A method for locating highway pavement defects based on drones

By integrating high-resolution video collectors and 5G communication terminals on the drone, combined with Beidou satellite positioning and improved DPM detection model, the automated detection and positioning of road surface diseases is achieved, and the low detection efficiency and poor accuracy in traditional methods are solved. The problem of the detection speed and accuracy are improved.

CN110894704BActive Publication Date: 2024-11-19SOUTHEAST UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN201911120174.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2019-11-15
Publication Date
2024-11-19
Estimated Expiration
2039-11-15

AI Technical Summary

Technical Problem

Traditional highway road disease detection methods have problems such as high detection cost, slow speed, a lot of manpower, and the detection accuracy is affected by personnel level and experience, resulting in low positioning efficiency and poor accuracy.

Method used

It adopts a low-altitude system based on drones, equipped with a high-resolution video collector, 5G communication terminal and Beidou satellite positioning system, and obtains pavement images and positioning information through low-altitude drones, and combines the improved DPM detection model for automated detection and positioning. Realize data transmission and information exchange.

Benefits of technology

It realizes rapid and accurate detection and positioning of road road diseases, reduces detection costs, improves detection efficiency and accuracy, and solves the problems of large labor consumption and limited accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN110894704B_ABST
    Figure CN110894704B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for locating road pavement defects based on drones, including: a drone system integrating a high-resolution video collector, a 5G mobile communication terminal, and a Beidou satellite positioning and navigation module; obtaining high-resolution road image data, drone flight altitude data, and drone location information; relying on the transmission of detection data of 5G mobile communication technology, and using a road surface disease detection model based on an improved DPM to detect and locate road surface diseases. The beneficial effects of the present invention are: effectively solving the shortcomings of the current road surface disease data acquisition method, such as low automation, high acquisition cost, and slow speed, as well as the low detection efficiency, large fluctuations in detection accuracy, large workload in the positioning process, and low positioning automation in the road surface disease detection and positioning method; the present invention has strong practicality and can be widely used for road surface disease detection and positioning of highways, national highways, other low-grade highways, and urban roads.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a method for locating road pavement defects based on an unmanned aerial vehicle, and in particular to a method for locating road pavement defects based on an unmanned aerial vehicle. Background Art

[0002] 5G communication technology is the abbreviation of the fifth generation of mobile communication technology. Like the 3G and 4G technologies that have been put into market application, 5G mobile communication technology also belongs to mobile communication technology that uses digital cellular networks. Compared with previous cellular data communication technologies, the advantages of 5G communication technology are that the data transmission speed and transmission capacity are much higher than the previous cellular networks. The fastest data transmission speed of commercial 5G mobile communication technology can reach 10Gbit / s, which is 100 times the data transmission rate of the current 4G LTE cellular network. In addition, 5G mobile communication technology also has the advantages of low network latency, fast response speed, and good compatibility with mobile network terminals. On June 6, 2019, the Ministry of Industry and Information Technology of the People's Republic of China officially issued commercial 5G licenses to China Telecom, China Mobile, China Unicom, and China Broadcasting Corporation, marking that China has entered the commercial era of 5G communication technology.

[0003] BeiDou Navigation Satellite System, also known as BDS, is a new generation of global satellite positioning and navigation system independently developed by China. It is the third mature satellite positioning and navigation system that can be put into use in the world after the US GPS satellite navigation system and the Russian GLONASS satellite navigation system. BeiDou satellite positioning and navigation system consists of three parts: space segment, ground segment and user segment. The space segment consists of inclined earth orbit synchronous satellites, medium earth orbit satellites and geostationary orbit satellites; the ground segment includes several ground stations such as communication control stations and satellite system monitoring stations, as well as line operation and management facilities and equipment; the user segment includes vehicle-mounted, ship-mounted, airborne, handheld BeiDou satellite positioning terminals used by users of the BeiDou satellite positioning and navigation system, as well as other chips, antennas, modules and other products that use or are compatible with the BeiDou satellite navigation system, as well as corresponding applications and services. The advantages of BeiDou satellite navigation are that compared with other satellite navigation systems, it has strong anti-blocking and anti-interference capabilities; it can provide navigation signals at multiple frequencies and has the advantages of higher positioning accuracy.

