A method for fine measurement of road rut based on structured light

Abstract

The application provides a kind of road rut fine measurement method based on structured light, first to the road point cloud data collected is denoised, slope and two-dimensional gridding processing, then the various features of point cloud data in gridding are mapped to gray scale image, finally based on image recognition and multi-feature fusion method realizes the identification and fine measurement of rut. Since the point cloud data collection is carried out by using the vehicle-mounted structured light collection equipment, the measurement of the degree of road rut under the dynamic traffic speed can be supported, and more rapid and fine rut detection and fine evaluation can be realized. At the same time, more fine road damage conditions are provided for the road maintenance department, and basis is provided for maintenance fund allocation and measure selection. Also, since the point cloud data is processed by two-dimensional gridding, the calculation amount of point cloud data processing and the dependence on artificial or high-precision instruments are greatly reduced, which has important significance for large-scale road performance high-frequency digital detection.

Description

Technical Field

[0001] This invention belongs to the field of road surface quality detection and automatic information acquisition technology, and specifically relates to a method for fine measurement of road ruts based on structured light. Background Technology

[0002] Ruts are permanent deformations on a road surface caused by repeated driving of vehicle tires along the roadway track. In recent years, with the rapid increase in traffic volume and the intensification of heavy and overloaded conditions, ruts have become one of the main factors affecting road performance. They not only directly affect driving safety and comfort, but also seriously endanger the integrity and stability of the road structure, shorten the road's service life, and increase maintenance costs.

[0003] Rutting detection technology can be mainly divided into two categories: manual detection and automated detection. Traditional manual contact-based rutting measurement methods, such as AASHTO rutting gauges, surface leveling meters, and hand-pushed profile measuring instruments, suffer from low measurement efficiency, and their accuracy is greatly affected by human factors, exhibiting random errors and making them unsuitable for large-scale rutting detection. With the rapid development of laser technology, 3D laser technology has gradually matured and begun to be applied in the field of road surface inspection. Current automated detection equipment is all laser equipment, including point laser rutting meters and line laser rutting meters. Point lasers use the principle of triangulation, with an unchangeable receiving direction, processing one point at a time, resulting in slow measurement speed. Line lasers project planar slit light through a projection source, projecting one structured light stripe at a time to obtain the depth of a cross section. However, vehicle-mounted line lasers are easily affected by vehicle bumps and vibrations, significantly reducing accuracy. Therefore, compared to the disadvantages of existing rutting measurement equipment, such as low efficiency and large errors, advanced surface structured light 3D snapshot technology provides the possibility for large-scale rutting detection.

[0004] Structured light technology obtains the three-dimensional coordinate information of the road surface at the same instant by taking high-frame-rate snapshots of the road surface. The analysis can produce a color depth image of the road surface, which realizes sub-millimeter high-precision measurement of ruts. It is unaffected by vehicle vibration and bumps, and meets the accuracy requirements to the greatest extent possible while ensuring large-scale vehicle measurement. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a method for refined measurement of road ruts based on structured light. The technical solution adopted in this invention is as follows:

[0006] This invention provides a method for fine measurement of road ruts based on structured light, characterized by the following steps: Step S1, installing a structured light acquisition device on a testing vehicle to photograph the road surface to be measured from a top-down angle and acquiring road point cloud data; Step S2, preprocessing the point cloud data; Step S3, performing two-dimensional rasterization processing on the preprocessed point cloud data; Step S4, extracting various features of the point cloud data within each raster and mapping the normalized features to grayscale images; Step S5, using image recognition methods to identify abnormal regions in the grayscale images as rut ​​regions; Step S6, extracting point cloud data within the rut regions and calculating the maximum depth, width, length, area, and volume of the rut regions as evaluation indicators for the ruts.

[0007] In the method for fine measurement of road ruts based on structured light provided by the present invention, there may also be the following technical feature, wherein the road point cloud data contains at least depth information, that is, the distance between the collected point and the collection device.

[0008] In the method for fine measurement of road ruts based on structured light provided by the present invention, the following technical features may also be included: the preprocessing includes noise reduction and slope reduction.

