阅读说明：本技术 一种基于无人机的公路路面病害的定位方法 (Unmanned aerial vehicle-based method for positioning road surface diseases ) 是由 赵池航 李�昊 袁守国 张澄 毛志坚 郑有凤 化丽茹 于 2019-11-15 设计创作，主要内容包括：本发明公布了一种基于无人机的公路路面病害的定位方法,包括：集成了高分辨率视频采集器、5G移动通信终端、北斗卫星定位导航模块的无人机系统；对路面高分辨率影像数据、无人机飞行高度数据、无人机位置信息的获取；依靠5G移动通信技术的检测数据的传输、利用基于改进DPM的路面病害检测模型对路面病害进行检测与定位。本发明的有益效果在于：有效解决了当前路面病害数据获取方法存在的自动化程度低,获取成本高、速度慢以及路面病害检测及定位方法存在的检测效率低、检测精确度波动大、定位过程工作量大、定位自动化程度低等缺点；实用性强,可广泛用于高速公路、国道、其他低等级公路以及城市道路的路面病害检测及定位。(The invention discloses a method for positioning road pavement diseases based on an unmanned aerial vehicle, which comprises the following steps: the unmanned aerial vehicle system integrates a high-resolution video collector, a 5G mobile communication terminal and a Beidou satellite positioning navigation module; acquiring high-resolution image data of a road, flight height data of an unmanned aerial vehicle and position information of the unmanned aerial vehicle; the detection data transmission of the 5G mobile communication technology is relied on, and the pavement damage detection model based on the improved DPM is utilized to detect and position the pavement damage. The invention has the beneficial effects that: the defects of low automation degree, high acquisition cost, low speed, low detection efficiency, large detection accuracy fluctuation, large workload in the positioning process, low positioning automation degree and the like of the conventional pavement disease data acquisition method are effectively overcome; the method has strong practicability, and can be widely used for detecting and positioning the pavement diseases of expressways, national roads, other low-grade roads and urban roads.)

1. A method for positioning road pavement diseases based on an unmanned aerial vehicle is characterized by comprising the following specific steps:

1.1) constructing a low-altitude unmanned aerial vehicle system, wherein a high-resolution video collector, a fifth generation mobile communication terminal and a Beidou satellite positioning terminal are configured on the low-altitude unmanned aerial vehicle system;

1.2) the unmanned aerial vehicle cruises at a certain flying speed and height, simultaneously acquires high-resolution image information of the road surface according to the shooting frequency of high-resolution image data of the road surface acquired once every 0.1 second, and simultaneously acquires the instant position information data of the unmanned aerial vehicle for detection through a Beidou satellite positioning system;

1.3) coding and storing the road surface high-resolution image information, and carrying out communication and data transmission between an unmanned aerial vehicle platform, a GPS measuring terminal and a data processing server by a fifth generation mobile communication technology;

1.4) the server preprocesses the road high-resolution image data after receiving the data, and detects and positions the diseases of the high-resolution road image acquired by the low-altitude unmanned aerial vehicle platform by using a road disease detection model based on an improved DPM model.

2. The unmanned aerial vehicle-based road pavement disease positioning method according to claim 1, characterized in that: the low-altitude unmanned aerial vehicle platform that the low-altitude unmanned aerial vehicle system who disposes high resolution video collector, fifth generation mobile communication terminal and big dipper satellite positioning terminal in step 1.1) adopted is civilian unmanned aerial vehicle, high resolution video collector indicates that the flying side iXU 180 type high resolution aerial survey camera of installation on the unmanned aerial vehicle platform through LEMO aviation interface, fifth generation mobile communication terminal indicates 5G communication terminal equipment, big dipper satellite positioning system terminal indicates UFireblrd UC6226 type big dipper positioning chip.