[0004] The DPM target detection algorithm is a component-based detection algorithm. The algorithm uses the HOG feature detection algorithm for the detection target and improves the HOG feature. It also combines the detection methods of sliding window and SVM classifier. The DPM algorithm has good stability in detecting the deformation of the target. Therefore, it is widely used in various image detection algorithms such as target classification, segmentation, and posture detection. It is a relatively mature algorithm among the current image detection algorithms.

[0005] The condition of road pavement damage is one of the important indicators for evaluating the health of roads. Road pavement damage can be divided into several main types of damage, such as cracks, transverse cracks, longitudinal cracks, block cracks, potholes, looseness, and rutting. The traditional method of detecting and locating road pavement damage mainly refers to the use of detection vehicles, laser radars, and detection personnel taking photos to obtain road photos, and then the staff completes the detection and classification of road pavement damage by manual judgment and positioning. This method has the problems of high detection cost, slow detection speed, large amount of manpower and material resources, and the detection accuracy is greatly affected by factors such as the detection level, work experience, and physiological condition of the detection personnel. This has greatly restricted the current road maintenance and management departments from discovering, detecting, and locating road damage conditions, and has become an urgent problem in the current field of road maintenance and management. Summary of the invention

[0006] In view of the shortcomings and defects of the existing pavement disease detection and positioning technology, the present invention proposes a method for positioning highway pavement diseases based on drones. First, a low-altitude drone system equipped with a high-resolution video collector, a fifth-generation mobile communication technology terminal and a Beidou satellite positioning system terminal is constructed. The low-altitude drone system that cruises over the road surface to be detected at a certain altitude and flight speed takes regular photos of the road surface to obtain high-resolution road surface image data, and integrates the horizontal, vertical and high three-dimensional coordinate position information of the drone obtained by the Beidou satellite positioning system; the fifth-generation mobile communication technology, that is, 5G technology, is used to complete the transmission, recording and storage functions of the pavement disease image and the pavement disease coordinate position data, and realize the data transmission and information exchange between the drone platform and the data processing server; after the server obtains the data, the pavement disease image is first pre-processed, and then the high-resolution pavement image obtained by the low-altitude drone platform is detected for disease using the pavement disease detection model based on the improved DPM, and the three-dimensional features of the pavement disease are obtained, and the pavement disease to be detected are determined. The location of the pavement disease is determined by the horizontal, vertical and high three-dimensional coordinate position information of the drone obtained by the Beidou satellite positioning system and the flight altitude of the drone.

[0007] The principle of the present invention is that a low-altitude UAV system cruising at a fixed height and fixed speed can take pictures of the road surface to be inspected at a certain shooting frequency through the high-resolution image acquisition equipment carried by the UAV, and the high-resolution image data obtained contains information such as the texture, contour, color, and undulation of the road surface. Through 5G mobile communication technology, it is possible to transmit the high-resolution image data of the road surface obtained by the UAV and the position information and status information of the UAV obtained by the Beidou satellite positioning terminal to the data processing server. After the data processing server preprocesses the road surface image, the road surface disease detection model based on the improved DPM performs road surface disease detection and positioning. The road surface detection model based on the improved DPM adopted by the present invention adopts a method of connecting two DPM detection models in series. The DPM-1 model is responsible for the detection of repaired road surface diseases, including the detection of three types of road surface diseases: road surface repaired transverse cracks, road surface repaired longitudinal cracks, and road surface repaired network cracks; the DPM-2 model is responsible for the detection of unrepaired road surface diseases, including unrepaired road surface transverse cracks, unrepaired road surface longitudinal cracks, unrepaired road surface network cracks, and other, a total of four types of road surface disease detection. During the inspection, the DPM-1 model is first used to inspect the image to be inspected. If the DPM-1 model detects the presence of a disease, the inspection is terminated and the inspection result is that there is a road disease. If the DPM-1 model does not detect the presence of a road disease, the image is transferred to the DPM-2 model for inspection. If the DPM-2 model detects the presence of a disease, the inspection is terminated and the inspection result is that there is a road disease. If the DPM-2 model does not detect the presence of a road disease, the inspection is terminated and the inspection result is that there is no road disease. When the inspection result is that there is a road disease, the location of the road disease is calculated through the latitude and longitude, elevation information and flight altitude of the drone at the corresponding inspection time, and the road disease is located.