[0009] In the structured light-based method for fine measurement of road ruts provided by this invention, the following technical feature is also included: denoising includes the following steps: Step S2-1a, calculating the average distance d from each point in the point cloud data to its k nearest points, and constructing a set D{d1,d2,…,d…} n}; Step S2-2a, normal fitting set D{d1,d2,…,d n} and calculate its mean μ and variance σ; step S2-2a, take d i Points outside the range [μ-3σ, μ+3σ] are treated as noise points and discarded.

[0010] In the structured light-based method for fine measurement of road ruts provided by this invention, the following technical feature is also included: the slope removal process is as follows: Step S2-1b, take the three-dimensional coordinate set O{(x1,y1,z1),(x2,y2,z2),…,(x n ,y n ,z n Step S2-2b: Fit the set O using the least squares method to obtain the corresponding three-dimensional plane parameters k. x ,k y Step S2-3b: Calculate the three-dimensional coordinate set O′{(x1,y1,z1-k) of the point cloud data after slope removal. x x1-k y y1-b),(x2,y2,z2-kx x2-k y y2-b),…,(x n ,y n ,z n -k x x n -k y y n -b)}.

[0011] In the method for fine measurement of road ruts based on structured light provided by the present invention, there may also be the following technical feature, wherein, in step S4, the feature types of the point cloud data include at least the height value of the point cloud after slope removal, the reflection intensity value of the point cloud, the surface normal, the principal curvature, the point feature histogram descriptor, the fast point feature histogram descriptor, and the 3D shape content descriptor.

[0012] In the method for fine measurement of road ruts based on structured light provided by the present invention, the following technical features are also provided: In step S4, the mapping process is as follows: Step S4-1, calculate the maximum and minimum values ​​of the feature values; Step S4-2, map the gray value of the feature maximum value position to 255, map the gray value of the feature minimum value position to 0, and map the remaining positions to integers [0, 255] according to a linear ratio, and round to the nearest integer when mapping to decimals.

[0013] In the structured light-based method for fine measurement of road ruts provided by the present invention, there may also be the following technical feature: the image recognition method adopts at least one of the convolutional neural network recognition method or the support vector machine recognition method.

[0014] In the structured light-based method for fine measurement of road ruts provided by this invention, the following technical feature is also included: the calculation process for the maximum depth of the rut region is as follows: Step S6-1, the average distance d from each point in the point cloud data of the rut region to its k nearest points along the z-axis is calculated. z All points form set D z {d z1 ,d z2 ,…,d zn Step S6-2, take set D z The point cloud corresponding to the maximum value in the data is taken as the highest point, and its z-axis coordinate is taken as the ground height z. o Step S6-3: Calculate the point cloud corresponding to the minimum z-axis value of the normalized rut area point cloud data as the lowest point, and take its z-axis coordinate as the height z2 at the maximum rut depth; Step S6-4: Take z... o -z2 represents the maximum depth of the rutted area.

[0015] In the structured light-based method for fine measurement of road ruts provided by this invention, the following technical feature is also included: the width of the rut region is determined by fitting the minimum bounding rectangle of the rut region in a grayscale image using the least squares method. The length of the shorter side of this minimum bounding rectangle is taken as the width w of the rut region, the length of the rut region is the length l of the first side of the minimum bounding rectangle, the area of ​​the rut region is w × l, and the volume of the rut region is... In the formula, This is the average depth. The average value of the z-axis of the point cloud within the rut region. Take the ground height z o and The difference is achievable.

[0016] Invention Function and Effect

[0017] According to the present invention, a method for refined measurement of road rutting based on structured light firstly denoises, de-slopes, and performs two-dimensional rasterization on the collected road point cloud data. Then, various features of the rasterized point cloud data are mapped into grayscale images. Finally, rutting identification and refined measurement are achieved based on image recognition and multi-feature fusion methods. Because this method uses an onboard structured light acquisition device to collect road point cloud data, it can support the measurement of road rutting severity under dynamic traffic flow speeds. On the one hand, it provides road maintenance departments with more detailed information on road damage, providing a basis for maintenance fund allocation and measure selection. On the other hand, it enables faster and more precise rutting detection and evaluation, thus ensuring travel safety. Furthermore, the use of a two-dimensional rasterization method for the point cloud data significantly reduces the computational load of point cloud data processing.