3. The unmanned aerial vehicle-based road pavement disease positioning method according to claim 1, characterized in that: in the step 1.2), a low-altitude unmanned aerial vehicle platform is used for carrying a high-resolution video collector to fly on a highway pavement to collect data, and meanwhile, the instant position information data of the unmanned aerial vehicle for detection is obtained through a Beidou satellite positioning system; the high-resolution image information refers to a high-resolution digital photo shot by a high-resolution video collector and is provided with shooting time, shooting attitude and unmanned aerial vehicle flying height positioning information files provided by a shooting unmanned aerial vehicle;

the specific data acquisition steps comprise:

3.1) the unmanned aerial vehicle platform cruises and flies along the road surface to be detected at the flying speed of 3m/s and the flying height of 7 m;

3.2) the unmanned aerial vehicle platform collects high-resolution image data of the road surface according to the fixed frequency of taking a picture every 0.1 second, the taken image is stored in a 1920 multiplied by 1080 resolution, and the storage format is a JPG picture format;

3.3) the Beidou satellite positioning terminal obtains the instant position information data of the unmanned aerial vehicle for detection, comprising X, Y, Z coordinates of the unmanned aerial vehicle, namely horizontal, vertical and height coordinates, and the representation form of the horizontal and vertical coordinates is represented according to the longitude and latitude positions.

4. The unmanned aerial vehicle-based road pavement disease positioning method according to claim 1, characterized in that: in the step 1.3), communication and data coding and transmission are completed through a fifth generation mobile communication technology; the specific process comprises the following steps:

4.1) storing the flight attitude, the working state and the flight height information of the unmanned aerial vehicle in a text file in txt format, wherein the text file is named according to the measurement time, and the specific naming format is' year/month/day-hour: and (2) minute: second: milliseconds "; the flight attitude, the working state and the flight height information of the unmanned aerial vehicle are stored in the text file through 9-digit digital codes, the first 4-digit numbers store the flight inclination angle of the unmanned aerial vehicle with radian as a unit, the 5-digit numbers represent the working state of the unmanned aerial vehicle, the normal working of the unmanned aerial vehicle is recorded as '1', the abnormal working is recorded as '2', and the 6-digit to 9-digit numbers store the flight height information of the unmanned aerial vehicle with 2-digit precision after a decimal point, wherein the unit is meter; a txt format text file representing the flight attitude, the working state and the flight height information of the unmanned aerial vehicle is uniformly stored in a folder named as 'txt-measurement date' by taking a day as a unit;

4.2) recoding and naming the high-resolution pavement image data collected by the high-resolution video collector carried by the unmanned aerial vehicle, wherein the naming format is a 34-bit character group consisting of numbers, English letters and hyphens; 1-18 bit characters represent unmanned aerial vehicle position information, wherein the first 6 bit characters represent unmanned aerial vehicle longitude coordinates, and are accurate to 2 bits after decimal point; 7 to 11 bits represent latitude coordinates of the unmanned aerial vehicle, and the latitude coordinates are accurate to 2 bits after decimal points; the height coordinate of the unmanned aerial vehicle is represented by 12 to 18 bits, the unit is m, and the height coordinate is accurate to 2 bits after a decimal point; position 19 is a hyphen; the 20-33 bit character represents the image acquisition time of the unmanned aerial vehicle, the naming format is 'year, month, day, hour, minute and second', the 34 th bit is the picture shooting serial number, and the picture is the picture order shot in the current second; the renamed images are uniformly stored in a folder named as 'image-measuring date' by taking a day as a unit;

4.3) through the 5G communication terminal installed on the unmanned aerial vehicle, the folder of txt files storing the unmanned aerial vehicle attitude, working state and flight height information and the folder where the recoded and named high-resolution pavement photos are located are sent to the data processing server by using the 5G mobile communication technology, and the data sending frequency is once every 24 hours.

5. The unmanned aerial vehicle-based road pavement disease positioning method according to claim 1, characterized in that: in the step 1.4), a pavement disease detection model based on an improved DPM model is utilized, the model is formed by connecting a DPM-1 model for detecting and repairing diseases and a DPM-2 model for detecting and not repairing diseases in series, the DPM-1 model is responsible for detecting three pavement disease types of transverse crack repairing, longitudinal crack repairing and net crack repairing, and the DPM-2 model is responsible for detecting and not repairing transverse crack, not repairing longitudinal crack, not repairing net crack and other four pavement diseases.

