CN116205865B - A pavement uniformity analysis method based on 2D and 3D fusion data - Google Patents

Abstract

The invention discloses a pavement uniformity analysis method based on two-dimensional and three-dimensional fusion data, which comprises the following steps: 1) Acquiring a two-dimensional image and three-dimensional depth data by using a three-dimensional near-ground integrated texture photographing device; 2) The upper computer carries out deformation calibration on the three-dimensional depth data to obtain a calibrated depth set Z ₁ The method comprises the steps of carrying out a first treatment on the surface of the 3) Based on the gray value of the two-dimensional image data, image segmentation based on the maximum inter-class gap is carried out on the two-dimensional image, so that a plurality of connected areas are obtained; 4) Calculating kurtosis P corresponding to the connected region _A Degree of deviation S _A Extremely poor D _A Information entropy E _A Self-similarity SSI _A The method comprises the steps of carrying out a first treatment on the surface of the 5) Calculating kurtosis P corresponding to the connected volume statistical set V _V Degree of deviation S _V Extremely poor D _V Information entropy E _V Self-similarity SSI _V The method comprises the steps of carrying out a first treatment on the surface of the 6) The average distance index ADI, maximum distance index MDI, and road local uniformity SSI, which characterize the road uniformity, are calculated. The invention can effectively monitor the uniformity of the pavement and provide a powerful criterion for construction quality evaluation.

Description

A road surface uniformity analysis method based on 2D and 3D fusion data

Technical Field

The invention relates to the field of pavement uniformity analysis, and in particular to a pavement uniformity analysis method based on two-dimensional and three-dimensional fusion data.

Background Art

The uniformity of the pavement significantly affects the occurrence of early pavement diseases and its long-term service life. It is mainly affected by the degree of segregation during the mixing, transportation and paving of the mixture and the uniformity of the pavement compaction.

Therefore, pavement uniformity is an important indicator for testing construction quality and an important method for testing the early status of roads; especially for special mixture compositions such as recycled materials, uniformity testing is particularly necessary.

However, existing construction acceptance often only tests macro indicators such as flatness, and does not consider the uniformity of the mixture.

Therefore, a pavement uniformity analysis method that integrates two-dimensional images and pavement texture depth distribution information is needed.

Summary of the invention

The purpose of the present invention is to provide a road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, comprising the following steps:

1) The road surface to be analyzed is divided into multiple large grids, and each large grid is divided into multiple sub-grids; the number of specifications and types of these sub-grids is recorded as n; each sub-grid is used as a scanning detection and analysis area; n is a positive integer;

2) Using a 3D near-ground integrated texture photography device to obtain 2D images and 3D depth data;

The three-dimensional near-ground integrated texture photography device includes a three-wheeled vehicle, a vehicle-mounted brain module, a three-dimensional camera sensor, an encoder, and a sensor clamping device;

The sensor clamping device is a beam-type gantry, which is provided with n clamping positions of different heights, corresponding to the scanning fields of n sub-grids of different specifications respectively;

A plurality of three-dimensional camera sensors are fixed on the positions of the sensor fixing device, respectively scan the corresponding scanning detection and analysis areas, and transmit the road surface image data to the vehicle brain module;

The road surface image data includes a two-dimensional image and three-dimensional depth data;

The encoder acquires the rotation frequency of the wheel axle of the three-wheeled trolley and transmits it to the on-board brain module;

The on-board brain module drives the three-wheeled vehicle to travel on the road surface to be analyzed and controls the travel trajectory of the three-wheeled vehicle;

The on-board brain module controls the operation of the three-dimensional camera sensor and the encoder;

The on-board brain module receives road image data and the rotation frequency of the wheel axle of the three-wheeled trolley, and uploads them to the host computer;

3) The host computer performs deformation calibration on the three-dimensional depth data to obtain a calibrated depth set Z ₁ ;

The calibrated three-dimensional depth information set _Z1 is as follows:

Z＝aX+bXY+cY+dX ² +e (5)

Z ₁ = Z ₀ - Z (6)

Where a, b, c, d are depth information coefficients; e is a constant; _Z0 is the three-dimensional depth information set before calibration; Z is the fitting optimization plane; X, Y are the two-dimensional coordinates of the current measurement point;

4) Based on the grayscale value of the two-dimensional image data, the two-dimensional image is segmented based on the maximum inter-class difference to obtain multiple connected regions;

5) Calculate the number of pixels A in the connected area, and based on the hierarchical statistical data of the connected area, calculate the kurtosis _PA , skewness _SA , range D _A , information entropy _EA and self-similarity SSI _A corresponding to the connected area;

6) Determine the depth threshold of the maximum inter-class gap based on the depth information;

7) Based on the depth threshold, count the connected volumes of convexities and concavities, and establish a connected volume statistics set V;

Calculate the kurtosis P _V , skewness S _V , range D _V , information entropy _EV and self-similarity SSI _V corresponding to the connected volume statistics set V;

8) Establish a feature set of N measuring points on the road surface to be analyzed; one measuring point corresponds to one scanning detection and analysis area;

Among them, the feature set of the i-th measuring point is as follows:

Fi＝{P _Ai, S _Ai, D _Ai, E _Ai, SSI _Ai, P _Vi, S _Vi, D _Vi , E _Vi , SSI _Vi }; (1)

Wherein, i=1, 2, ..., N;

9) Calculate the Euclidean distance d _ij between each pair of measuring points;

10) Calculate the average distance index ADI, maximum distance index MDI and local road uniformity SSI used to characterize road surface uniformity, namely:

MDI＝max _i，j≤N d _ij -min _i，j≤N d _ij (3)

Where N is the total number of measuring points; SSI _i is the local uniformity index.