[0008] The technical solution provided by the present invention is:

[0009] A method for locating highway pavement defects based on unmanned aerial vehicles (UAVs), including: construction of an UAV system, acquisition of pavement and UAV data, data preprocessing, construction of a pavement detection model based on an improved DPM, and pavement defect detection and positioning.

[0010] Among them, the construction of the drone system includes installing and integrating high-resolution image collectors, 5G communication terminals and Beidou satellite positioning modules on low-altitude drone platforms, with the purpose of building a low-altitude drone system with data collection, real-time positioning and data transmission functions. Pavement and drone data acquisition refers to obtaining high-resolution photos of highway pavement and information such as the flight altitude and location of drones through drones, with the purpose of providing data for the detection and positioning of pavement defects. Data preprocessing refers to normalizing the size of highway pavement photos before detection, with the purpose of making the image format meet the requirements of the detection model for the image format. The construction of the pavement detection model based on the improved DPM includes model construction, training data set construction and model training, with the purpose of making the pavement detection model based on the improved DPM have better recognition accuracy and detection speed for highway pavement defects. Pavement defect detection and positioning refers to detecting highway pavement defects and determining the location of highway pavement defects through the trained pavement detection model based on the improved DPM.

[0011] The pavement disease detection method provided by the present invention specifically comprises the following steps:

[0012] 1) Build a low-altitude UAV system for detection. The low-altitude UAV system needs to be equipped with a high-resolution video collector, a fifth-generation mobile communication terminal, and a Beidou satellite positioning terminal. The high-resolution video collector refers to the Phase One iXU 180 high-resolution aerial survey camera installed on the UAV platform through the LEMO aviation interface, the fifth-generation mobile communication terminal refers to the 5G communication terminal equipment, and the Beidou satellite positioning system terminal refers to the UFirebird UC6226 Beidou positioning chip;

[0013] 2) The low-altitude UAV system constructed in step 1) cruises over the road surface to be inspected at a speed of 3m / s and an altitude of 7m. The UAV platform collects high-resolution image data of the road surface at a fixed frequency of taking a photo every 0.1 seconds, and the captured images are saved in JPG image format. The saved resolution is 1920×1080. While the UAV system is collecting pictures of the road surface, the Beidou satellite positioning module obtains the location information of the UAV;

[0014] 3) Installing a 5G communication terminal on the drone, transmitting the drone working status, drone location information, and pavement damage image data obtained in step 2) to the data processing server according to the set naming, coding, and transmission rules;

[0015] 4) After receiving the data transmitted in step 3), the background processor pre-processes the high-resolution road surface image transmitted back by the drone. At the same time, the road surface disease detection model based on the improved DPM is trained;

[0016] 5) The road surface image preprocessed in step 4) is subjected to road surface disease detection by a road surface disease detection model composed of two detection models based on improved DPM, DPM-1 and DPM-2, which are connected in series, to obtain a road surface disease detection result;

[0017] 6) If the detection result of step 5) is that there is a road surface disease, the location of the road surface disease is calculated based on the drone position information and flight altitude information when the image data showing the presence of the road surface disease is taken.

[0018] For the above-mentioned road pavement disease detection and positioning method. Further, the drone location information involved in step 2) specifically refers to the horizontal and vertical coordinates and elevation information of the drone, wherein the horizontal and vertical coordinate information of the drone is represented by the longitude and latitude coordinates of the drone, and the elevation information of the drone is represented by the elevation coordinates based on the 1985 National Elevation Datum as the elevation reference surface.

[0019] For the above-mentioned road pavement disease detection and location method, further, the data naming, coding and transmission rules involved in step 3) specifically refer to:

[0020] a) Save the UAV flight attitude, working status, and flight altitude information in a txt format text file. The text file is named according to the measurement time. The specific naming format is "year / month / day-hour: minute: second: millisecond";

[0021] The text file uses a 9-digit code to save the UAV's flight attitude, working status, and flight altitude information. The first 4 digits save the UAV's flight tilt angle in radians, and the 5th digit indicates the UAV's working status. Normal operation of the UAV is recorded as "1" and abnormal operation is recorded as "2". The 6th to 9th digits save the UAV's flight altitude information accurate to 2 decimal places in meters. The txt format text file that shows the UAV's flight attitude, working status, and flight altitude information is uniformly saved in a folder named "txt-measurement date" in units of days;