6. The method for positioning the road pavement diseases based on the unmanned aerial vehicle as claimed in claim 5, wherein the method comprises the following steps: in the step 1.4), a pavement disease detection model based on an improved DPM model is used for detecting and positioning diseases of the preprocessed high-resolution pavement image, and the specific detection and flow comprises the following steps:

6.1) preprocessing the high-resolution pavement image transmitted back by the unmanned aerial vehicle, wherein the preprocessing process comprises the steps of copying an original image, storing a copy in a folder named as 'detection date + detection', carrying out resize processing on the image copy in the folder, modifying the resolution into a 512 x 500 format, and renaming the image copy from 1 to small in numerical sequence to be used as a sample for roadside disease detection and positioning;

6.2) detecting the road surface diseases and detecting the road surface diseases of the positioning samples by using a road surface disease detection model based on an improved DPM model, wherein the specific detection process comprises the steps of firstly detecting an image to be detected by using the DPM-1 model, if the DPM-1 model detects that the diseases exist, ending the detection, and simultaneously storing the image into a folder corresponding to the disease type, wherein the detection result is that the road surface diseases exist. If the DPM-1 model does not detect the existence of the road surface diseases, the image is transferred to a DPM-2 model for detection, if the DPM-2 model detects the existence of the road surface diseases, the detection is finished, the detection result shows that the road surface diseases exist, and meanwhile, the image is stored in a folder corresponding to the types of the diseases; if the DPM-2 model does not detect the existence of the pavement diseases, the detection is finished, and the detection result shows that the pavement diseases do not exist;

6.3) if the detected result of the road image to be detected is that the road is damaged, positioning the position of the road damaged by a road damaged detection model based on an improved DPM model, wherein the specific positioning process is to compare the shooting time of high-resolution image data where the road damaged is and the longitude coordinate X of the unmanned aerial vehicle obtained by the Beidou satellite positioning system at the time₁Latitude coordinate Y₁And elevation information Z₁Unmanned aerial vehicle flight height information H recorded by unmanned aerial vehicle system₁And then, calculating the position of the road surface disease.

7. The method for positioning the road pavement diseases based on the unmanned aerial vehicle as claimed in claim 6, wherein the method comprises the following steps: calculating the position of the road pavement disease in the step 6.3), specifically, calculating a longitude coordinate X of the position of the road pavement disease according to a formula (1), calculating a latitude coordinate Y of the position of the road pavement disease according to a formula (2), and calculating an elevation coordinate Z of the position of the road pavement disease according to a formula (3)

X＝X₁Formula (1)

Y＝Y₁Formula (2)

Z＝Z₁-H₁Formula (3)

In the formula X₁Unmanned aerial vehicle longitude coordinate and unmanned aerial vehicle longitude coordinate Y obtained by representing Beidou satellite positioning system₁Unmanned aerial vehicle latitude coordinate, Z, obtained by representing Beidou satellite positioning system₁Unmanned aerial vehicle height coordinate H obtained by representing Beidou satellite positioning system₁And the unmanned aerial vehicle flying height information recorded by the unmanned aerial vehicle system is represented.

Technical Field

The invention relates to a method for positioning a road surface disease based on an unmanned aerial vehicle, in particular to a method for positioning a road surface disease based on an unmanned aerial vehicle.

Background

The 5G communication technology is an abbreviation of the fifth generation mobile communication technology. The 5G mobile communication technology is also a mobile communication technology using a digital cellular network, like 3G and 4G technologies already put into market, and compared with the conventional cellular data communication technology, the 5G mobile communication technology has the advantage that the data transmission speed and the transmission capacity are much higher than those of the conventional cellular network. The fastest data transmission speed of the commercial 5G mobile communication technology can reach 10Gbit/s, which is 100 times of the data transmission speed of the current 4G LTE cellular network. In addition, the 5G mobile communication technology also has the advantages of low network delay, high response speed, good adaptability with a mobile network terminal and the like. 6.2019, 6.6.8, the ministry of industry and informatization of the people's republic of China formally releases commercial 5G license plates to operators of China telecom, China Mobile, China Unicom, China radio and television, marking China to enter the commercial era of 5G communication technology.