Furthermore, the three-dimensional near-ground integrated texture photography device also includes a remote control module for the vehicle;

The car remote control module generates the car remote control data and transmits it to the on-board brain module, thereby realizing the remote control of the three-wheeled car.

Furthermore, the vehicle-mounted brain module is mounted on a three-wheeled vehicle, and includes a vehicle control system, a photography control module, a local data storage module, and a communication module;

The vehicle control system drives the three-wheeled vehicle to travel on the road surface to be analyzed and controls the travel trajectory of the three-wheeled vehicle;

The photography control module is used to control the acquisition mode of the three-dimensional camera sensor; the acquisition mode includes a uniform speed acquisition mode and an encoder-based speed coordination control mode;

The data local storage module is used to receive and store road surface image data collected by the three-dimensional camera sensor;

The communication module is used to receive data from the remote control module of the car and transmit it to the car control system.

Furthermore, the three-wheeled vehicle also includes a flexible trailer hook;

The flexible trailer hook is used to connect the three-wheeled trolley to the trailer, so that the trailer can move with the three-wheeled trolley;

When the trailer moves with the three-wheeled vehicle, the front wheels of the three-wheeled vehicle move upward and are retracted.

Further, the three-dimensional camera sensor includes a camera sensor and a line laser sub-sensor;

The line laser sub-sensor is used to monitor the three-dimensional depth information of the ground;

The camera sensor is used to monitor the two-dimensional image of the ground.

Furthermore, the kurtosis P(J), skewness S(J), range D(J), information entropy E(J) and self-similarity SSI(J) are as follows:

D(J)＝max J-min J (9)

E(J)＝-∑ _J∈J P(j)log ₂ j (10)

SSI _i =SSI(I _i ) _i SSI(H _i ) _i (12)

Wherein, J represents the data matrix composed of the number of pixels in the connected area A, the data matrix composed of the statistical set of connected volumes V, the data matrix composed of the 2D image brightness data I _i , and the data matrix composed of the 3D depth data H _i, respectively; m and n are the numbers of the corresponding subsets based on the equal division of the existing data matrix; B is the number of equal divisions; α, β, and γ are constants; μ represents the mean, and σ represents the standard deviation; SSI(I _i ) _i represents the local uniformity of the road corresponding to the 2D image brightness data I _i ; SSI(H) _i represents the local uniformity of the road corresponding to the 3D depth data H _i ;

Among them, the parameters l(m, n), c(m, n), and t(m, n) are as follows:

c ₁ =(K ₁ *L) ² (17)

c ₂ =(K ₂ *L) ² (18)

In the formula, c ₁ , c ₂ , c ₃ are intermediate parameters; K ₁ , K ₂ are constants; L is the length of the matrix numerical interval; _{μ m} , μ _n are means; σ _mn is the standard deviation.

Furthermore, the Euclidean distance d _ij is as follows:

Among them, d _ij is the Euclidean distance between two measuring points, _fik , _fjk is the value of the kth index of the feature set of measuring points i and j.

Furthermore, the average distance index ADI represents the overall uniformity level of the road section, the maximum distance index MDI represents the extreme unevenness level, and the local uniformity situation of the road SSI represents the local unevenness level;

The larger the values of ADI and MDI, the worse the uniformity; the value range of SSI is 0 to 1, and the larger the value, the better the uniformity.

Further, the step of performing image segmentation based on the maximum inter-class difference on the two-dimensional image includes:

4.1) Taking all pixel grayscale data within the measurement range as statistical objects, solve the pixel grayscale distribution probability P _r (K _rt ) corresponding to different grayscale levels i;

4.2) Calculate the inter-class variance of the pixel grayscale distribution corresponding to any two different grayscale levels;

Among them, the inter-class variance between the pixel grayscale cluster corresponding to grayscale r and the pixel grayscale cluster corresponding to grayscale t is As shown below:

Wherein, m _G is the average gray value of the measurement range; m _r (K _rt ) and m _t (K _rt ) are the average gray values of the pixel gray clusters corresponding to gray level r and gray level t; r≠t; r, t＝1, 2, …, L; L is the number of gray levels;

4.3) Solve the inter-class variance Reach the maximum grayscale threshold K _rtmax ;

4.4) Segment the two-dimensional image based on all grayscale thresholds.

Further, the step of solving the depth threshold of the maximum inter-class gap based on the depth information includes:

6.1) Take all depth data within the measurement range as statistical objects and solve the distribution probability P _g (h _fg ) corresponding to different depth levels g;

6.2) Calculate the inter-class variance of the depth distribution corresponding to any two different depths;

Among them, the inter-class variance of the deep clusters corresponding to the depth level g and the deep clusters corresponding to the depth level f is As shown below:

Wherein, m _h is the average depth of the measurement range; m _f (K _fg ) and m _g (K _fg ) are the average depth values of the depth clusters corresponding to depth level f and depth level g; f ≠ g; f, g = 1, 2, …, H; H is the number of depth levels;

6.3) Solve the between-class variance The maximum depth threshold _hfgmax is reached.

