CN116500643A

CN116500643A - Road deformation disease detection method and system based on single-line laser point cloud

Info

Publication number: CN116500643A
Application number: CN202310434332.XA
Authority: CN
Inventors: 朱俊清; 蒋舜; 卜天翔; 马涛; 张伟光; 王志鹏
Original assignee: Southeast University
Current assignee: Southeast University
Priority date: 2023-04-21
Filing date: 2023-04-21
Publication date: 2023-07-28

Abstract

The invention discloses a method and a system for detecting highway deformation diseases based on single-line laser point clouds, which are characterized in that firstly, pose of an unmanned aerial vehicle at different moments is recorded, and laser radar acquires road point cloud data in a cross section form and then fits a road surface reference cross section; fitting a road reference surface and a road curved surface model, determining the position of a disease point cloud, and dividing a 10m pavement area; and finally, calculating the maximum depth and deformation volume parameters of the diseases from a three-dimensional angle, and drawing up an index and evaluating the highway deformation diseases. According to the invention, the calculation of the deformation disease depth is realized from a three-dimensional angle, and compared with the traditional detection of the pavement deformation disease, the method has higher precision, and the three-dimensional evaluation index of the pavement deformation disease is provided, so that the evaluation method of the deformation disease is enriched, and the position of the road section with the disease to be maintained is primarily positioned.

Description

Road deformation disease detection method and system based on single-line laser point cloud

Technical Field

The invention belongs to the field of road engineering, and particularly relates to a road deformation disease detection method and system based on a single-line laser point cloud.

Background

Along with the development of economy at a high speed, asphalt roads gradually become the main road surface type of expressways in China. However, the increasing traffic volume, overload and heavy load problems are increasingly aggravated, so that various damages are caused to asphalt pavements, particularly to highway asphalt pavements.

The deformation disease is a disease with obvious deformation caused by external force influence on the pavement, and is one of common diseases of asphalt highway pavement. On the one hand, the deformation diseases cause a great amount of deformation of the road surface, and the abrupt change of the elevation of the driving is caused, so that the comfortableness and the stability of the driving are influenced, on the other hand, the deformation diseases cause the bulge or the dent of the road surface structure, a series of other diseases are caused, the drainage problem is caused, the strength and the service life of the road surface are reduced, and the safety of the road surface is reduced. Therefore, it is necessary to perform systematic and comprehensive evaluation on road surface deformation disasters and to make targeted maintenance measures.

At present, the evaluation indexes of deformation diseases at home and abroad are often based on depth and width indexes of cross section diseases, and no unified deformation disease evaluation indexes exist, meanwhile, due to the defect of a pavement detection technology, the precision and the density of collected data are difficult to ensure, and the damage of the deformation diseases is rarely considered from a three-dimensional angle.

In the past, due to the lag of electronic technology, manual methods are the main method for detecting pavement diseases, and the depth and width of ruts are mainly measured through a ruler, however, manual measurement depends on experience and has large errors, and are eliminated nowadays. In recent years, with the continuous development of electronic technology, deformation disease detection methods based on deep learning or image processing are gradually put into use, mainly using a detection vehicle as a mounting device, and realizing detection of road deformation disease by arranging a plurality of multi-line laser radars. However, this approach still has some drawbacks, such as: the method is easy to be influenced by other vehicles on the road, the cost and maintenance cost for detecting the vehicles and the multi-line radar are high, and the vehicle speed requirement is strict. Although the accuracy of the data can be ensured by the way of measuring the road point cloud by the laser radar, the mounting device limits the exertion of radar performance. Compared with a detection vehicle, the unmanned aerial vehicle can achieve lower cost, lighter weight and higher efficiency, and is a better laser radar mounting device in theory. Meanwhile, the unmanned aerial vehicle detects the deformation of the road surface from high altitude, and the single-line radar is enough to meet the detection requirement. Compared with the multi-line radar, the single-line laser radar has the advantages of simpler and simpler data processing and higher operand, and after certain point cloud data processing, the single-line radar can generate high-density and accurate point cloud.

Disclosure of Invention

In order to solve the technical problems mentioned in the background art, the invention provides a highway deformation disease detection method and system based on a single-line laser point cloud.

In order to achieve the technical purpose, the technical scheme of the invention is as follows:

a highway deformation disease detection method based on a single-line laser point cloud comprises the following steps:

s1, a single-line laser radar based on triangular ranging is mounted on an unmanned aerial vehicle, pose of the unmanned aerial vehicle at different moments is recorded on the way that the unmanned aerial vehicle flies along a specified route, and road point cloud data in a cross section form are collected.

S2, noise interference exists in the data acquisition process, so that a series of isolated points and interference points are generated. Denoising the collected road point cloud data, effectively removing outliers to obtain single-frame point clouds of the road, screening out outer lane point clouds, fitting a standard cross section in a straight line mode, inserting the inner lane single-frame point clouds according to the slope of the fitted straight line, and splicing the insertion point clouds and the outer lane point clouds to obtain the three-dimensional point clouds of the road.

S3, dividing the road into pavement areas with specific intervals according to the three-dimensional point cloud in the step S2, fitting a standard running surface of the road through a thin plate spline interpolation method, positioning areas where diseases are located on the divided pavement areas so as to conveniently study the severity of the diseases, and determining an actual curved surface equation of the road through the thin plate spline interpolation method.

S4, calculating the maximum subsidence of the divided disease areas, the maximum protrusion of the disease and the corresponding deformation volume, determining the severity of the disease, and judging whether maintenance is needed.

Further, the specific steps of step S2 are as follows:

s201, adopting a radius filtering algorithm, making a circle on the center of each point cloud, reserving points contained in the circle if the number of points is larger than a fixed value, deleting the points if the number of points is smaller than the fixed value, adopting a statistical filtering algorithm, solving the average value of distances from each point cloud to all points in the field K, solving the average value mu and the variance sigma of all the average values, setting mu+nv as a threshold value, and taking the threshold value as a specified multiple, thus preliminarily eliminating the noise data of the point cloud of the road, and obtaining single-frame point clouds of the road; and extracting accurate transverse surface point cloud based on the road transverse slope and the road lowest point height Cheng Queding road surface point cloud range.