Big dipper satellite positioning system, the english name is: the BeiDou Navigation Satellite System, BDS for short, is a new generation of Satellite positioning and Navigation System independently developed in china, and is a third global mature Satellite positioning and Navigation System that can be used in the united states of america and russian GLONASS. The Beidou satellite positioning navigation system consists of a space section, a ground section and a user section. The space section consists of an inclined earth orbit synchronous satellite, a middle circle earth orbit satellite and an earth static orbit satellite; the ground section comprises a plurality of ground stations such as a communication control station, a satellite system monitoring station and the like, and line operation management facility equipment; the user section comprises vehicle-mounted, ship-mounted, airborne and handheld Beidou satellite positioning terminals used by users of the Beidou satellite positioning navigation system, and other products such as chips, antennas and modules using or compatible with the Beidou satellite navigation system, corresponding application programs and services and the like. The Beidou satellite navigation has the advantages that compared with other satellite navigation systems, the Beidou satellite navigation has strong shielding resistance and interference resistance; the navigation signal of a plurality of frequency points can be provided, and the method has the advantages of higher positioning precision and the like.

The DPM target detection algorithm is a component-based detection algorithm, adopts a detection algorithm for detecting HOG characteristics of a target, improves the HOG characteristics, and combines a detection method of a sliding window and an SVM classifier. The DPM algorithm has good detection stability on the deformation of the target, so that the DPM algorithm is widely applied to image detection algorithms such as classification, segmentation, posture detection and the like of various targets, belongs to the current image detection algorithm, and is a mature algorithm.

The road pavement disease condition is one of important indexes for evaluating the health condition of a road, and the road pavement disease can be divided into a plurality of main disease types such as cracks, transverse cracks, longitudinal cracks, block cracks, pits, loose, tracks and the like. The traditional method for detecting and positioning the road pavement diseases mainly comprises the steps of obtaining pavement photos by using methods such as a detection vehicle, a laser radar and a detector for photographing, and then finishing the detection and classification of the road pavement diseases by using a method of manual judgment and positioning by workers. The method has the problems of high detection cost, low detection speed, large amount of manpower and material resources, and great influence on detection precision by factors such as detection level, working experience, physiological conditions and the like of detection personnel. The discovery, detection and positioning of the road disease condition by the current road surface maintenance management department are greatly limited, and the problem to be solved urgently exists in the current road maintenance management field.

Disclosure of Invention

Aiming at the defects of the existing pavement disease detection and positioning technology, the invention provides a method for positioning the pavement disease of the highway based on an unmanned aerial vehicle. Firstly, a low-altitude unmanned aerial vehicle system which is provided with a high-resolution video collector and is provided with a fifth-generation mobile communication technology terminal and a Beidou satellite positioning system terminal is constructed. The method comprises the steps that a low-altitude unmanned aerial vehicle system cruising above a road surface to be detected according to a certain height and flight speed shoots the road surface at regular time to obtain high-resolution road surface image data, and unmanned aerial vehicle transverse, longitudinal and high three-dimensional coordinate position information obtained by a Beidou satellite positioning system is fused; the fifth generation mobile communication technology, namely 5G technology is utilized to complete the functions of sending, recording and storing the pavement disease image and the pavement disease coordinate position data, and data transmission and information exchange between the unmanned aerial vehicle platform and the data processing server are realized; after the server obtains data, firstly, preprocessing is carried out on a road surface disease image, then, a road surface disease detection model based on improved DPM is utilized to carry out disease detection on a high-resolution road surface image obtained by a low-altitude unmanned aerial vehicle platform, three-dimensional characteristics of a road surface disease are obtained, whether the road surface to be detected is diseased or not is judged, and the position of the road surface disease is determined through unmanned aerial vehicle horizontal, vertical and high three-dimensional coordinate position information and unmanned aerial vehicle flight height obtained by a Beidou satellite positioning system.