[0022] b) Recode and name the high-resolution road image data collected by the high-resolution video collector carried by the drone. The naming format is a 34-character group consisting of numbers, English letters, and hyphens. Characters 1 to 18 represent the drone's location information, of which the first 6 characters represent the drone's longitude coordinates, accurate to 2 decimal places; 7 to 11 characters represent the drone's latitude coordinates, accurate to 2 decimal places; 12 to 18 characters represent the drone's elevation coordinates, in meters, accurate to 2 decimal places; the 19th character is a hyphen; characters 20 to 33 represent the drone's image acquisition time, and the naming format is "year-month-day-hour-minute-second". The 34th character is the photo shooting sequence number, indicating the position of the photo taken within the current second. The renamed images are uniformly saved in a folder named "image-date of measurement" in units of days;

[0023] Regarding the above-mentioned road pavement disease detection and positioning method, further, the specific contents of image preprocessing and training of the road pavement disease detection model based on the improved DPM in step 4) refer to:

[0024] a) The image preprocessing process specifically includes copying the original image, saving the copy in a folder named "test date + test", resizing the image copy in the folder, and changing the resolution to 512×500 format to meet the size requirements of the pavement disease detection model based on the improved DPM for the detection sample. At the same time, the image copy is renamed in numerical order from small to large starting from 1 to serve as a sample for roadside disease detection and positioning;

[0025] b) The specific process of training the pavement disease detection model based on the improved DPM refers to training the pavement disease detection model based on the improved DPM through a certain university's pavement disease data set that includes seven types of disease types, including pavement repaired transverse cracks, pavement repaired longitudinal cracks, pavement repaired mesh cracks, pavement unrepaired transverse cracks, pavement unrepaired longitudinal cracks, pavement unrepaired mesh cracks, and others, with a total data volume of more than 10,000 pictures and more than 1,000 pictures of each disease type, so that its detection speed and accuracy meet the detection requirements. During model training, 50 pictures of each type of disease data in the pavement disease data set of a certain university are selected as training positive samples, 80% of the data are randomly selected for the model training process training set, and the remaining 20% ​​of the data are used for the model training verification sample set, and then 200 pictures of pavement without disease are randomly selected as training negative samples;

[0026] For the above-mentioned highway pavement disease detection and positioning method. Further, the pavement disease detection process involved in step 5) specifically refers to first using the DPM-1 model to detect the image to be inspected. If the DPM-1 model detects the existence of the disease, the detection is terminated, and the detection result is the existence of pavement disease. At the same time, the image is saved in a folder corresponding to the disease type. If the DPM-1 model does not detect the existence of the pavement disease, the image is transferred to the DPM-2 model for detection. If the DPM-2 model detects the existence of the disease, the detection is terminated, and the detection result is the existence of the pavement disease. At the same time, the image is saved in a folder corresponding to the disease type; if the DPM-2 model does not detect the existence of the pavement disease, the detection is terminated, and the detection result is the absence of the pavement disease.

[0027] Regarding the above-mentioned road pavement disease detection and positioning method, further, the pavement disease positioning process involved in step 6) specifically refers to:

[0028] a) Compare the shooting time of the high-resolution image data of the road surface disease and the longitude coordinate X of the UAV obtained by the Beidou satellite positioning system at that time. 1 , Latitude coordinate Y 1 , and elevation information Z 1 and the UAV flight altitude information H recorded by the UAV system itself 1 ;

[0029] b) The longitude coordinate X of the location of the pavement defect is calculated according to formula (1), the latitude coordinate Y of the location of the pavement defect is calculated according to formula (2), and the elevation coordinate Z of the location of the pavement defect is calculated according to formula (3):

[0030] X=X 1 Formula (1)

[0031] Y=Y 1 Formula (2)

[0032] Z=Z 1 -H 1 Formula (3)