S202, screening point clouds 0.5m-3m outside the cross section direction in the single frame point clouds by adopting a method for extracting the point clouds outside the single frame data based on the point cloud coordinates, using the point clouds as emergency lane point clouds, fitting the emergency lane point clouds in a straight line mode to form a standard cross section, and inserting the single frame point clouds inside the emergency lane according to the point spacing in the road range according to the slope of the straight line to simulate a standard running surface.

S203, because the unmanned aerial vehicle on which the radar is mounted can deviate and twist due to wind power factors in the operation running process, the specific running track and position of the mounting equipment can be known by means of the sensor, and the three-dimensional position of the acquired cross section can be positioned. Based on the single-line radar pose recorded in the step S1, the cross-section point cloud extracted in the step S201, and the single-frame point cloud of the step S202, the three-dimensional road point cloud data and the reference driving surface three-dimensional point cloud data are formed by stitching the point cloud data of the continuous frames through the coordinate transformation from NED (North East Down) to LLA (longitude, latitude, altitude, longitude and latitude high coordinate system).

Further, the specific steps of step S3 are as follows:

s301, based on the emergency lane point cloud acquired in the step S202, extracting point cloud plane coordinates, fitting the whole line type of a road in a cubic curve mode, reflecting the trend of the road, dividing the curve into 10m small sections, extracting x and y coordinates at the sections and a normal plane of the curve at the sections, taking points, which are positioned at the inner sides of 20m away from the trend curve, of the normal plane on an xOy projection straight line, and determining a road area clamped by the normal plane, so that the division of a 10m section of road area is realized on a reference driving plane; the normal plane point taking and sitting marks are as follows:

QD＝[[x ₀₁ ，y ₀₁ ，x ₀₂ ，y ₀₂ ]，......]

Wherein x is ₀₁ Representing the abscissa, y, of the intersection of the trend curve with the first normal plane ₀₁ An ordinate, x, representing the intersection of the trend curve with the first normal plane ₀₂ An abscissa, y, representing the point taken inside the first normal plane ₀₂ Representing the ordinate of the point taken inside the first normal plane.

S302, because the point cloud of the emergency lane relatively perfectly stores the road elevation information of the designed place, the subsequent operation is convenient, the reference running surface of the road is fitted by adopting a plane fitting algorithm based on the three-dimensional point cloud data of the reference running surface in the step S203, and the parameter A related to the plane equation of the reference running surface is obtained ₀ 、A ₁ 、A ₂ The specific formula is as follows:

z＝A ₀ x+A ₁ y+A ₂

wherein A is ₀ 、A ₁ 、A ₂ The parameters of the plane with respect to the x, y coordinates and the intercept of the z coordinate when the x, y coordinates are equal to 0, respectively.

S303, calculating the elevation difference between the three-dimensional road point cloud data and the reference plane corresponding to the xOy coordinate based on the three-dimensional road point cloud data in the step S203 and the reference road surface equation of the road in the step S302.

S304, dividing an xOy coordinate of a road reference surface into grids of 1dm multiplied by 1dm based on a grid method, and if the absolute value of a point cloud elevation difference is larger than a preset threshold value and a point position is positioned below the reference surface, the point position is a sinking disease point, and a grid area where the point position is positioned is a sinking disease area; if the absolute value of the point cloud elevation difference is larger than a preset threshold value and the point position is positioned above the reference plane, the point position is a raised disease point, and the grid area where the point position is positioned is a raised disease area; thus preliminarily dividing the sinking disease area Sa and the raised disease area Sb.

S305, fitting a three-dimensional curved surface equation of the road disease area by a thin plate spline interpolation method based on the subsidence disease area and the raised disease area point cloud, wherein the specific formula is as follows:

U(x)＝r ² lnr

wherein p (x, y) is any point on the curved surface, U (x) is a radial basis function, ||p-p _i I represents the distance of point p to a control point, knowing control points 1, 2, 3 …, N, ω _i Representing the weighting of different radial basis groups, m ₀ 、m ₁ 、m ₂ Is a parameter of the plane.

S306, establishing a control point matrix and a height matrix of point cloud data, wherein the specific formula is as follows:

(1) Control point matrix

Where n is the number of control points, and the second and third columns represent the (x, y) coordinates of the control points.

(2) Height matrix

Wherein v is ₁ To v _n Representing the coordinates in the z-direction of each control point.

S307, calculating radial basis function values of any two control points, wherein the specific formula is as follows:

wherein r is _ij Represents the distance between control points i and j, U (r _ij ) Corresponding distance r to radial basis function _ij Is a value of (2).

S308, defining a matrix L as:

the above matrix has the following relationship:

Y＝L*(ω ₁ ，…ω _N ，m ₀ ，m ₁ ，m ₂ ) ^T 。

substituting a condition function about a control point based on a thin plate spline interpolation principle, calculating all parameters of a road three-dimensional curved surface equation, and finishing interpolation, wherein a specific matrix is as follows:

Wherein omega _ij Weighting the j-th radial basis on the i-th segment, m _i0 ，m _i1 ，m _i2 For m on the ith segment ₀ ，m ₁ ，m ₂ Coefficients.

Further, the specific steps of step S4 are as follows:

s401, extracting a point location M, N corresponding to the maximum elevation difference in the step S3, taking a grid where the M, N point location is located and 8 grids around the grid as a selected area, and solving a gradient of the point location M, N based on a steepest descent/ascent method, wherein the specific formula is as follows:

f＝z _{actual practice is that of} -z _Datum

Wherein f is the difference between the elevation of the actual curved surface of the road and the quasi-running surface of the road bed, and z _{Actual practice is that of} Z is the elevation of the actual curved surface of the road _Datum The road bed is used as a quasi-driving surface; the known minimum point set in the selected disease area is M, and the point obtained by k times of iteration is recorded as M ^k ，P ^k Represents M ^k Direction d of maximum change rate of point curved surface ^k Representing the gradient.