The principle of the invention is as follows: the low-altitude unmanned aerial vehicle system cruising at a fixed height and a fixed speed can shoot a road surface to be detected according to a certain shooting frequency through high-resolution image acquisition equipment carried on the unmanned aerial vehicle, and the obtained high-resolution image data contains information such as texture, outline, color, fluctuation degree and the like of the road surface. Through 5G mobile communication technology, can realize transmitting the unmanned aerial vehicle position information, the state information that the unmanned aerial vehicle obtained the high resolution image data of road surface and big dipper satellite positioning terminal that unmanned aerial vehicle obtained to data processing server. After the data processing server preprocesses the pavement image, the pavement disease detection and positioning are carried out by a pavement disease detection model based on the improved DPM. The pavement detection model based on the improved DPM adopts a method of connecting double DPM detection models in series, and the DPM-1 model is responsible for detecting the pavement repairing diseases, including detection of three pavement disease types of pavement repairing transverse cracks, pavement repairing longitudinal cracks and pavement repairing net cracks; the DPM-2 model is responsible for detecting the diseases of the unrepaired pavement, and comprises detection of transverse unrepaired pavement cracks, longitudinal pavement cracks, net cracks and the like, wherein the detection of the four types of pavement diseases comprises detection of four types of pavement diseases. When detecting, firstly, detecting an image to be detected by using a DPM-1 model, if the DPM-1 model detects that a disease exists, ending the detection, wherein the detection result is that the road disease exists, if the DPM-1 model does not detect that the road disease exists, transferring the image to a DPM-2 model for detection, if the DPM-2 model detects that the disease exists, ending the detection, and the detection result is that the road disease exists; and if the DPM-2 model does not detect the existence of the pavement diseases, ending the detection, wherein the detection result is that the pavement diseases do not exist. And when the detection result shows that the road surface diseases exist, calculating the position of the road surface diseases according to the longitude and latitude, the elevation information and the flying height of the unmanned aerial vehicle corresponding to the detection time, and completing the positioning of the road surface diseases.

The technical scheme provided by the invention is as follows:

a method for positioning road pavement diseases based on an unmanned aerial vehicle comprises the following steps: the method comprises the technical links of unmanned aerial vehicle system construction, pavement and unmanned aerial vehicle data acquisition, data preprocessing, pavement detection model construction based on improved DPM, pavement disease detection and positioning and the like.

The unmanned aerial vehicle system is constructed by installing and integrating a high-resolution image collector, a 5G communication terminal and a Beidou satellite positioning module on a low-altitude unmanned aerial vehicle platform, and aims to construct a low-altitude unmanned aerial vehicle system with data acquisition, real-time positioning and data transmission functions. The data acquisition of the road surface and the unmanned aerial vehicle means that the high-resolution picture of the road surface and the information such as the flight height and the position of the unmanned aerial vehicle are obtained through the unmanned aerial vehicle, and the purpose is to provide data for the detection and the positioning of the road surface diseases. The data preprocessing refers to normalizing the size of the road surface photo before detection, so that the picture format meets the requirement of a detection model on the picture format. The construction of the road surface detection model based on the improved DPM comprises model construction, training data set construction and model training, and the aim is to enable the road surface detection model based on the improved DPM to have better identification precision and detection speed for road surface diseases. The pavement disease detection and positioning means that the pavement disease of the highway is detected and the position of the pavement disease of the highway is determined through a trained pavement detection model based on improved DPM.

The pavement disease detection method provided by the invention specifically comprises the following steps:

1) a low-altitude unmanned aerial vehicle system for detection is constructed, and a high-resolution video collector, a fifth-generation mobile communication terminal and a Beidou satellite positioning terminal need to be configured for the low-altitude unmanned aerial vehicle system. The high-resolution video collector refers to a flying side iXU 180 type high-resolution aerial survey camera installed on an unmanned aerial vehicle platform through an LEMO aerial interface, the fifth generation mobile communication terminal refers to 5G communication terminal equipment, and the Beidou satellite positioning system terminal refers to a UFirebairdUC 6226 type Beidou positioning chip;

2) the low-altitude unmanned aerial vehicle system constructed in the step 1) performs cruise flight on the road surface to be detected at the flight speed of 3m/s and the flight height of 7 m. The unmanned aerial vehicle platform collects high-resolution image data of the road surface according to the fixed frequency of taking a picture every 0.1 second, and the shot image is stored in a JPG image format. The save resolution is 1920 × 1080. The method comprises the steps that when an unmanned aerial vehicle system collects a highway pavement picture, a Beidou satellite positioning module obtains position information of the unmanned aerial vehicle;

3) installing a 5G communication terminal on the unmanned aerial vehicle, and transmitting the working state of the unmanned aerial vehicle, the position information of the unmanned aerial vehicle and the pavement disease image data acquired in the step 2) to a data processing server according to set naming, coding and transmission rules;