[0033] c) Determine the location of the pavement defect based on its longitude, latitude and elevation coordinates.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] Low-altitude UAV systems can be used to obtain high-definition video image data of road pavement diseases. Compared with the traditional method of obtaining road images by shooting with inspection vehicles, aerial photography, and manual shooting, it has the advantages of fast detection speed, stable detection frequency, low detection cost, less labor demand, and all-weather operation. Relying on 5G mobile communication technology, it can solve the shortcomings of the current 4G mobile communication technology, such as slow transmission speed and large transmission capacity restrictions. The use of 5G mobile communication technology and equipment can realize fast and large-capacity data transmission and information communication between the road disease detection site and the background processing server. Relying on the Beidou satellite positioning module and the road disease detection model based on the improved DPM, the detection and positioning functions of road diseases can be automatically completed by the model, with high detection accuracy, fast detection speed, and accurate positioning; it can solve the shortcomings of the current method of relying on inspectors to manually detect and locate road diseases by checking road diseases, such as low detection efficiency, poor detection accuracy, large labor consumption, high labor cost, and the detection effect is greatly affected by factors such as the technical level, work experience, and physiological condition of the inspectors.

[0036] The drone-based highway pavement disease positioning method proposed in the present invention can realize the full-process automated detection operation of highway pavement diseases from data acquisition to positioning, including highway pavement data acquisition, data transmission, disease detection, and disease positioning, which can greatly improve the level of highway maintenance and management services and reduce the cost of highway maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of the method for locating highway pavement defects based on drones proposed in the present invention.

[0038] Figure 2 This is a detection principle diagram of the method for locating highway pavement defects based on drones proposed in the present invention.

[0039] Figure 3 This is a structural diagram of the drone system constructed by the present invention.

[0040] Figure 4 Schematic diagram of the improved DPM detection algorithm.

[0041] Figure 5 This is a pavement disease dataset from a certain university used for training the pavement detection model based on the improved DPM.

[0042] Figure 6 Flow chart for pavement disease detection and location based on the improved DPM pavement detection model. DETAILED DESCRIPTION

[0043] The specific implementation process of the present invention is further described below with reference to the accompanying drawings and through specific examples, but the scope of application of the present invention is not limited in any way.

[0044] Figure 1 The flowchart of the method for locating road pavement defects based on drones proposed by the present invention includes the following steps:

[0045] Step 1: Install the Phase One iXU 180 high-resolution aerial survey camera, Huawei 5G mobile communication terminal, and UFirebird UC6226 Beidou positioning chip on the low-altitude drone to build a low-altitude drone system;

[0046] Step 2: The low-altitude drone constructed in step 1 cruises on the road surface to be inspected at a speed of 3m / s and a flight altitude of 7m, and takes a high-definition photo of the road surface every 0.1 seconds. The photo is saved in 1920×1080 resolution and JPEG format. When there are height-restricted facilities on the road surface or the terrain conditions do not allow it, the flight altitude of the drone can be adjusted according to the actual situation;

[0047] Step 3: When the drone system performs the shooting process described in step 2, the data such as the drone's flight altitude and flight status are synchronously saved;

[0048] Step 4: When the drone system is performing the shooting process described in step 2, the airborne Beidou satellite positioning module obtains the longitude, latitude, and altitude information of the drone in real time;

[0049] Step 5: Recode and name the data and information involved in Step 2, Step 3, and Step 4;

[0050] Step 6: The onboard 5G mobile communication terminal transmits the data and information re-encoded and named in step 5 to the data processing server;

[0051] Step 7: The data processing server preprocesses the pavement disease image data to meet the format and resolution requirements of the detection model; data preprocessing requires copying the original image and saving the copy in a folder named "Detection Date + Detection", resizing the image copy in the folder, changing the resolution to 512×500 format, and renaming the image copy in numerical order starting from 1 and from small to large. At the same time, the data processing server uses a university pavement disease dataset to train the pavement disease model based on the improved DPM.

[0052] like Figure 6The basic principle of the DPM image feature detection algorithm shown is that for any detection image, the DPM features of the image are first extracted, and then the input image is upsampled by the Gaussian pyramid, and then the DPM features of the sampled image are extracted again. The DPM feature map of the original input sample image and the trained root filter are convolved to obtain the response map of the root filter. The DPM feature map of the 2x resolution image and the trained component filter are then convolved to obtain the response map of the component filter. Then the fine Gaussian pyramid is downsampled. After the operation, the response maps of the root filter and the component filter have the same resolution, and then the weighted average calculation is performed to obtain the final response map.