[0018] The structured light-based method for fine measurement of road rutting in this invention can achieve rapid and fine measurement of road rutting at traffic speeds, improve the detection speed of road rutting, reduce reliance on manual labor or high-precision instruments, and is of great significance for large-scale high-frequency digital detection of road performance. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method for fine measurement of road ruts based on structured light in an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of point cloud data after slope removal in an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of point cloud data rasterization and grayscale mapping in an embodiment of the present invention. Detailed Implementation

[0022] This invention relates to a method for fine measurement of road ruts based on structured light. The method mainly utilizes vehicle-mounted structured light acquisition equipment to collect road point cloud data, and achieves rut ​​identification and fine measurement through two-dimensional rasterization processing, image recognition, and multi-feature fusion.

[0023] To make the technical means, creative features, objectives and effects of this invention easy to understand, the following describes the structured light-based method for fine measurement of road ruts in conjunction with embodiments and accompanying drawings.

[0024] <Example>

[0025] Figure 1 This is a flowchart of a method for fine measurement of road ruts based on structured light in an embodiment of the present invention.

[0026] like Figure 1 As shown, the structured light-based method for fine measurement of road ruts includes the following steps:

[0027] Step S1: Install the structured light acquisition device on the inspection vehicle and take pictures of the road surface to be tested from a top-down angle to collect road point cloud data.

[0028] To ensure sampling accuracy and density, the structured light acquisition device needs to be adjusted in installation position. It can be installed and fixed in front of or behind the inspection vehicle, shooting vertically towards the ground to ensure a longitudinal sampling accuracy greater than 5mm and a sampling density greater than 90,000 / square meter. In this embodiment, the structured light acquisition device is installed on the rear roof of the inspection vehicle, with the lens shooting vertically downwards to acquire point cloud data including elevation information and reflection intensity information. The elevation (depth) information refers to the distance of the sampled point from the acquisition device.

[0029] Step S2: Preprocess the point cloud data.

[0030] In this embodiment, the point cloud data is first denoised, and the process is as follows:

[0031] Step S2-1a: Calculate the average distance d from each point in the point cloud data to its four nearest points, and construct a set D{d1,d2,…,d…} n};

[0032] Step S2-2a, normal fitting set D{d1,d2,…,d n} and calculate its mean μ = 15 and variance σ = 1.24;

[0033] Step S2-2a, take d i Points outside the range [μ-3σ,μ+3σ], i.e., [11.28,18.72], are considered noise points and are removed.

[0034] Then, the point cloud data is deslope processed:

[0035] Step S2-1b: Obtain the three-dimensional coordinate set O{(x1,y1,z1),(x2,y2,z2),…,(x n ,y n ,z n )};

[0036] Step S2-2b: Fit the set O using the least squares method to obtain the corresponding three-dimensional plane parameters k. x =0.036,k y =-0.008, b=57.031;

[0037] Step S2-3b: Calculate the three-dimensional coordinate set O′{(x1,y1,z1-k) of the point cloud data after slope removal. x x1-k y y1-b),(x2,y2,z2-k x x2-k y y2-b),…,(x n ,y n ,z n -k x x n -k y y n -b)}, removing point cloud data before and after slope, for example Figure 2 As shown.

[0038] Step S3: Perform two-dimensional rasterization processing on the x,y plane of the preprocessed point cloud data.

[0039] Figure 3 This is a schematic diagram of point cloud data rasterization and grayscale mapping in an embodiment of the present invention.

[0040] In this embodiment, the two-dimensional rasterization process is as follows: the preprocessed point cloud data is projected onto the x,y plane, and then the plane is rasterized using a 5×5 grid. The processing result is as follows. Figure 3 As shown.

[0041] Step S4: Extract various features of the point cloud data in each grid, and map the normalized features to grayscale images respectively.

[0042] The features of point cloud data include, but are not limited to, the following: the height value of the point cloud after slope removal, the reflection intensity value of the point cloud, the surface normal, the principal curvature, the point feature histogram descriptor (PFH), the fast point feature histogram descriptor (FPFH), and the 3D shape content descriptor, etc.