4) and after receiving the data transmitted in the step 3), the background processor performs preprocessing on the high-resolution pavement image transmitted back by the unmanned aerial vehicle. Meanwhile, a pavement disease detection model based on the improved DPM is trained;

5) carrying out pavement disease detection on the pavement image preprocessed in the step 4) through a pavement disease detection model formed by connecting DPM-1 and DPM-2 in series based on improved DPM detection models to obtain a pavement disease detection result;

6) and if the detection result in the step 5) indicates that the road surface disease exists, calculating to obtain the position of the road surface disease according to the position information and the flight height information of the unmanned aerial vehicle when the image data with the detected road surface disease exists is shot.

The method aims at detecting and positioning the road pavement diseases. Further, the unmanned aerial vehicle position information related in the step 2) specifically refers to horizontal and vertical coordinates and elevation information of the unmanned aerial vehicle, wherein the horizontal and vertical coordinate information of the unmanned aerial vehicle is represented by longitude and latitude coordinates of the unmanned aerial vehicle, and the elevation information of the unmanned aerial vehicle is represented by an elevation coordinate taking the 1985 national elevation standard as an elevation reference surface.

The method aims at detecting and positioning the road pavement diseases. Further, the data naming, encoding and transmission rules related in step 3) specifically refer to:

a) the flight attitude, the working state and the flight height information of the unmanned aerial vehicle are stored in a text file in a txt format, the text file is named according to the measurement time, and the specific naming format is' year/month/day-hour: and (2) minute: second: milliseconds ";

the flight attitude, the working state and the flight height information of the unmanned aerial vehicle are stored in the text file through 9-digit digital codes, the flight inclination angle of the unmanned aerial vehicle with radian as a unit is stored in the first 4-digit codes, the working state of the unmanned aerial vehicle is represented by the 5 th digit, the normal work of the unmanned aerial vehicle is recorded as '1', and the abnormal work is recorded as '2'. The 6 th to 9 th digits store the flying height information of the unmanned aerial vehicle which is accurate to 2 bits behind the decimal point, and the unit is meter. A txt format text file representing the flight attitude, the working state and the flight height information of the unmanned aerial vehicle is uniformly stored in a folder named as 'txt-measurement date' by taking a day as a unit;

b) recoding and naming the high-resolution road surface image data acquired by a high-resolution video acquisition device carried by the unmanned aerial vehicle, wherein the naming format is a 34-bit character group consisting of numbers, English letters and hyphens. 1-18 bit characters represent unmanned aerial vehicle position information, wherein the first 6 bit characters represent unmanned aerial vehicle longitude coordinates, and are accurate to 2 bits after decimal point; 7 to 11 bits represent latitude coordinates of the unmanned aerial vehicle, and the latitude coordinates are accurate to 2 bits after decimal points; the height coordinate of the unmanned aerial vehicle is represented by 12 to 18 bits, the unit is m, and the height coordinate is accurate to 2 bits after a decimal point; position 19 is a hyphen; the 20-to-33-bit character indicates the image acquisition time of the unmanned aerial vehicle, the naming format is 'year, month, day, hour, minute and second', the 34 th bit is the picture taking serial number, and the picture is indicated as the picture order taken in the current second. The renamed images are uniformly stored in a folder named as 'image-measuring date' by taking a day as a unit;

the method aims at detecting and positioning the road pavement diseases. Further, the specific contents of image preprocessing and training of the road surface disease detection model based on the improved DPM in the step 4) refer to:

a) the image preprocessing flow specifically comprises the steps of copying an original image, storing a copy in a folder named as detection date + detection, and carrying out resize processing on the image copy in the folder, wherein the resolution is modified into a 512 x 500 format, so that the size requirement of a pavement damage detection model based on improved DPM on a detection sample is met. At the same time, renaming the image copies according to the numerical sequence from 1 to small to large to be used as a sample for roadside disease detection and positioning;