The technical effect of the present invention is unquestionable. The present invention is based on two- and three-dimensional road texture acquisition technology, integrates and analyzes two-dimensional images and road texture depth distribution information, and constructs a non-destructive automatic detection method for road uniformity, porosity, etc. This method can effectively monitor road uniformity and provide a powerful basis for construction quality assessment.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a schematic diagram of image monitoring;

Figure 2 is a schematic diagram of the operation of a three-wheeled vehicle;

FIG3 is a schematic diagram of two-dimensional brightness data Ii (grayscale image);

FIG4 is a schematic diagram of three-dimensional depth information data Hi;

Figure 5 shows the image segmentation result based on the maximum inter-class gap;

Figure 6 shows the depth threshold for solving the maximum inter-class gap based on depth information.

DETAILED DESCRIPTION

The present invention is further described below in conjunction with the embodiments, but it should not be understood that the above subject matter of the present invention is limited to the following embodiments. Without departing from the above technical ideas of the present invention, various substitutions and changes are made according to the common technical knowledge and customary means in the art, which should all be included in the protection scope of the present invention.

Embodiment 1:

Referring to FIG. 1 to FIG. 6 , a road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data is characterized by comprising the following steps:

1) Divide the road surface to be analyzed into multiple large grids, and divide each large grid into multiple sub-grids; the number of specifications and types of these sub-grids is recorded as n; each sub-grid is used as a scanning detection and analysis area;

2) Using a 3D near-ground integrated texture photography device to obtain 2D images and 3D depth data;

The three-dimensional near-ground integrated texture photography device includes a three-wheeled vehicle, a vehicle-mounted brain module, a three-dimensional camera sensor, an encoder, and a sensor clamping device;

The sensor clamping device is a beam-type gantry, which is provided with n clamping positions of different heights, corresponding to the scanning fields of n sub-grids of different specifications respectively;

A plurality of three-dimensional camera sensors are fixed on the positions of the sensor fixing device, respectively scan the corresponding scanning detection and analysis areas, and transmit the road surface image data to the vehicle brain module;

The road surface image data includes a two-dimensional image and three-dimensional depth data;

The encoder acquires the rotation frequency of the wheel axle of the three-wheeled trolley and transmits it to the on-board brain module;

The on-board brain module drives the three-wheeled vehicle to travel on the road surface to be analyzed and controls the travel trajectory of the three-wheeled vehicle;

The on-board brain module controls the operation of the three-dimensional camera sensor and the encoder;

The on-board brain module receives road image data and the rotation frequency of the wheel axle of the three-wheeled trolley, and uploads them to the host computer;

3) The host computer performs deformation calibration on the three-dimensional depth data to obtain a calibrated depth set Z ₁ ;

4) Based on the grayscale value of the two-dimensional image data, the two-dimensional image is segmented based on the maximum inter-class difference to obtain multiple connected regions. The specific steps are as follows:

First, all pixel grayscale data within the measurement range are taken as statistical objects. L represents the corresponding grayscale level, usually between 0 and 255. The sum of the number of pixels at each grayscale level is the total number of pixels in the image.

Secondly, solve the distribution probability _Pi corresponding to different gray levels i, set K as the segmentation threshold to be solved, the average gray values of the number of pixels corresponding to the two groups of gray levels higher than K and lower than K are _m1 (K) and _m2 (K), and the probabilities are _P1 (K) and _P2 (K), respectively. The average gray value of the entire measurement range (image range) is _mG .

The between-class variance is Solve so that The maximum K value K* is used as the grayscale threshold for image segmentation and binarization.

Based on the grayscale threshold, the two-dimensional image is segmented.

5) Calculate the number of pixels A in the connected area, and based on the hierarchical statistical data of the connected area, calculate the kurtosis _PA , skewness _SA , range D _A , information entropy _EA and self-similarity SSI _A corresponding to the connected area;

6) Determine the depth threshold of the maximum inter-class gap based on the depth information; the specific steps are as follows:

First, all depth data within the measurement range are taken as statistical objects, H represents the corresponding depth value distribution interval, and the sum of the number of pixels at each depth level is the total number of pixels in the image.

Secondly, solve the distribution probability _Ph corresponding to different depths h, set h* as the depth plane segmentation threshold to be solved, the plane depths of the two groups of grayscales corresponding to the number of pixels above h* and below h* are _m1 (h) and _m2 (h), and the probabilities are _P1 (h) and _P2 (h), respectively. The average depth value of the entire measurement range (image range) is _mh .

The between-class variance is Solve so that The maximum h value h* is used as the depth threshold for image segmentation binarization.

7) Based on the depth threshold, count the connected volumes of convexities and concavities, and establish a connected volume statistics set V;

Calculate the kurtosis P _V , skewness S _V , range D _V , information entropy _EV and self-similarity SSI _V corresponding to the connected volume statistics set V;

8) Establish a feature set of N measuring points on the road surface to be analyzed; one measuring point corresponds to one scanning detection and analysis area.