[0043] In this embodiment, the height value of the point cloud is selected as the point cloud data feature. The average height of the point cloud data within each grid is extracted. The grayscale value of the pixel corresponding to the largest grid is set to 255, the grayscale value of the pixel corresponding to the smallest grid is set to 0, and the grayscale values ​​of the pixels corresponding to the remaining grids are interpolated proportionally to obtain the mapped grayscale image. Figure 3 As shown.

[0044] Step S5: Use image recognition methods such as convolutional neural networks or support vector machines to identify abnormal regions in grayscale images as rut ​​regions.

[0045] In this embodiment, the cyclic local binarization (LBP) algorithm is first used to extract the local binary pattern (LBP) features, i.e., texture features, of the mapped grayscale image. The formula for calculating the binarization feature of each pixel is as follows:

[0046]

[0047] In the formula, (x c ,y c ) is the center pixel, i c It is the grayscale value, i p It represents the grayscale value of adjacent pixels, and s is a sign function:

[0048] Then, the binarized features are input into a support vector machine (SVM) to classify whether the region belongs to a rut region.

[0049] Step S6: Extract the basic parameters of the point cloud data within the rut area, and use them to calculate the maximum depth, width, length, area, and volume of the rut area as evaluation indicators for the rut.

[0050] In this embodiment, the basic parameters include ground height, height at the maximum depth of the ruts, and the average value of the z-axis of the point cloud within the region. The length and width of the minimum bounding rectangle. The extraction process for these basic parameters is as follows:

[0051] The calculation process for ground height and height at maximum depth is as follows:

[0052] First, calculate the average distance d from each point in the point cloud data of the rut area to its k nearest points along the z-axis. z All points form set D z {d z1 ,d z2 ,…,d zn}

[0053] Then, take set D. z The point cloud corresponding to the maximum value in the data is taken as the highest point, and its z-axis coordinate is taken as the ground height z.o .

[0054] Finally, the point cloud corresponding to the minimum z-axis value of the normalized rut area point cloud data is calculated as the lowest point, and its z-axis coordinate is taken as the height z2 at the maximum depth of the rut.

[0055] The minimum bounding rectangle is obtained by fitting the rut region in the grayscale image using the least squares method. The length of the long side and the length of the short side of the minimum bounding rectangle are used as the length l and width w of the rut region, respectively.

[0056] The results of the above basic parameters calculated in this embodiment are shown in Table 1 below:

[0057]

[0058] Table 1

[0059] In this embodiment, z is taken as o -z2 represents the maximum depth of the rut region. The area of ​​the rut region is w×l, and the volume of the rut region is... in, The average depth is calculated by taking the ground height z. o The average value of the z-axis of the point cloud within the rut region The difference is achievable.

[0060] The detailed evaluation of the detected ruts was calculated based on the basic parameters calculated in Table 1 above, and the results are shown in Table 2 below:

[0061]

[0062] Table 2

[0063] Functions and effects of the embodiments

[0064] According to the structured light-based method for fine measurement of road ruts provided in this embodiment, the collected road point cloud data is first denoised, de-slopeed, and rasterized in two dimensions. Then, various features of the rasterized point cloud data are mapped into grayscale images. Finally, the identification and fine measurement of ruts are achieved based on image recognition and multi-feature fusion methods.

[0065] In this embodiment, because a vehicle-mounted structured light acquisition device is used to collect road surface point cloud data, it can support the measurement of road rutting degree under dynamic traffic flow speeds. On the one hand, this provides road maintenance departments with more detailed information on road damage, providing a basis for the allocation of maintenance funds and the selection of measures. On the other hand, it enables faster and more precise rutting detection and evaluation, thus ensuring travel safety. Furthermore, the use of a two-dimensional rasterization method for the point cloud data significantly reduces the computational load of point cloud data processing.

[0066] In summary, the structured light-based method for fine measurement of road rutting in this embodiment can achieve rapid and fine measurement of road rutting at traffic speeds, improve the detection speed of road rutting, reduce reliance on manual labor or high-precision instruments, and is of great significance for large-scale high-frequency digital detection of road performance.