b) the concrete process of training the pavement disease detection model based on the improved DPM refers to that the pavement disease detection model based on the improved DPM is subjected to model training through a certain university pavement disease data set which comprises pavement repairing transverse cracks, pavement repairing longitudinal cracks, pavement repairing net cracks, pavement non-repairing transverse cracks, pavement non-repairing longitudinal cracks, pavement non-repairing net cracks, other seven disease types, total data amount of which is more than 10000 pieces, and data of each disease type of which is more than 1000 pieces, so that the detection speed and the accuracy of the detection model meet the detection requirements. During model training, in the data of each disease type in a certain university pavement disease data set, 50 pieces of each type of disease data are selected as training positive samples, 80% of data are randomly extracted to be used in a model training flow training set, the rest 20% of data are used in a model training verification sample set, and 200 pieces of pavement photos without diseases are randomly selected to be used as training negative samples;

the method aims at detecting and positioning the road pavement diseases. Further, the road surface disease detection process involved in the step 5) specifically includes firstly detecting an image to be detected by using the DPM-1 model, if the DPM-1 model detects that a disease exists, ending the detection, and storing the image into a folder corresponding to the disease type, wherein the detection result indicates that the road surface disease exists. If the DPM-1 model does not detect the existence of the road surface diseases, the image is transferred to a DPM-2 model for detection, if the DPM-2 model detects the existence of the road surface diseases, the detection is finished, the detection result shows that the road surface diseases exist, and meanwhile, the image is stored in a folder corresponding to the types of the diseases; and if the DPM-2 model does not detect the existence of the pavement diseases, ending the detection, wherein the detection result is that the pavement diseases do not exist.

The method aims at detecting and positioning the road pavement diseases. Further, the positioning process of the pavement diseases related in the step 6) specifically refers to:

a) comparing the shooting time of the high-resolution image data where the road surface disease is located with the longitude coordinate X of the unmanned aerial vehicle obtained by the Beidou satellite positioning system at the time₁Latitude coordinate Y₁And elevation information Z₁Unmanned aerial vehicle flight height information H recorded by unmanned aerial vehicle system₁；

b) Calculating a longitude coordinate X of a position of the road surface defect according to the formula (1), calculating a latitude coordinate Y of the position of the road surface defect according to the formula (2), and calculating an elevation coordinate Z of the position of the road surface defect according to the formula (3):

X＝X₁formula (1)

Y＝Y₁Formula (2)

Z＝Z₁-H₁Formula (3)

c) And determining the position of the pavement disease according to the longitude coordinate, the latitude coordinate and the elevation coordinate of the pavement disease.

Compared with the prior art, the invention has the beneficial effects that:

compared with the traditional method for acquiring road surface images by means of shooting by means of a detection vehicle, aerial photography and manual shooting, the method has the advantages of being high in detection speed, stable in detection frequency, low in detection cost, low in manpower requirement, capable of performing all-weather operation and the like, can overcome the defects that the transmission speed is low, the transmission capacity is greatly limited and the like in the current 4G mobile communication technology by means of the 5G mobile communication technology, and can achieve rapid and large-capacity data transmission and information communication between a road surface disease detection site and a background processing server by means of the 5G mobile communication technology and equipment. By means of the Beidou satellite positioning module and the pavement damage detection model based on the improved DPM, the functions of automatically detecting and positioning pavement damage by means of the model can be realized, the detection precision is high, the detection speed is high, and the positioning is accurate; the method can solve the defects of low detection efficiency, poor detection precision, large manpower consumption, high manpower cost and larger influence of factors such as the technical level, the working experience, the physiological condition and the like on the detection effect of the current method for positioning the pavement diseases by detecting the pavement diseases through detecting the pavement diseases by depending on the detectors.

The method for positioning the road pavement diseases based on the unmanned aerial vehicle can realize full-flow automatic detection operation of acquiring and positioning the road pavement diseases by data acquisition of the road pavement data, transmitting the data, detecting the diseases and positioning the diseases, can greatly improve the level of road maintenance and management service, and reduces the cost of the road maintenance.

Drawings

Fig. 1 is a flowchart of a method for positioning a road pavement defect based on an unmanned aerial vehicle according to the present invention.

Fig. 2 is a detection schematic diagram of the positioning method for road pavement diseases based on the unmanned aerial vehicle provided by the invention.

Fig. 3 is a structural diagram of the unmanned aerial vehicle system constructed by the present invention.

Fig. 4 is a schematic diagram of an improved DPM detection algorithm.