Among them, the feature set of the i-th point is as follows:

Fi＝{P _Ai, S _Ai, D _Ai, E _Ai, SSI _Ai, P _Vi, S _Vi, D _Vi , E _Vi , SSI _Vi }; (1)

Wherein, i=1, 2, ..., N;

9) Calculate the Euclidean distance d _ij between each pair of measuring points;

10) Calculate the average distance index ADI, maximum distance index MDI and local road uniformity SSI used to characterize road surface uniformity, namely:

MDI＝max _i，j≤N d _ij -min _i，j≤N d _ij (3)

Where N is the total number of measuring points; SSI _i is the local uniformity index.

The three-dimensional near-ground integrated texture photography device also includes a trolley remote control module;

The car remote control module generates the car remote control data and transmits it to the on-board brain module, thereby realizing the remote control of the three-wheeled car.

The vehicle-mounted brain module is mounted on a three-wheeled vehicle, and includes a vehicle control system, a photography control module, a local data storage module, and a communication module;

The vehicle control system drives the three-wheeled vehicle to travel on the road surface to be analyzed and controls the travel trajectory of the three-wheeled vehicle;

The photography control module is used to control the acquisition mode of the three-dimensional camera sensor; the acquisition mode includes a uniform acquisition mode and an encoder-based speed coordination control mode. The uniform acquisition mode sets the acquisition rate; the coordinated acquisition mode sets the scanning sampling interval.

The data local storage module is used to receive and store road surface image data collected by the three-dimensional camera sensor;

The communication module is used to receive data from the remote control module of the car and transmit it to the car control system.

The three-wheeled vehicle also includes a flexible trailer hook;

The flexible trailer hook is used to connect the three-wheeled trolley to the trailer, so that the trailer can move with the three-wheeled trolley;

When the trailer moves with the three-wheeled vehicle, the front wheels of the three-wheeled vehicle move upward and are retracted.

The three-dimensional imaging sensor comprises an imaging sensor and a line laser sub-sensor;

The line laser sub-sensor is used to monitor the three-dimensional depth information of the ground;

The camera sensor is used to monitor the two-dimensional image of the ground.

In step 3), the calibrated three-dimensional depth information set _Z1 is as follows:

Z＝aX+bXY+cY+dX ² +e (5)

Z ₁ = Z ₀ - Z (6)

Where a, b, c, d are coefficients; e is a constant; _Z0 is the three-dimensional depth information set before calibration; Z is the fitting optimization plane; X and Y are the two-dimensional coordinates of the current measuring point.

Kurtosis P(J), skewness S(J), range D(J), information entropy E(J) and self-similarity SSI(J) are as follows:

D(J)＝max J-min J (9)

E(J)＝-∑ _j∈J P(j)log ₂ j (10)

SSI _i =SSI(I _i ) _i SSI(H _i ) _i (12)

Where J represents the data matrix composed of the number of pixels in the connected area A, the data matrix composed of the connected volume statistics set V, the data matrix composed of the two-dimensional image brightness data I _i , and the data matrix composed of the three-dimensional depth data _Hi ; m and n are the numbers of the corresponding subsets based on the equal division of the existing data matrix; B is the number of equal divisions; α, β, and γ are constants; μ represents the mean, and σ represents the standard deviation.

Among them, the parameters l(m, n), c(m, n), and t(m, n) are as follows:

c ₁ =(K ₁ *L) ² (17)

c ₂ =(K ₂ *L) ² (18)

Wherein, c ₁ , c ₂ , and c ₃ are intermediate parameters.

When J represents the data matrix composed of the number of pixels in the connected area A, α＝0, β＝1, γ＝1;

When J represents the data matrix composed of the connected volume statistics set V, α＝1, β＝1, γ＝1.

The Euclidean distance d _ij is as follows:

Among them, d _ij is the Euclidean distance between two measuring points, _fik , _fjk is the value of the kth index of the feature set of measuring points i and j.

The average distance index ADI represents the overall uniformity level of the road section, the maximum distance index MDI represents the extreme unevenness level, and the local uniformity of the road SSI represents the local unevenness level;

The larger the values of ADI and MDI, the worse the uniformity; the value range of SSI is 0 to 1, and the larger the value, the better the uniformity.

Embodiment 2:

Referring to FIG. 1 to FIG. 6 , a road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data includes the following steps:

1) The collected road surface data is divided into 1m×1m grids. Each cross section contains at least 3 grids. The longitudinal section samples at least 10 grids based on a fixed interval, usually 30 grids. The number of grids is set to N.

2) Each 1 square meter grid is divided into nine equal parts of 3×3, and each sub-grid is a scanning, detection and analysis object, with the size set to 30cm×30cm, 20cm×20cm, and 10cm×10cm.

3) Use three-dimensional near-ground integrated texture photogrammetry to remotely control the car to acquire data.

The composition and data acquisition scope are shown in the figure, including five parts: a three-wheeled vehicle, an on-board brain module, a three-dimensional camera sensor, an encoder, and a sensor clamping device.

The three-wheeled vehicle includes its own transmission device, remote control, motor and other equipment, as well as a flexible trailer hook at the front for towing; an internal light source photography channel for projecting lasers and preventing data acquisition from being blocked; the front wheels have the function of moving upward and retracting, and can be folded up in a towing situation;

The on-board brain module includes a control system that controls the car to run according to the trajectory, a photography control module that obtains wheel encoder data and controls the camera sensor acquisition mode, a data local storage module, and a communication module for remote control (remote control);

The 3D camera sensor uses the principle of laser photography triangulation and consists of centrally controlled camera and line laser sub-sensors;

The encoder is used to obtain the rotation frequency of the wheel shaft;

The sensor clamping device is mainly used to fix the three-dimensional camera sensor. It is a beam-type gantry with three fixed heights, corresponding to the three scanning fields of view of 30cm×30cm, 20cm×20cm, and 10cm×10cm respectively.