S402, obtaining z according to the formula of the step S305 _Datum Value of z _Datum The value formula solves the differentiation to obtain the following formula:

d ^k the final expression is:

wherein N is the number of control points fitted by the reference surface, Q is the number of control points fitted by the actual curved surface of the road, M ^k (x, y) is the starting point of the current iteration; x is x _i ，y _i Coordinates corresponding to the control points; r is M ^k (x, y) distance to each control point.

S403, solving the elevation of the next point by taking 0.01m as a step length, and iterating until an extreme point appears, wherein the elevation difference between the extreme point and a reference plane corresponding to the xOy coordinate is the disease depth/disease height on the pavement area.

S404, regularly taking 9 points in each disease grid, and calculating the average value of the elevationGet->The product of the area of the grid and the area of the gridDeformation volume V _i The specific formula is as follows:

wherein S is the area of the grid 4cm ² ；V _i The deformation volume of the grid with deformation disease is from left to right and then from bottom to top in each road surface area.

Adding all the deformed volumes of the grids to obtain the subsidence/protrusion volume V of the pavement area _a ，V _b 。

And S405, based on the pavement disease depth/disease height and the pavement area subsidence/protrusion volume, completing three-dimensional evaluation of the pavement deformation disease.

Furthermore, the invention also provides a highway deformation disease detection system based on the single-line laser point cloud, which comprises

And the information acquisition module is used for loading the single-line laser radar on the unmanned aerial vehicle, recording the poses of the unmanned aerial vehicle at different moments on the way that the unmanned aerial vehicle flies along a specified route, and acquiring the road point cloud data in the cross section form.

The road three-dimensional point cloud acquisition module is used for denoising the acquired road point cloud data to obtain a single-frame point cloud of a road, screening an outer lane point cloud, fitting a standard cross section in a straight line mode, inserting the inner lane single-frame point cloud according to the slope of the fitted straight line, and splicing the insertion point cloud and the outer lane point cloud to obtain the three-dimensional point cloud of the road.

The disease area positioning module is used for dividing the road into road surface areas with specific intervals according to the road three-dimensional point cloud, fitting the standard running surface of the road through a thin plate spline interpolation method, positioning the area where the disease is located on the divided road surface areas, and determining an actual curved surface equation of the road.

And the disease area deformation volume calculation module is used for calculating the maximum subsidence of the divided disease area, the maximum protrusion of the disease and the corresponding deformation volume.

Further, in the road three-dimensional point cloud acquisition module, the specific steps are as follows:

step 1, adopting a radius filtering algorithm, making a circle on the center of each point cloud, reserving points contained in the circle if the number of points is larger than a fixed value, deleting the points if the number of points is smaller than the fixed value, adopting a statistical filtering algorithm, solving the average value of distances from each point cloud to all points in the field K, solving the average value mu and the variance sigma of all the average values, setting mu+n sigma as a threshold value, and taking n as a specified multiple, and taking the threshold value as a screening value, thereby preliminarily eliminating the noise data of the point cloud of the road to obtain single-frame point cloud of the road; and extracting accurate transverse surface point cloud based on the road transverse slope and the road lowest point height Cheng Queding road surface point cloud range.

And 2, screening point clouds which are 0.5m-3m outside the cross section direction in the single frame point clouds by adopting a method for extracting the point clouds outside the single frame data based on the point cloud coordinates, using the point clouds as emergency lane point clouds, fitting the emergency lane point clouds in a straight line mode to form a standard cross section, and inserting the single frame point clouds inside the emergency lane according to the point spacing in the road range according to the slope of the straight line.

And 3, based on single-line radar pose, cross-section point cloud and single-frame point cloud, splicing point cloud data of continuous frames through coordinate transformation from NED to LLA coordinate system to form three-dimensional road point cloud data and reference driving surface three-dimensional point cloud data.

Further, in the disease area positioning module, the specific steps are as follows:

step 1, based on emergency lane point cloud, extracting point cloud plane coordinates, fitting the whole line type of a road in a cubic curve mode, dividing the curve into 10m small sections, extracting x and y coordinates at the sections and a normal plane of the curve at the sections, taking points, which are positioned at the inner side of 20m away from a trend curve, of the normal plane on an xOy projection straight line, and determining a road area clamped by the normal plane, thereby realizing the division of a 10m section of road surface area on a reference driving surface; the normal plane point taking and sitting marks are as follows:

QD＝[[x ₀₁ ，y ₀₁ ，x ₀₂ ，y ₀₂ ]，......]

Step 2, fitting the reference running surface of the road by adopting a plane fitting algorithm based on the three-dimensional point cloud data of the reference running surface to obtain a parameter A of a plane equation of the reference running surface ₀ 、A ₁ 、A ₂ The specific formula is as follows:

z＝A ₀ x+A ₁ y+A ₂

And 3, calculating the elevation difference between the three-dimensional road point cloud data and the reference surface corresponding to the xOy coordinate based on the three-dimensional road point cloud data and the road reference running surface equation.

Step 4, dividing the xOy coordinate of the road reference surface into grids of 1dm multiplied by 1dm based on a grid method, and if the absolute value of the point cloud elevation difference is larger than a preset threshold value and a point position is positioned below the reference surface, the point position is a sinking disease point, and the grid area where the point position is positioned is a sinking disease area; if the absolute value of the point cloud elevation difference is larger than a preset threshold value and the point position is positioned above the reference plane, the point position is a raised disease point, and the grid area where the point position is positioned is a raised disease area; thus preliminarily dividing the sinking disease area Sa and the raised disease area Sb.