[0053] like Figure 5 The pavement disease dataset of a certain university shown in the figure contains seven types of pavement disease data, including repaired pavement transverse cracks, repaired pavement longitudinal cracks, repaired pavement network cracks, unrepaired pavement transverse cracks, unrepaired pavement longitudinal cracks, unrepaired pavement network cracks, and other pavement disease data, with more than 1,000 images of each type of disease data. When training the model, first select 50 images of each type of disease data as training positive samples, randomly select 80% of them as model training data sets, and the remaining 20% ​​as model training validation sets, and then select 200 pavement image data without pavement diseases as training negative samples.

[0054] Step 8: Use the pavement defect detection model based on the improved DPM to detect and locate defects on the pavement image data preprocessed in step 7.

[0055] The specific process of disease detection and location is as follows: Figure 6 As shown, first, the preprocessed image data is used to detect the image to be inspected using the trained DPM-1 model. The DPM-1 model is used to detect three types of defects, namely, repaired transverse cracks, repaired longitudinal cracks, and repaired network cracks. If the detection result is that a defect exists, the defect is located. If the detection result is that no defect exists, the data is transferred to the trained DPM-2 model for detection of four other types of defects, namely, unrepaired transverse cracks, unrepaired longitudinal cracks, unrepaired network cracks, and unrepaired transverse cracks. If the detection result is that a pavement defect exists, the pavement defect is located. If the detection result is that no defect exists, the model's detection result output is that no pavement defect exists.