4) Store two-dimensional brightness data and three-dimensional depth information at the same time.

5) Deformation calibration of depth information

The driving direction of the collected single-point image is L, the scanning direction is T, and the depth information direction is Z.

First, the surface fitting was performed after removing the extremely convex and concave areas of the 10×10 area window.

Fitting optimization plane Z, the original depth set is Z ₀ , and the calibrated depth set is Z ₁ .

Z＝aX+bXY+cY+ ^dX2 +e

Z ₁ = Z ₀ - Z

6) Image segmentation based on the maximum inter-class gap is performed based on the grayscale value of the grayscale image, and the data obtained is shown in the figure below.

The number of pixels A in the connected area within the measurement range (i.e., each test grid) is statistically calculated, and the corresponding kurtosis _PA , skewness _SA , range D _A , information entropy _EA , and self-similarity _SSIA are calculated based on the hierarchical statistical data of the connected areas.

Based on the depth information, the depth threshold of the maximum inter-class gap is solved, and the data obtained is shown in the figure below.

7) Based on the threshold surface, the connected volume statistics set V of the convex and concave is obtained to obtain the distribution statistics of the volume data, and the corresponding kurtosis _PV , skewness _SV , range _DV , information entropy _EV and self-similarity _SSIV are calculated.

8) The calculation formulas for the corresponding indicators of the measuring points are as follows, where J represents the corresponding data matrix of A, V, I _i , _Hi .

D(J)＝maxJ-minJ

Among them, m and n are the numbers of the corresponding subsets based on the equal division of the existing data matrix, B is the number of equal divisions, and the equal division ensures that the actual space x and y corresponding to the size of each submatrix is not less than 50 mm, and the number of B is not less than 4.

c ₁ =(K ₁ *L) ²

c ₂ =(K ₂ *L) ²

Generally, K1=0.01, K2=0.03, L=matrix value interval range, which is the dynamic range of pixel values or depth values, which is 255 for grayscale images.

When it is a grayscale image, α＝0, β＝1, γ＝1;

When it is a depth matrix, α＝1, β＝1, γ＝1;

Local uniformity index SSI _i = SSI(I _i ) _i SSI(H _i ) _i

9) Obtain the feature set of N measuring points belonging to the same road section. The feature set of the i-th point is as follows:

Fi={ _PAi , S _Ai , D _Ai , E _Ai , SSI _Ai , P _Vi , S _Vi , D _Vi , E _Vi , SSI _Vi },

Where i=1~N.

10) Calculate the Euclidean distance d _ij between each pair of N measuring points based on the normalized set of 10 statistical characteristic indicators.

Among them, d _ij is the Euclidean distance between two measuring points, _fik , _fjk is the value of the kth index of the feature set of measuring points i and j.

The test road data set D={d _ij }, i=1~N, j=1~N.

11) Calculate the average distance index ADI, the maximum distance index MDI and the local uniformity of the road SSI.

Where N is the total number of measuring points, usually 30, which can be increased or decreased but not less than 10;

ADI represents the overall uniformity level of the road section, MDI represents the extreme unevenness level, and SSI represents the local unevenness level. The three are used comprehensively to evaluate the uniformity of the road surface. The larger the values of ADI and MDI, the worse the uniformity; the value range of SSI is 0 to 1, and the larger the value, the better the uniformity.

Embodiment 3:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data includes the following steps:

1) The road surface to be analyzed is divided into multiple large grids, and each large grid is divided into multiple sub-grids; the number of specifications and types of these sub-grids is recorded as n; each sub-grid is used as a scanning detection and analysis area; n is a positive integer;

2) Using a 3D near-ground integrated texture photography device to obtain 2D images and 3D depth data;

The three-dimensional near-ground integrated texture photography device includes a three-wheeled vehicle, a vehicle-mounted brain module, a three-dimensional camera sensor, an encoder, and a sensor clamping device;

The sensor clamping device is a beam-type gantry, which is provided with n clamping positions of different heights, corresponding to the scanning fields of n sub-grids of different specifications respectively;

A plurality of three-dimensional camera sensors are fixed on the positions of the sensor fixing device, respectively scan the corresponding scanning detection and analysis areas, and transmit the road surface image data to the vehicle brain module;

The road surface image data includes a two-dimensional image and three-dimensional depth data;

The encoder acquires the rotation frequency of the wheel axle of the three-wheeled trolley and transmits it to the on-board brain module;

The on-board brain module drives the three-wheeled vehicle to travel on the road surface to be analyzed and controls the travel trajectory of the three-wheeled vehicle;

The on-board brain module controls the operation of the three-dimensional camera sensor and the encoder;

The on-board brain module receives road image data and the rotation frequency of the wheel axle of the three-wheeled trolley, and uploads them to the host computer;

3) The host computer performs deformation calibration on the three-dimensional depth data to obtain a calibrated depth set Z ₁ ;

The calibrated three-dimensional depth information set _Z1 is as follows:

Z＝aX+bXY+cY+dX ² +e (5)