Step 5, fitting a three-dimensional curved surface equation of the road disease area by a thin plate spline interpolation method based on the point clouds of the subsidence disease area and the raised disease area, wherein the specific formula is as follows:

U(x)=r ² lnr

Step 6, establishing a control point matrix and a height matrix of the point cloud data, wherein the specific formula is as follows:

(1) Control point matrix

(2) Height matrix

Step 7, calculating radial basis function values of any two control points, wherein the specific formula is as follows:

Step 8, defining a matrix L as:

the above matrix has the following relationship:

Y＝L*(ω ₁ ，…ω _N ，m ₀ ，m ₁ ，m ₂ ) ^T 。

Wherein omega _ij Weighting the j-th radial basis on the i-th segment, m _i0 ,m _i1 ，m _i2 For m on the ith segment ₀ ，m ₁ ，m ₂ Coefficients.

Further, in the disease area deformation volume calculation module, the specific steps are as follows:

step 1, extracting a point location M, N corresponding to the maximum elevation difference, taking a grid where the M, N point location is located and 8 grids around the grid as a selected area, and solving a gradient of the point location M, N based on a steepest descent/ascent method, wherein a specific formula is as follows:

f＝z _{actual practice is that of} -z _Datum

d ^k the final expression is:

And 3, solving the elevation of the next point by taking 0.01m as a step length, and iterating until an extreme point appears, wherein the elevation difference between the extreme point and a reference plane corresponding to the xOy coordinate is the disease depth/disease height on the pavement area.

Step 4, regularly taking 9 points in each disease grid, and calculating the average value of the elevationGet->The product of the mesh area and the deformation volume V of the mesh _i The specific formula is as follows:

And 5, based on the pavement disease depth/disease height and the pavement area subsidence/protrusion volume, completing three-dimensional evaluation of the pavement deformation disease.

Compared with the prior art, the invention adopts the technical proposal and has the following remarkable technical effects:

the invention fits the equation of the road curved surface and the road bed quasi-running surface, simply and effectively positions the road disease area, completes the calculation of the deformation disease depth and the deformation volume on the three-dimensional level, and provides comprehensive and effective evaluation indexes by reference to the specification. Based on unmanned aerial vehicle laser radar acquisition technology, adopt lighter weight, low cost's loading equipment, adopt the less single line laser radar of data volume simultaneously, designed the three-dimensional deformation disease extraction algorithm of low operand, reduced the operation time, increased the efficiency of patrolling and examining, ensured deformation disease degree of depth calculation's precision to realize the evaluation of deformation class disease from three-dimensional angle.

Meanwhile, the invention fits the road surface equation by using a thin plate spline interpolation method, writes a gradient formula of the equation on the basis of the equation, combines the principle of the steepest descent/ascent method, further improves the accuracy of deformation disease depth detection, provides comprehensive deformation disease evaluation indexes by referring to the past specifications, and provides certain guiding assistance for the selection of subsequent maintenance road sections and the calculation of the use amount of maintenance materials.

Drawings

FIG. 1 is a schematic overall flow chart of an embodiment of the present invention.

Fig. 2 is a schematic diagram of a cross-sectional point cloud of a frame acquired by a single-line laser radar according to an embodiment of the present invention.

Fig. 3 is a three-dimensional point cloud image after point cloud stitching according to an embodiment of the present invention.

Fig. 4 is a schematic diagram of calculating a detection depth of a point cloud disease area according to an embodiment of the present invention.

Fig. 5 is a schematic view of a micro-grid point of view for calculating a lesion volume according to an embodiment of the present invention.

Detailed Description

In order to make the technical means, creation characteristics, achievement purposes and effects realized by the invention easy to understand, the following describes in detail a highway deformation disease detection method based on a single-line laser point cloud with reference to the embodiments and the drawings.

The invention discloses a highway deformation disease detection method based on a single-line laser point cloud, wherein a flow chart is shown in figure 1, and the method specifically comprises the following steps:

S1, in the embodiment, the acquisition equipment is a large-area m600pro unmanned aerial vehicle, the laser radar is a dock lms511 single-line laser radar, the laser radar is mounted below the unmanned aerial vehicle, the unmanned aerial vehicle is located above a road to shoot, and the flight speed is set to be 1m/S to carry out inspection.

During the flight of the unmanned aerial vehicle along a prescribed route, the single-line radar detects at the frequency of 25hz, the angular resolution of each detection is 0.1667 degrees, the number of data points acquired by each frame is about 1100, the pose of the unmanned aerial vehicle at different moments is recorded, and the laser radar acquires road point cloud data of 10m sections in the cross section form.

S2, denoising collected road point cloud data, effectively removing outliers to obtain single-frame point clouds of a road, screening outer lane point clouds, fitting a standard cross section in a straight line mode, inserting the inner lane single-frame point clouds according to the slope of the fitted straight line, and splicing the insertion point clouds and the outer lane point clouds to obtain three-dimensional point clouds of the road, wherein the method comprises the following specific steps of:

s201, road point cloud data are road cross section data which are acquired by single-line laser radar and are more than thousand frames, as shown in fig. 2. And using open3d to read a single-frame point cloud, firstly selecting the lowest point in the point cloud data, wherein the lowest point is the lowest point of the road surface point cloud, and primarily screening the road surface data by taking the point elevation to +30cm as a boundary. The method comprises the steps of adopting a radius filtering algorithm, taking the center of each point cloud as a circle, reserving the points contained in the circle when the number of the points is larger than a fixed value, deleting the points when the number of the points is smaller than the fixed value, setting the fixed value to be 2cm in the embodiment, adopting a statistical filtering algorithm, solving the average value of distances from each point cloud to all points in the field, solving the average value mu and variance sigma of all average values, setting mu+n sigma as a threshold value, taking n as a specified multiple, taking the threshold value as the value of the radius filtering algorithm, screening the point cloud, and preliminarily eliminating single-frame point cloud noise data by combining with the normal vector of the road surface elevation and the point cloud, so as to obtain the single-frame point cloud of the road.