Claims

1. A method for locating road pavement defects based on drones, characterized in that: The specific steps include: 1.1) Building a low-altitude UAV system, which is equipped with a high-resolution video collector, a fifth-generation mobile communication terminal and a Beidou satellite positioning terminal; 1.2) The drone cruises at a certain flight speed and altitude, and collects high-resolution road surface image information at a shooting frequency of once every 0.1 seconds, and at the same time obtains the real-time location information data of the detection drone through the Beidou satellite positioning system; 1.3) Encode and store high-resolution road image information, and use the fifth-generation mobile communication technology to communicate and transmit data between the UAV platform, GPS measurement terminal and data processing server; 1.4) After receiving the data, the server pre-processes the high-resolution road image data and uses the road disease detection model based on the improved DPM model to detect and locate the high-resolution road image acquired by the low-altitude UAV platform; The low-altitude UAV platform used by the low-altitude UAV system configured with a high-resolution video collector, a fifth-generation mobile communication terminal and a Beidou satellite positioning terminal in step 1.1) is a civilian UAV, the high-resolution video collector refers to a Phase One iXU 180 high-resolution aerial survey camera installed on the UAV platform through a LEMO aviation interface, the fifth-generation mobile communication terminal refers to a 5G communication terminal device, and the Beidou satellite positioning system terminal refers to a UFirebird UC6226 Beidou positioning chip; In step 1.2), a low-altitude UAV platform is used to carry a high-resolution video collector to fly on the road surface to collect data, and at the same time, the real-time location information data of the detection UAV is obtained through the Beidou satellite positioning system; the high-resolution image information refers to the high-resolution digital photos taken by the high-resolution video collector, and contains the shooting time, shooting posture and UAV flight altitude positioning information files provided by the shooting UAV; The specific data collection steps include: 3.1) The UAV platform cruises along the road surface to be inspected at a speed of 3m / s and a flight altitude of 7m; 3.2) The UAV platform takes a photo at a fixed frequency of 0.1 seconds to collect high-resolution image data of the road surface. The images taken are saved at a resolution of 1920×1080 in the JPG format; 3.3) The Beidou satellite positioning terminal obtains the real-time location information data of the detection drone, including the X, Y, and Z coordinates of the drone, i.e., the horizontal, vertical, and altitude coordinates. The horizontal and vertical coordinates are expressed in terms of longitude and latitude. In step 1.3), communication and data encoding and transmission are completed through the fifth generation mobile communication technology; the specific process includes: 4.1) Save the UAV flight attitude, working status, and flight altitude information in a txt format text file. The text file is named according to the measurement time, and the specific naming format is "year / month / day-hour: minute: second: millisecond"; the UAV flight attitude, working status, and flight altitude information are saved in the text file through a 9-digit code. The first 4 digits save the UAV flight tilt angle in radians, and the fifth digit indicates the UAV working status. Normal operation of the UAV is recorded as "1" and abnormal operation is recorded as "2". The 6th to 9th digits save the UAV flight altitude information accurate to 2 decimal places, in meters; the txt format text files representing the UAV flight attitude, working status, and flight altitude information are uniformly saved in a folder named "txt-date of measurement" in units of days; 4.2) The high-resolution road image data collected by the high-resolution video collector carried by the drone is recoded and named. The naming format is a 34-character group consisting of numbers, English letters, and hyphens. Characters 1 to 18 represent the drone's location information, of which the first 6 characters represent the drone's longitude coordinates, accurate to 2 decimal places; 7 to 11 characters represent the drone's latitude coordinates, accurate to 2 decimal places; 12 to 18 characters represent the drone's elevation coordinates, in meters, accurate to 2 decimal places; the 19th character is a hyphen; characters 20 to 33 represent the drone's image acquisition time, and the naming format is "year-month-day-hour-minute-second", and the 34th character is the photo shooting sequence number, indicating the position of the photo taken within the current second; the renamed images are uniformly saved in a folder named "image-measurement date" in units of days; 4.3) Through the 5G communication terminal installed on the drone, the folder containing the txt file containing the drone's attitude, working status, and flight altitude information, as well as the folder containing the recoded and named high-resolution road photos, is sent to the data processing server using 5G mobile communication technology. The data is sent once every 24 hours. In step 1.4), a pavement disease detection model based on an improved DPM model is used. The model is composed of a DPM-1 model for detecting repaired diseases and a DPM-2 model for detecting unrepaired diseases connected in series. The DPM-1 model is responsible for detecting three types of pavement diseases: repaired transverse cracks, repaired longitudinal cracks, and repaired lattice cracks. The DPM-2 model is responsible for detecting unrepaired transverse cracks, unrepaired longitudinal cracks, unrepaired lattice cracks, and the other four types of pavement diseases. In step 1.4), the pavement defect detection model based on the improved DPM model is used to detect and locate defects on the pre-processed high-resolution pavement image. The specific detection and process include: 6.1) Preprocess the high-resolution road surface images transmitted back by the drone. The preprocessing process is to copy the original image, save the copy in a folder named "test date + test", resize the image copy in the folder, change the resolution to 512×500 format, and rename the image copy in numerical order from 1 to large, so as to serve as samples for roadside disease detection and location; 6.2) Use the pavement disease detection model based on the improved DPM model to perform pavement disease detection on the pavement disease detection and positioning samples. The specific detection process is: first use the DPM-1 model to detect the image to be inspected. If the DPM-1 model detects the existence of the disease, the detection is terminated, and the detection result is that the pavement disease exists. At the same time, the image is saved in a folder corresponding to the disease type. If the DPM-1 model does not detect the existence of the pavement disease, the image is transferred to the DPM-2 model for detection. If the DPM-2 model detects the existence of the disease, the detection is terminated, and the detection result is that the pavement disease exists. At the same time, the image is saved in a folder corresponding to the disease type; if the DPM-2 model does not detect the existence of the pavement disease, the detection is terminated, and the detection result is that the pavement disease exists. At the same time, the image is saved in a folder corresponding to the disease type; if the DPM-2 model does not detect the existence of the pavement disease, the detection is terminated, and the detection result is that the pavement disease does not exist; 6.3) If the detection result of the road surface image to be inspected is that there is a defect, the road surface defect detection model based on the improved DPM model will locate the location of the road surface defect. The specific positioning process is to compare the shooting time of the high-resolution image data where the road surface defect is located, and the longitude coordinate X1, latitude coordinate Y1, and elevation information Z1 of the drone obtained by the Beidou satellite positioning system at that time, and the drone flight altitude information H1 recorded by the drone system itself, and then calculate the location of the road surface defect; The location of the road pavement defect in step 6.3) is calculated by, specifically, calculating the longitude coordinate X of the location of the road pavement defect according to formula (1), calculating the latitude coordinate Y of the location of the road pavement defect according to formula (2), and calculating the elevation coordinate Z of the location of the road pavement defect according to formula (3). X=X1 Formula (1) Y=Y1 Formula (2) Z=Z1-H1 Formula (3) In the formula, X1 represents the longitude coordinate of the UAV obtained by the Beidou satellite positioning system, Y1 represents the latitude coordinate of the UAV obtained by the Beidou satellite positioning system, Z1 represents the altitude coordinate of the UAV obtained by the Beidou satellite positioning system, and H1 represents the flight altitude information of the UAV recorded by the UAV system itself.