Z ₁ = Z ₀ - Z (6)

Where a, b, c, d are depth information coefficients; e is a constant; _Z0 is the three-dimensional depth information set before calibration; Z is the fitting optimization plane; X, Y are the two-dimensional coordinates of the current measurement point;

4) Based on the grayscale value of the two-dimensional image data, the two-dimensional image is segmented based on the maximum inter-class difference to obtain multiple connected regions;

5) Calculate the number of pixels A in the connected area, and based on the hierarchical statistical data of the connected area, calculate the kurtosis _PA , skewness _SA , range D _A , information entropy _EA and self-similarity SSI _A corresponding to the connected area;

6) Determine the depth threshold of the maximum inter-class gap based on the depth information;

7) Based on the depth threshold, count the connected volumes of convexities and concavities, and establish a connected volume statistics set V;

Calculate the kurtosis P _V , skewness S _V , range D _V , information entropy _EV and self-similarity SSI _V corresponding to the connected volume statistics set V;

8) Establish a feature set of N measuring points on the road surface to be analyzed; one measuring point corresponds to one scanning detection and analysis area;

Among them, the feature set of the i-th measuring point is as follows:

Fi={ _PAi , S _Ai , D _Ai , E _Ai , SSI _Ai , P _Vi , S _Vi , D _Vi , E _Vi , SSI _Vi }; (1

Wherein, i=1, 2, ..., N;

9) Calculate the Euclidean distance d _ij between each pair of measuring points;

10) Calculate the average distance index ADI, maximum distance index MDI and local road uniformity SSI used to characterize road surface uniformity, namely:

MDI＝max _i，j≤N d _ij -min _i，j≤N d _ij (3)

Where N is the total number of measuring points; SSI _i is the local uniformity index.

Embodiment 4:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, the main content of which is shown in Example 3, wherein the three-dimensional near-ground integrated texture photography device also includes a car remote control module;

The car remote control module generates the car remote control data and transmits it to the on-board brain module, thereby realizing the remote control of the three-wheeled car.

Embodiment 5:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, the main content of which is shown in Example 3, wherein the vehicle-mounted brain module is mounted on a three-wheeled vehicle, including a vehicle control system, a photography control module, a data local storage module and a communication module;

The vehicle control system drives the three-wheeled vehicle to travel on the road surface to be analyzed and controls the travel trajectory of the three-wheeled vehicle;

The photography control module is used to control the acquisition mode of the three-dimensional camera sensor; the acquisition mode includes a uniform speed acquisition mode and an encoder-based speed coordination control mode;

The data local storage module is used to receive and store road surface image data collected by the three-dimensional camera sensor;

The communication module is used to receive data from the remote control module of the car and transmit it to the car control system.

Embodiment 6:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, the main content of which is shown in Example 3, wherein the three-wheeled vehicle also includes a flexible trailer hook;

The flexible trailer hook is used to connect the three-wheeled trolley to the trailer, so that the trailer can move with the three-wheeled trolley;

When the trailer moves with the three-wheeled vehicle, the front wheels of the three-wheeled vehicle move upward and are retracted.

Embodiment 7:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, the main content of which is shown in Example 3, wherein the three-dimensional camera sensor includes a camera sensor and a line laser sub-sensor;

The line laser sub-sensor is used to monitor the three-dimensional depth information of the ground;

The camera sensor is used to monitor the two-dimensional image of the ground.

Embodiment 8:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, the main content of which is shown in Example 3, wherein the kurtosis P(J), skewness S(J), range D(J), information entropy E(J) and self-similarity SSI(J) are respectively as follows:

D(J)＝max J-min J (9)

E(J)＝-∑ _j∈J P(j)log ₂ j (10)

SSI _i =SSI(I _i ) _i SSI(H _i ) _i (12)

Wherein, J represents the data matrix composed of the number of pixels in the connected area A, the data matrix composed of the connected volume statistics set V, the data matrix composed of the two-dimensional image brightness data I _i , and the data matrix composed of the three-dimensional depth data _Hi ; m and n are the numbers of the corresponding subsets based on the equal division of the existing data matrix; B is the number of equal divisions; α, β, and γ are constants; μ represents the mean, and σ represents the standard deviation; SSI(I _i ) _i represents the local uniformity of the road corresponding to the two-dimensional image brightness data I _i ; SSI(H) _i represents the local uniformity of the road corresponding to the three-dimensional depth data _Hi ;

Among them, the parameters l(m,n), c(m,n), and t(m,n) are as follows:

c ₁ =(K ₁ *L) ² (17)

c ₂ =(K ₂ *L) ² (18)

In the formula, c ₁ , c ₂ , c ₃ are intermediate parameters; K ₁ , K ₂ are constants; L is the length of the matrix numerical interval; _{μ m} , μ _n are means; σ _mn is the standard deviation.

Embodiment 9:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, the main content of which is shown in Example 3, wherein the Euclidean distance d _ij is as follows:

Among them, d _ij is the Euclidean distance between two measuring points, _fik , _fjk is the value of the kth index of the feature set of measuring points i and j.

Embodiment 10:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, the main content of which is shown in Example 3, wherein the average distance index ADI represents the overall uniformity level of the road section, the maximum distance index MDI represents the extreme unevenness level, and the local uniformity situation SSI of the road represents the local unevenness level;

The larger the values of ADI and MDI, the worse the uniformity; the value range of SSI is 0 to 1, and the larger the value, the better the uniformity.

