; thereby obtaining the inspection value of each inspection device and updating it to step 4; Step 6: Whenever there is an inspection device in an idle state, repeat steps 2 to 5 until the inspection of all paths is completed.

3. The road disease inspection data analysis system based on the Internet of Things and artificial intelligence according to claim 2 is characterized in that: The specific process of analyzing the saturation degree of traffic flow on the path is as follows: Select any one of the routes, divide the route into multiple sections, obtain the traffic volume of each section in real time, and compare and analyze it with the set traffic volume interval to divide the section corresponding to the traffic volume into a highly saturated section, a moderately saturated section, and a low saturated section; count the number of highly saturated sections, moderately saturated sections, and low saturated sections in the route respectively, and sum and calculate the traffic volume corresponding to the highly saturated section, moderately saturated section, and low saturated section to obtain a high saturated flow value, a moderately saturated flow value, and a low saturated flow value; The number of highly saturated sections, the number of moderately saturated sections, the number of low saturated sections, the high saturated flow value, the medium saturated flow value and the low saturated flow value are normalized and their numerical values ​​are taken, and the numerical values ​​are calculated by formula to obtain the path saturation value.

4. The road disease inspection data analysis system based on the Internet of Things and artificial intelligence according to claim 2 is characterized in that: The specific process of analyzing the efficiency of inspection equipment is as follows: Obtain the historical inspection times of each inspection device, as well as the path corresponding to each inspection, the total risk value of the path, the length of the path, the start time of the inspection, and the end time of the inspection; calculate the time difference between the start time and the end time of the inspection to obtain the inspection duration, thereby obtaining the inspection duration corresponding to each inspection; The total risk value, path length and inspection time of the path are normalized and their values ​​are taken, and the efficiency value of each inspection is obtained by formulating the values; The efficiency value of each inspection is averaged to obtain the efficiency mean of the inspection equipment, and the interval distance between each idle inspection equipment and the target path is obtained; the efficiency mean and the interval distance are normalized and their values ​​are taken, and the values ​​are formulated and analyzed to obtain the inspection value of the inspection equipment.

5. The road disease inspection data analysis system based on the Internet of Things and artificial intelligence according to claim 1 is characterized in that: The central server exchanges data with the inspection equipment through the communication network. The various sensors carried on the inspection equipment conduct inspections along the corresponding paths to collect road images and their corresponding location information in real time, and upload them to the central server for storage through the communication network. The central server also has multiple disease type sets, each of which stores multiple disease images corresponding to the road disease type, as well as the location information, three characteristic elements, disease parameters and disease severity values ​​corresponding to each disease image.

Images