The normal vector of the point cloud is the normal vector of a plane formed by fitting a certain point and the point cloud in the adjacent area, and the direction of the normal vector is mainly distributed in the z coordinate for the road surface. The normal vector is calculated by the normal function of open3d, and the modulo length of the normal vector is corrected to 1, and the direction is the positive direction of the z coordinate.

If the z coordinate data of the normal vector is larger than 0.9, the point is a disease-free road point; and extracting the maximum and minimum values of coordinates of the single-frame point cloud in the y direction, wherein all the point clouds in the range are regarded as road points. Based on the road cross slope and the road lowest point height Cheng Queding road surface point cloud range, accurate cross-section point cloud is extracted and used for splicing three-dimensional point cloud data.

S202, since the emergency lane mostly does not generate deformation diseases in the road running process, the cross section of the outside emergency lane is taken as a straight line to be prolonged, a result similar to the standard cross section can be obtained, and the emergency lane can be regarded as the standard cross section under the condition of lacking original design data.

And (3) reading the processed road surface point cloud by using open3d, screening the point cloud which is 0.5m-3m outside the cross section direction in the single frame point cloud by adopting a method for extracting the point cloud outside the single frame data based on the point cloud coordinates, using the point cloud as an emergency lane point cloud, fitting the emergency lane point cloud in a straight line mode to form a standard cross section, and inserting the single frame point cloud inside the emergency lane according to the slope of the straight line and the point spacing of 2cm in the road range to simulate the standard running surface. Recording the inserted point cloud and the number of the frames in an array DY:

Dy= [ [ x, y, z, (frame number) ], … … ].

S203, based on the single-line radar pose recorded in the step S1, the cross-section point cloud extracted in the step S201 and the single-frame point cloud of the step S202, the point cloud data of the continuous frames are spliced through the coordinate transformation from NED to LLA, so that three-dimensional road point cloud data and reference driving surface three-dimensional point cloud data are formed.

In order to facilitate point cloud modeling, pile number setting and region division, single-line radar pose recorded in step S1, cross-section point cloud extracted in step S201 and single-frame point cloud of step S202 are converted into an array DY1 through numpy, and a column of data is newly added after all point cloud data to mark the number of frames in which the point cloud is originally located. DY1 is:

dy1= [ [ x, y, z, (frame number) ], and.

S3, dividing the road into pavement areas with an interval of 10m according to the three-dimensional point cloud in the step S2, fitting a standard running surface of the road by a thin plate spline interpolation method, positioning areas where diseases are located on the divided pavement areas so as to conveniently study the severity of the diseases, and determining an actual curved surface equation of the road by the thin plate spline interpolation method, wherein the specific steps are as follows:

fig. 3 is a three-dimensional point cloud image of the present embodiment, and a black part of the circled area is a disease point cloud area of the present embodiment.

S301, based on the plane coordinates of the emergency lane point cloud acquired in the step S202, using a python library function to complete a cubic curve mode to fit the whole line type of a road, marking the pile number of the curve, dividing the curve into 10m segments, extracting x and y coordinates of the segment and a normal plane of the curve of the segment, taking a point of the normal plane, which is at the inner side of 20m away from the trend curve on an xOy projection straight line, determining a road area clamped by the normal plane, and recording the coordinates in a numpy array, thereby realizing division of a 10m section of road area on a reference running surface. The normal plane point taking and sitting marks are as follows:

QD＝[[x ₀₁ ，y ₀₁ ，x ₀₂ ，y ₀₂ ]，......]

S302, adopting a plane fitting algorithm based on the three-dimensional point cloud data of the reference driving surface in the step S203Fitting the reference running surface of the road to obtain a parameter A related to the reference running surface ₀ 、A ₁ 、A ₂ The specific formula is as follows:

z=A ₀ x+A ₁ y+A ₂

S303, calculating the elevation difference between the three-dimensional road point cloud data and a reference plane equation under the corresponding xOy coordinate based on the three-dimensional road point cloud data and the reference road plane equation of the S302 road. If the value is negative, it is stated that the point is located below the reference plane, otherwise it is located above the reference plane.

S304, dividing the xOy coordinates of the road reference surface into grids of 1dm multiplied by 1dm based on a grid method.

In the array DY1 described in step S203, a new row of data is added to determine a point cloud without deformation disease. If the absolute value of the point cloud elevation difference is larger than a preset threshold value and the point position is positioned below the reference plane, the point position is a sinking disease point, and the grid area where the point position is positioned is a sinking disease area; if the absolute value of the point cloud elevation difference is larger than a preset threshold value and the point is located above the reference plane, the point is a raised disease point, and the grid area where the point is located is a raised disease area. The threshold value set in this example is + -5 mm.

The array DY1 is added with a row of data to obtain DY2:

dy2= [ [ x, y, z, (frame number), (no/sink/bump) ]

In order to divide the position of the point cloud and facilitate subsequent operation, the disease point cloud is divided in a grid mode. The information of the grid is represented by an array WG, specifically:

Wg= [ [ x, y, (sink/bump), (road surface area ordinal) ], ]

Wherein the first two columns of columns represent the coordinate positions of the grid, for example (2, 3) represents an area surrounded by x=2cm, x=4cm, y=4cm and y=6cm in the coordinate system; the third column of data is used for marking the subsidence and the protrusion of the grid, and the fourth column of the array DY2 is used for marking the subsidence and the protrusion of the gridData, if point clouds in the grid are not subsidence point clouds or protrusion point clouds, no disease area exists in the grid; if the point cloud in the grid has sinkage or protrusion, the grid is a sinkage disease area Sa or protrusion disease area Sb, and the sinkage volume V is calculated subsequently _a Bump volume V _b The method comprises the steps of carrying out a first treatment on the surface of the The fourth column of data is the road surface area ordinal number of the grid with the disease, the road surface area of the grid can be determined according to the position relation between the grid and the normal plane (the straight line after being projected to the xOy plane) in the step S301, so that the fourth column of data of the grid is assigned, and for the grid without the disease, the calculation is simplified, and the default is 0.