Fig. 5 is a diagram of a road surface damage data set for university for road surface detection model training based on improved DPM.

Fig. 6 is a flowchart of road surface disease detection and positioning based on the improved DPM road surface detection model.

Detailed Description

The following further describes the implementation of the present invention by way of specific examples, with reference to the accompanying drawings, but not in any way limiting the scope of applicability of the present invention.

Fig. 1 is a flowchart of a method for positioning a road pavement disease based on an unmanned aerial vehicle, which includes the following steps:

step 1: installing an aeronautical survey camera iXU 180 type high resolution, a mobile communication terminal 5G and a UFirebig Dipper positioning chip UC6226 on the low-altitude unmanned aerial vehicle to construct a low-altitude unmanned aerial vehicle system;

step 2: the low-altitude unmanned aerial vehicle constructed in the step 1 cruises on the road surface of the road to be detected according to the flying speed of 3m/s and the flying height of 7m, and takes a high-definition picture of the road surface every 0.1 second, wherein the picture is stored in a format of 1920 multiplied by 1080; when height-limiting facilities exist on the road surface or the landform conditions are not allowed, the flying height of the unmanned aerial vehicle can be adjusted according to the actual situation;

and step 3: when the unmanned aerial vehicle system carries out the shooting process in the step 2, the data such as the flight height, the flight state and the like of the unmanned aerial vehicle are synchronously stored;

and 4, step 4: when the unmanned aerial vehicle system carries out the shooting process in the step 2, the airborne Beidou satellite positioning module acquires longitude, latitude and elevation information of the unmanned aerial vehicle in real time;

and 5: recoding and naming each item of data and information related in the step 2, the step 3 and the step 4;

step 6: the airborne 5G mobile communication terminal transmits the data and the information recoded and named in the step 5 to the data processing server;

and 7: the data processing server preprocesses the pavement disease image data to enable the pavement disease image data to meet the requirements of the detection model on format and resolution; the data preprocessing requires that an original image is copied, the copy is stored in a folder named as 'detection date + detection', the image copy in the folder is subjected to resize processing, the resolution is modified into a 512 x 500 format, and the image copy is renamed according to the numerical sequence from 1 to small and big. Meanwhile, the data processing server uses a road surface damage data set of a university to train a road surface damage model based on the improved DPM.

As shown in fig. 6, the basic principle of the DPM image feature detection algorithm is to, for any detected image, first extract the DPM features of the image, then perform gaussian pyramid upsampling on the input image, and then extract the DPM features of the sampled image again. And performing convolution processing on the DPM characteristic diagram of the original input sample image and the trained root filter to obtain a response diagram of the root filter. And performing convolution processing on the DPM characteristic diagram of the 2-time resolution image and the trained component filter to obtain a response diagram of the component filter. And then downsampling the fine Gaussian pyramid. The response graphs of the root filter and the component filter after operation have the same resolution, and then the final response graph is obtained through weighted average calculation processing.

The college pavement disease data set shown in fig. 5 includes seven types of pavement disease data, including transverse crack repair, longitudinal crack repair, net crack repair, transverse crack unrepaired pavement, longitudinal crack unrepaired pavement, net crack unrepaired pavement, and other types of pavement disease data, where each type of disease data exceeds 1000 pieces of disease data. During model training, 50 pieces of each type of disease data are selected as training positive samples, 80% of the disease data are randomly selected as a model training data set, the rest 20% of the disease data are selected as a model training verification set, and 200 pieces of road image data without road diseases are selected as training negative samples.

And 8: and 7, carrying out disease detection and positioning on the road surface image data preprocessed in the step 7 by using a road surface disease detection model based on improved DPM.

The specific flow of disease detection and location is shown in fig. 6, firstly, the preprocessed image data is detected by using a trained DPM-1 model to detect an image to be detected, three types of disease detection of transverse crack repair, longitudinal crack repair and net crack repair are carried out by the DPM-1 model, if the detection result is that a disease exists, the disease is located, if the detection result is that no disease exists, the data is transferred to the trained DPM-2 model to carry out four types of disease detection of transverse crack repair, longitudinal crack repair, network repair and the like, if the detection result is that a road disease exists, the road disease is located, and if the detection result is that no disease exists, the detection result of the model is output as that no road disease exists.