[0067] The above embodiments are only used to illustrate specific implementations of the present invention, and the present invention is not limited to the scope of the description of the above embodiments.

Claims

1. A method for fine measurement of road ruts based on structured light, characterized in that, include: Step S1: Install the structured light acquisition device on the inspection vehicle and take pictures of the road surface to be tested from a top-down angle to collect road point cloud data; Step S2, preprocess the point cloud data, the preprocessing including noise reduction and slope reduction; Step S3: Perform two-dimensional rasterization processing on the x,y plane of the preprocessed point cloud data; Step S4: Extract various features of the point cloud data in each grid, and map the normalized features to grayscale images respectively. The feature types of the point cloud data include at least the height value of the point cloud after slope removal, the reflection intensity value of the point cloud, the surface normal, the principal curvature, the point feature histogram descriptor, the fast point feature histogram descriptor, and the 3D shape content descriptor. Step S5: Use an image recognition method to identify abnormal regions in the grayscale image as rut ​​regions. The image recognition method shall employ at least one of the following: a convolutional neural network recognition method or a support vector machine recognition method. Step S6: Extract point cloud data from the rut region to calculate the maximum depth, width, length, area, and volume of the rut region as evaluation indicators for the ruts. The calculation process for the maximum depth of the rut region is as follows: Step S6-1: Calculate the average distance d from each point in the point cloud data of the rutted area to its k nearest points along the z-axis. z All points form set D z {d z1 ,d z2 ,…,d zn }; Step S6-2, take set D z The point cloud corresponding to the maximum value in the data is taken as the highest point, and its z-axis coordinate is taken as the ground height z. o ; Step S6-3: Calculate the point cloud corresponding to the minimum z-axis value of the normalized rut area point cloud data as the lowest point, and take its z-axis coordinate as the height z2 at the maximum depth of the rut. Step S6-4, take z o -z2 represents the maximum depth of the rutted area. The width of the rut region is determined by fitting the minimum bounding rectangle of the rut region in the grayscale image using the least squares method, and the length of the shorter side of this minimum bounding rectangle is taken as the width w of the rut region. The length of the rutted region is the length l of the field side of the minimum bounding rectangle. The area of ​​the rutted region is w×l. The volume of the rutted area is in, This is the average depth. The average value of the z-axis of the point cloud within the rut region. Take the ground height z o and The difference is achievable.

2. The method for fine measurement of road ruts based on structured light according to claim 1, characterized in that: in, The road surface point cloud data includes at least depth information, that is, the distance between the collected point and the collection device.

3. The method for fine measurement of road ruts based on structured light according to claim 1, characterized in that: in, The noise reduction includes the following steps: Step S2-1a: Calculate the average distance d from each point in the point cloud data to its k nearest points, and construct a set D{d1,d2,…,d...} n }; Step S2-2a, normal fitting set D{d1,d2,…,d n } and calculate its mean μ and variance σ; Step S2-3a, take d i Points outside the range [μ-3σ, μ+3σ] are treated as noise points and discarded.

4. The method for fine measurement of road ruts based on structured light according to claim 3, characterized in that: in, The process of reducing the slope is as follows: Step S2-1b: Obtain the three-dimensional coordinate set O{(x1,y1,z1),(x2,y2,z2),…,(x n ,y n ,z n )}; Step S2-2b: Fit the set O using the least squares method to obtain the corresponding three-dimensional plane parameters k. x ,k y ,b; Step S2-3b: Calculate the three-dimensional coordinate set O′{(x1,y1,z1-k) of the point cloud data after slope removal. x x1-k y y1-b),(x2,y2,z2-k x x2-k y y2-b),…,(x n ,y n ,z n -k x x n -k y y n -b)}.

5. The method for fine measurement of road ruts based on structured light according to claim 1, characterized in that: in, In step S4, the mapping process is as follows: Step S4-1: Calculate the maximum and minimum values ​​of the eigenvalues; Step S4-2: Map the gray value at the maximum feature value position to 255, map the gray value at the minimum feature value position to 0, and map the remaining positions to integers in the range [0, 255] according to a linear ratio. When mapping to decimals, round to the nearest integer.

Images