The method for judging the position relation between the grid and the normal plane comprises the following steps: taking the midpoint G of the disease grid, and calculating the vector from the point G to the point recorded in the QD array; calculating the vector inner product of the adjacent normal planes from the initial surface, and if the vector inner product is negative, G is arranged between the adjacent normal planes corresponding to the vectors, so that the pavement area where the grid is positioned is determined; if the vector inner product is positive, G is outside the adjacent normal plane corresponding to the vector, and the vector inner product of the next group of normal planes and G is calculated.

U(x)＝r ² lnr

S306, the three-dimensional curved surface equation has +3 parameters of point cloud number, a control point matrix and a height matrix of point cloud data are established, and the specific formula is as follows:

(1) Control point matrix

(2) Height matrix

S307, calculating radial basis function values of any two control points, substituting the radial basis function values into a condition function of thin plate spline interpolation, so as to calculate all parameters of an equation and complete interpolation, wherein the specific formula is as follows:

S308, defining a matrix L as:

the above matrix has the following relationship:

Y＝L*(ω ₁ ，…ω _N ，m ₀ ，m ₁ ，m ₂ ) ^T 。

S4, calculating the maximum subsidence of the divided disease areas, the maximum protrusion of the disease and the corresponding deformation volume, determining the severity of the disease, and judging whether maintenance is needed or not, wherein the specific steps are as follows:

s401, solving points M and N corresponding to maximum and minimum values of the elevation difference, grids where the points M and N are located and the elevation difference by adopting max and min functions of python, wherein an extreme value in the vicinity of the point M, N is the maximum value of deformation on the pavement area.

And taking the grid where the point position M, N is and 8 grids around the point position M, N as a selected area, and solving an extremum for the actual model of the road in the area by using a steepest descent/ascent method, wherein the extremum is the disease depth/disease height on the road surface area. The gradient from point M, N is solved, as shown in fig. 4, the left graph selects the grid containing the maximum elevation difference point, and the right graph uses the steepest descent/ascent method in the region to solve the maximum elevation difference point of the model. The specific formula is as follows:

f＝z _{actual practice is that of} -z _Datum

Wherein f is the difference between the elevation of the actual curved surface of the road and the quasi-running surface of the road bed, and z _{Actual practice is that of} Z is the elevation of the actual curved surface of the road _Datum The road bed is used as a quasi-driving surface; the known minimum point set in the selected disease area is M, and the point obtained by k times of iteration is recorded as M ^k ，P ^k Represents M ^k Direction d of maximum change rate of point curved surface ^k Representing the gradient;

d ^k the final expression is:

S403, in order to ensure the solving precision, the searching step length is lambda=1cm, and M is the sum of ^k In the direction d ^k Search for M ^k+1 ＝M ^k +λd ^k . And (3) checking whether the modulus of the gradient of the original point position is smaller than 0.01, if so, the point is an extreme point, and if not, iterating until the modulus of the gradient is smaller than 0.01, and ending the iteration. Outputs f and M ^k I.e. the maximum value H of the deformation depth _a Or H _b Coordinates corresponding to the maximum value. The maximum depth of disease calculated by this method in this example was 82.9mm.

S404, regularly taking 9 points in each disease grid, wherein the taking method is shown in figure 5, and calculating the elevation mean valueTaking outThe product of the mesh area and the deformation volume V of the mesh _i The specific formula is as follows:

wherein,,is the average value of the elevation sum of the nine points, S is the grid area of 4cm ² ；V _i The deformation volume of the grid with deformation disease is from left to right and then from bottom to top in each road surface area.

If V _i If positive, the grid calculates the volume of the protrusion; if V _i Negative, a settled volume is obtained. According to V _i The positive and negative of the road surface area and the deformation volume of the grid are added to the whole grid, namely the subsidence/protrusion volume V of the road surface area _a ，V _b 。

The volume index V _a ，V _b The large road section possibly develops serious diseases in the future, and plays a certain role in helping and guiding the selective maintenance of the subsequent road.

The calculated sinking disease volume of this example is 0.1532m ³ No raised disease was detected.

The evaluation indexes of deformation diseases such as a pack, a wave, a rut, a sink and the like are mostly divided into 5mm, 15mm and 25 mm. Maximum deformation depth H of the invention _a ，H _b Likewise, the deformation severity was divided by 5mm, 15mm, 25mm as boundaries, as shown in Table 1. The difference is that the index of the common deformation disease is the distance from the virtual drawing method or the virtual ruler method to the lowest point, the maximum deformation depth H of the invention _a ，H _b The distance from the road reference surface to the disease minimum point is fit. Thus, for the same road section, the maximum deformation depth H _a ，H _b The calculated value is relatively smaller.

TABLE 1 disease severity score table

In the detection of the actual road surface deformation disease, H is calculated for each road section by referring to Table 1 _a ，H _b ，V _a ，V _b And the severity is listed in table 2 for evaluating the deformation of the whole road. Wherein H is _a ，H _b The severity of the road deformation disease is determined, va, vb are the volumes of the sinkage and the protrusion. For the volume index V _a ，V _b The two can provide certain information support for the scheme of maintenance and monitoring. For road sections with middle and low severity, if a larger deformation volume exists, the possibility that diseases of the road sections are subsequently developed into diseases with high severity is larger, and important attention is paid to follow-up monitoring and maintenance, and remarks are marked.

TABLE 2 depth of deformation and volume summary table for pavement area

The embodiment of the invention also provides a highway deformation disease detection system based on the single-line laser point cloud, which comprises an information acquisition module, a road three-dimensional point cloud acquisition module, a disease area positioning module, a disease area deformation volume calculation module and a computer program capable of running on a processor. It should be noted that each module in the above system corresponds to a specific step of the method provided by the embodiment of the present invention, and has a corresponding functional module and beneficial effect of executing the method. Technical details not described in detail in this embodiment may be found in the methods provided in the embodiments of the present invention.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein. The solutions in the embodiments of the present application may be implemented in various computer languages, for example, object-oriented programming language Java, and an transliterated scripting language JavaScript, etc.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiments and all such alterations and modifications as fall within the scope of the application.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present application without departing from the spirit or scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims and the equivalents thereof, the present application is intended to cover such modifications and variations.