Embodiment 11:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, the main content of which is shown in Example 3, wherein the step of performing image segmentation on the two-dimensional image based on the maximum inter-class difference includes:

4.1) Taking all pixel grayscale data within the measurement range as statistical objects, solve the pixel grayscale distribution probability P _r (K _rt ) corresponding to different grayscale levels i;

4.2) Calculate the inter-class variance of the pixel grayscale distribution corresponding to any two different grayscale levels;

Among them, the inter-class variance between the pixel grayscale cluster corresponding to grayscale r and the pixel grayscale cluster corresponding to grayscale t is As shown below:

Wherein, m _G is the average gray value of the measurement range; m _r (K _rt ) and m _t (K _rt ) are the average gray values of the pixel gray clusters corresponding to gray level r and gray level t; r≠t; r, t＝1, 2, …, L; L is the number of gray levels;

4.3) Solve the inter-class variance Reach the maximum grayscale threshold K _rtmax ;

4.4) Segment the two-dimensional image based on all grayscale thresholds.

Embodiment 12:

A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, the main content of which is shown in Example 3, wherein the step of solving the depth threshold of the maximum inter-class difference based on depth information includes:

6.1) Take all depth data within the measurement range as statistical objects and solve the distribution probability P _g (h _fg ) corresponding to different depth levels g;

6.2) Calculate the inter-class variance of the depth distribution corresponding to any two different depths;

Among them, the inter-class variance of the deep clusters corresponding to the depth level g and the deep clusters corresponding to the depth level f is As shown below:

Wherein, m _h is the average depth of the measurement range; m _f (K _fg ) and m _g (K _fg ) are the average depth values of the depth clusters corresponding to depth level f and depth level g; f ≠ g; f, g = 1, 2, …, H; H is the number of depth levels;

6.3) Solve the between-class variance The maximum depth threshold _hfgmax is reached.

Claims (10)

1. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data, which is characterized by including the following steps: 1) Split the road surface to be analyzed into multiple large grids, and then split each large grid into multiple sub-grids; the number of specification types of these sub-grids is recorded as n; each sub-grid is treated as a scan. Detection and analysis area; n is a positive integer; 2) Use a three-dimensional near-earth integrated texture photography device to obtain two-dimensional images and three-dimensional depth data; The three-dimensional near-ground integrated texture photography device includes a three-wheeled trolley, a vehicle-mounted brain module, a three-dimensional camera sensor, an encoder, and a sensor clamping device; The sensor clamping position fixing device is a beam-type gantry, which is provided with n clamping positions with different heights, respectively corresponding to the scanning fields of n sub-grids of different specifications; Multiple three-dimensional camera sensors are fixed on the card positions of the sensor card position fixing device, respectively scan the corresponding scanning detection and analysis areas, and transmit the road surface image data to the on-board brain module; The road surface image data includes two-dimensional images and three-dimensional depth data; The encoder obtains the rotation frequency of the three-wheeled vehicle axle and transmits it to the on-board brain module; The vehicle-mounted brain module drives the three-wheeled car to drive on the road to be analyzed, and controls the driving trajectory of the three-wheeled car; The vehicle-mounted brain module controls the work of the three-dimensional camera sensor and encoder; The vehicle-mounted brain module receives road surface image data and three-wheeled vehicle axle rotation frequency, and uploads them to the host computer; 3) The host computer performs deformation calibration on the three-dimensional depth data to obtain the calibrated three-dimensional depth information set Z ₁ ; The calibrated three-dimensional depth information set Z ₁ is as follows: Z＝aX+bXY+cY+dX ² +e (5) Z ₁ =Z ₀ -Z (6) In the formula, a, b, c, d are depth information coefficients; e is a constant; Z ₀ is the three-dimensional depth information set before calibration; Z is the fitting optimization plane; X and Y are the two-dimensional coordinates of the current measuring point; 4) Based on the gray value of the two-dimensional image data, perform image segmentation on the two-dimensional image based on the maximum inter-class gap to obtain multiple connected regions; 5) Calculate the number of pixels A in the connected area, and based on the hierarchical statistical data of the connected area, calculate the kurtosis P _A , skewness S _A , range D _A , information entropy EA and self-similarity SSI _A corresponding to the connected area; 6) Solve the depth threshold for the maximum inter-class gap based on depth information; 7) Based on the depth threshold, count the connected volumes of bulges and depressions, and establish a connected volume statistical set V; Calculate the kurtosis P _V , skewness S _V , range D _V , information entropy E _V and self-similarity SSI _V corresponding to the connected volume statistical set V; 8) Establish a feature set of N measuring points on the road surface to be analyzed; one measuring point corresponds to a scanning detection and analysis area; Among them, the feature set of the i-th measuring point is as follows: Fi={ _PAi , S _Ai , D _Ai , E _Ai , SSI _Ai , P _Vi , S _Vi , D _Vi , E _Vi , SSI _Vi }; (1) In the formula, i=1, 2,...,N; 9) Calculate the Euclidean distance d _ij between two measuring points; 10) Calculate the average distance index ADI, the maximum distance index MDI and the local uniformity of the road SSI used to characterize the uniformity of the road surface, namely: MDI＝max _i，j≤N d _ij -min _i，j≤N d _ij (3) In the formula, N is the total number of measuring points; SSI _i is the local uniformity index. 2. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data according to claim 1, characterized in that the three-dimensional near-ground integrated texture photography device also includes a car remote control module; The car remote control module generates the car remote control data and transmits it to the vehicle brain module, thereby realizing the remote control of the three-wheeled car. 3. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data according to claim 2, characterized in that the vehicle-mounted brain module is mounted on a three-wheeled trolley and includes a trolley control system, a photography control module, and a data local computer. storage module and communication module; The trolley control system drives the three-wheeled trolley to drive on the road surface to be analyzed, and controls the driving trajectory of the three-wheeled trolley; The photography control module is used to control the collection mode of the three-dimensional camera sensor; the collection mode includes a uniform speed collection mode and an encoder-based speed coordination control mode; The data local storage module is used to receive and store road surface image data collected by the three-dimensional camera sensor; The communication module is used to receive data from the car's remote control module and transmit it to the car control system. 4. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data according to claim 1, characterized in that the three-wheeled trolley further includes a flexible trailer hook; The flexible trailer hitch is used to connect the three-wheeled trolley to the trailer so that the trailer moves with the three-wheeled trolley; When the trailer moves with the three-wheeled trolley, the front wheels of the three-wheeled trolley move upward and stow away. 5. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data according to claim 1, characterized in that the three-dimensional camera sensor includes a camera sensor and a line laser sub-sensor; The line laser sub-sensor is used to monitor the three-dimensional depth information of the ground; The camera sensor is used to monitor two-dimensional images of the ground. 6. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data according to claim 1, characterized in that kurtosis P(J), skewness S(J), range D(J), information entropy E(J) and self-similarity SSI(J) are as follows: D(J)＝max J-min J (9) E(J)＝-∑ _j'∈J P(j')log ₂ j' (10) SSI _i =SSI(I _i ) _i SSI(H _i ) _i (12) In the formula, J respectively represents the data matrix composed of the number of pixels in the connected area A, the data matrix composed of the connected volume statistics set V, the data matrix composed of the two-dimensional image brightness data I _i , and the data matrix composed of the three-dimensional depth data H _i ; m, n is the number of the corresponding subset based on the existing data matrix equalization; B is the number of equal parts; α, β, and γ are constants; μ represents the mean, σ represents the standard deviation; SSI (I _i ) _i represents the brightness of the two-dimensional image The local uniformity of the road corresponding to the data I _i ; SSI(H) _i represents the local uniformity of the road corresponding to the three-dimensional depth data _Hi ; Among them, the parameters l(m, n), c(m, n), t(m, n) are as follows: c ₁ = (K ₁ *L) ² (17) c ₂ =(K ₂ *L) ² (18) In the formula, c ₁ , c ₂ and c ₃ are intermediate parameters; K ₁ and K ₂ are constants; L is the length of the matrix numerical interval; μ _m and μ _n are the mean; σ _mn is the standard deviation. 7. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data according to claim 1, characterized in that the Euclidean distance d _ij is as follows: Among them, d _ij is the Euclidean distance between two measuring points, f _ik , and f _jk is the value of the kth index of the feature set of measuring points i and j. 8. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data according to claim 1, characterized in that the average distance index ADI represents the overall uniformity level of the road section, and the maximum distance index MDI represents the extreme uneven level, Road local uniformity condition SSI represents the local unevenness level; The larger the values of ADI and MDI, the worse the uniformity; the value range of SSI is 0 ~ 1, the larger the value, the better the uniformity. 9. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data according to claim 1, characterized in that the step of performing image segmentation on the two-dimensional image based on the maximum inter-class gap includes: 4.1) Use all pixel grayscale data within the measurement range as statistical objects to solve for the pixel grayscale distribution probability P _r (K _rt ) corresponding to different grayscale levels; 4.2) Calculate the inter-class variance of the pixel gray distribution corresponding to any two different gray levels; Among them, the inter-class variance between the pixel gray cluster corresponding to gray level r and the pixel gray cluster corresponding to gray level t As follows: In the formula, m _G is the average gray value of the measurement range; m _r (K _rt ) and m _t (K _rt ) are the average gray value of the pixel gray cluster corresponding to gray level r and gray level t; r≠t; r, t=1, 2,..., L; L is the number of gray levels; 4.3) Solve the inter-class variance Reach the maximum gray threshold K _rtmax ; 4.4) Segment the two-dimensional image according to all grayscale thresholds. 10. A road surface uniformity analysis method based on two-dimensional and three-dimensional fusion data according to claim 1, characterized in that the step of solving the depth threshold of the maximum inter-class gap based on depth information includes: 6.1) Use all depth data within the measurement range as statistical objects and solve for the distribution probability P _g (h _fg ) corresponding to different depth levels g; 6.2) Calculate the inter-class variance of the depth distribution corresponding to any two different depths; Among them, the inter-class variance of the depth clustering corresponding to depth level g and the depth clustering corresponding to depth level f As follows: In the formula, m _h is the average depth of the measurement range; m _f (h _fg ) and m _g (h _fg ) are the average depth values of the depth clusters corresponding to depth level f and depth level g; f ≠ g; f, g =1,2,…,H; H is the number of depth levels; 6.3) Solve the inter-class variance The maximum depth threshold h _fgmax is reached.