Claims

1. The highway deformation disease detection method based on the single-line laser point cloud is characterized by comprising the following steps of:

s1, a single-line laser radar is mounted on an unmanned aerial vehicle, pose of the unmanned aerial vehicle at different moments is recorded in the process that the unmanned aerial vehicle flies along a specified route, and road point cloud data in a cross section form are collected;

s2, denoising the collected road point cloud data to obtain a single-frame point cloud of the road, screening an outer lane point cloud, fitting a standard cross section in a straight line mode, inserting an inner lane single-frame point cloud according to the slope of the fitted straight line, and splicing the inserted point cloud and the outer lane point cloud to obtain a three-dimensional point cloud of the road;

s3, dividing the road into pavement areas with specific intervals according to the three-dimensional point cloud in the step S2, fitting a standard running surface of the road by a thin plate spline interpolation method, positioning areas where diseases are located on the divided pavement areas, and determining an actual curved surface equation of the road;

S4, calculating the maximum subsidence of the divided disease areas, the maximum protrusion of the disease and the corresponding deformation volume.

2. The method for detecting highway deformation diseases based on single-line laser point cloud according to claim 1, wherein the specific steps of step S2 are as follows:

s201, adopting a radius filtering algorithm, making a circle on the center of each point cloud, reserving points contained in the circle if the number of points is larger than a fixed value, deleting the points if the number of points is smaller than the fixed value, adopting a statistical filtering algorithm, solving the average value of all the point distances in the field for each point cloud, solving the average value mu and the variance sigma of all the average values, setting mu+n as a threshold value, and taking the threshold value as a specified multiple, thus preliminarily eliminating the noise data of the point cloud data of the road, and obtaining single-frame point cloud of the road; extracting accurate transverse plane point cloud based on the road transverse slope and the road lowest point height Cheng Queding road surface point cloud range;

s202, screening point clouds 0.5m-3m outside the cross section direction in the single frame point clouds by adopting a method for extracting the point clouds outside the single frame data based on the point cloud coordinates, using the point clouds as emergency lane point clouds, fitting the emergency lane point clouds in a straight line mode to form a standard cross section, and inserting the single frame point clouds inside the emergency lane according to the point spacing in the road range according to the slope of the straight line;

3. The method for detecting highway deformation diseases based on single-line laser point cloud according to claim 2, wherein the specific steps of step S3 are as follows:

s301, based on the emergency lane point cloud acquired in the step S202, extracting point cloud plane coordinates, fitting the whole line type of a road in a cubic curve mode, dividing the curve into 10m small sections, extracting x and y coordinates at the section and a normal plane of the curve at the section, taking points, which are far from the inner side of a trend curve by 20m, of the normal plane on an xOy projection straight line, and determining a road area clamped by the normal plane, so that the division of a 10m section of road area on a reference driving plane is realized; the normal plane point taking and sitting marks are as follows:

QD＝[[X ₀₁ ,y ₀₁ ，x ₀₂ ,y ₀₂ ]，......]

wherein x is ₀₁ Representing the abscissa, y, of the intersection of the trend curve with the first normal plane ₀₁ An ordinate, x, representing the intersection of the trend curve with the first normal plane ₀₂ An abscissa, y, representing the point taken inside the first normal plane ₀₂ An ordinate representing a point taken inboard of the first normal plane;

s302, fitting a reference running surface of a road by adopting a plane fitting algorithm based on the three-dimensional point cloud data of the reference running surface in the step S203 to obtain a parameter A of a plane equation of the reference running surface ₀ 、A ₁ 、A ₂ The specific formula is as follows:

z＝A _o x+A ₁ y+A ₂

wherein A is ₀ 、A ₁ 、A ₂ Parameters of the plane on x and y coordinates are respectively the intercept of the z coordinate when the x and y coordinates of the plane are equal to 0;

s303, calculating the elevation difference between the three-dimensional road point cloud data and a reference plane corresponding to xOy coordinates based on the three-dimensional road point cloud data in the step S203 and the reference road surface equation of the road in the step S302;

s304, dividing an xOy coordinate of a road reference surface into grids of 1dm multiplied by 1dm based on a grid method, and if the absolute value of a point cloud elevation difference is larger than a preset threshold value and a point position is positioned below the reference surface, the point position is a sinking disease point, and a grid area where the point position is positioned is a sinking disease area; if the absolute value of the point cloud elevation difference is larger than a preset threshold value and the point position is positioned above the reference plane, the point position is a raised disease point, and the grid area where the point position is positioned is a raised disease area; thus preliminarily dividing the sinking disease area Sa and the raised disease area Sb;

U(x)＝r ² ln r

Wherein p (x, y) is any point on the curved surface, U (x) is a radial basis function, ||p-p _i I represents the distance of point p to a control point, knowing control points 1, 2, 3 …, N, ω _i Representing the weighting of different radial basis groups, m ₀ 、m ₁ 、m ₂ Is a parameter of the plane;

(1) Control point matrix

Wherein n is the number of control points, and the second and third columns represent the (x, y) coordinates of the control points;

(2) Height matrix

Wherein v is ₁ To v _n Representing the coordinates in the z direction of each control point;

wherein r is _ij Represents the distance between control points i and j, U (r _ij ) Corresponding distance r to radial basis function _ij Is a value of (2);

s308, defining a matrix L as:

the above matrix has the following relationship:

Y＝L*(ω ₁ ，…ω _N ，m ₀ ，m ₁ ，m ₂ ) ^T ；

4. The method for detecting highway deformation diseases based on single-line laser point cloud according to claim 3, wherein the specific steps of step S4 are as follows:

S401, extracting a point location M, N corresponding to the maximum elevation difference, taking a grid where the M, N point location is and 8 grids around the grid as a selected area, and solving a gradient of the point location M, N based on a steepest descent/ascent method, wherein the specific formula is as follows:

f＝z _{actual practice is that of} -z _Datum

s402, according to step S305Formula, get z _Datum Value of z _Datum The value formula solves the differentiation to obtain the following formula:

d ^k the final expression is:

wherein N is the number of control points fitted by the reference surface, Q is the number of control points fitted by the actual curved surface of the road, M ^k (x, y) is the starting point of the current iteration; x is x _i ，y _i Coordinates corresponding to the control points; r is M ^k (x, y) distance to each control point;

s403, solving the elevation of the next point by taking 0.01m as a step length, and iterating until an extreme point appears, wherein the elevation difference between the extreme point and a reference plane corresponding to xOy coordinates is the disease depth/disease height on the pavement area;

S404, regularly taking 9 points in each disease grid, and calculating the average value of the elevationGet->The product of the mesh area and the deformation volume V of the mesh _i The specific formula is as follows:

wherein S is the area of the grid 4cm ² ；V _i The deformation volume of the grid with deformation diseases from left to right and from bottom to top in each road surface area;

adding all the deformed volumes of the grids to obtain the subsidence/protrusion volume V of the pavement area _a ，V _b ；

5. Highway deformation disease detecting system based on single line laser point cloud, characterized by comprising

The information acquisition module is used for loading the single-line laser radar on the unmanned aerial vehicle, recording the poses of the unmanned aerial vehicle at different moments in the middle of the flight of the unmanned aerial vehicle along a specified route, and acquiring road point cloud data in a cross section form;

the road three-dimensional point cloud acquisition module is used for denoising the acquired road point cloud data to obtain a single-frame point cloud of a road, screening an outer lane point cloud, fitting a standard cross section in a straight line mode, inserting the single-frame point cloud of an inner lane according to the slope of the fitted straight line, and splicing the insertion point cloud and the outer lane point cloud to obtain a three-dimensional point cloud of the road;

The disease area positioning module is used for dividing the road into road surface areas with specific intervals according to the road three-dimensional point cloud, fitting a standard running surface of the road through a thin plate spline interpolation method, positioning areas where the disease is located on the divided road surface areas and determining an actual curved surface equation of the road;

6. The highway deformation disease detection system based on the single-line laser point cloud according to claim 5, wherein the road three-dimensional point cloud acquisition module comprises the following specific steps:

step 1, adopting a radius filtering algorithm, making a circle on the center of each point cloud, reserving points contained in the circle if the number of points is larger than a fixed value, deleting the points if the number of points is smaller than the fixed value, adopting a statistical filtering algorithm, solving the average value of distances from each point cloud to all points in the field K, solving the average value mu and the variance sigma of all the average values, setting mu+n sigma as a threshold value, and taking n as a specified multiple, and taking the threshold value as a screening value, thereby preliminarily eliminating the noise data of the point cloud of the road to obtain single-frame point cloud of the road; extracting accurate transverse plane point cloud based on the road transverse slope and the road lowest point height Cheng Queding road surface point cloud range;

Step 2, screening point clouds which are 0.5m-3m outside the cross section direction in the single frame point clouds by adopting a method for extracting the point clouds outside the single frame data based on the point cloud coordinates, using the point clouds as emergency lane point clouds, fitting the emergency lane point clouds in a straight line mode to form a standard cross section, and inserting the single frame point clouds inside the emergency lane according to the point spacing in the road range according to the slope of the straight line;

7. The highway deformation disease detection system based on the single-line laser point cloud as set forth in claim 5, wherein the disease area positioning module comprises the following specific steps:

QD＝[[x ₀₁ ,y ₀₁ ，x ₀₂ ,y ₀₂ ]，......]

z＝A ₀ x+A ₁ y+A ₂

wherein A is ₀ 、A ₁ 、A ₂ The parameters of the plane about x and y coordinates and the intercept of the z coordinate when the x and y coordinates are equal to 0 are respectively;

step 3, calculating the elevation difference between the three-dimensional road point cloud data and a reference plane corresponding to the xOy coordinate based on the three-dimensional road point cloud data and a road reference running surface equation;

step 4, dividing the xOy coordinate of the road reference surface into grids of 1dm multiplied by 1dm based on a grid method, and if the absolute value of the point cloud elevation difference is larger than a preset threshold value and a point position is positioned below the reference surface, the point position is a sinking disease point, and the grid area where the point position is positioned is a sinking disease area; if the absolute value of the point cloud elevation difference is larger than a preset threshold value and the point position is positioned above the reference plane, the point position is a raised disease point, and the grid area where the point position is positioned is a raised disease area; thus preliminarily dividing the sinking disease area Sa and the raised disease area Sb;

U(x)＝r ² lnr

(1) Control point matrix

(2) Height matrix

step 8, defining a matrix L as:

the above matrix has the following relationship:

Y＝L*(ω ₁ ，·…ω _N ，m ₀ ，m ₁ ，m ₂ ) ^T ；

8. The highway deformation disease detection system based on single-line laser point cloud according to claim 5, wherein in the disease area deformation volume calculation module, the specific steps are as follows:

f＝z _{actual practice is that of} -z _Datum

s402, obtaining according to the formula of the step S305z _Datum Value of z _Datum The value formula solves the differentiation to obtain the following formula:

d ^k the final expression is:

step 3, solving the elevation of the next point by taking 0.01m as a step length, and iterating until an extreme point appears, wherein the elevation difference between the extreme point and a reference plane corresponding to the xOy coordinate is the disease depth/disease height on the pavement area;

wherein,,is the average value of the elevation sum of the nine points, S is the grid area of 4cm ² ；V _i The deformation volume of the grid with deformation diseases is from left to right and then from bottom to top in each road surface area;