Intelligent detection method for steel slag mixing amount of steel slag asphalt based on laser point cloud

Abstract

The invention discloses a steel slag asphalt steel slag mixing amount intelligent detection method based on a laser point cloud, and the method comprises the steps: steel slag asphalt pavement point cloud collection, abnormal data cleaning, a vibration noise suppression program based on empirical mode decomposition, a texture parameter extraction program, building of a steel slag mixing amount data set, and construction and training of a detection model. And verifying the detection model. According to the invention, laser point cloud and road engineering are combined together, and automatic and intelligent analysis and detection of the steel slag mixing amount of the steel slag asphalt pavement based on laser point cloud data are realized.

Description

Technical Field

[0001] The invention belongs to the field of detection technology, and in particular relates to an intelligent detection method for steel slag asphalt steel slag content based on laser point cloud. Background Art

[0002] my country is a major steel industry country, and its steel production ranks first in the world. Using steel slag to replace or partially replace the stone in asphalt pavement as aggregate to form steel slag asphalt pavement is an effective way to reuse steel slag. Steel slag can replace or partially replace stone, and the replacement ratio is called the steel slag content. If all the stone is replaced by steel slag, the steel slag content is 100%. The steel slag content is an important technical indicator. As shown in the table, when the steel slag content changes, the high temperature performance, low temperature performance, water stability performance, fatigue performance and other indicators of the mixture are significantly changed.

[0003] Table 1 Performance test results of AC-13 steel slag asphalt mixture with different dosages

[0004]

[0005] At present, steel slag asphalt pavement is still in its infancy. Various places have successively carried out experimental steel slag asphalt pavement paving projects, but there is still a distance to form a complete set of design-testing-maintenance technical systems. Steel slag content detection is an important part of the technical system, and this technology is still blank.

[0006] As an advanced measurement method, the three-dimensional laser rangefinder has the advantages of being fast, real-time, high-density, high-precision, dynamic and active, and can obtain fine undulation data on the surface of an object. According to the differences in texture morphology of steel slag asphalt pavements with different steel slag contents, it is a feasible idea to obtain the pavement point cloud based on the three-dimensional laser rangefinder, extract the two-dimensional and three-dimensional parameters of the pavement, and further judge the steel slag content of the surface layer. The three-dimensional laser rangefinder is used in large-scale detection of steel slag asphalt pavements, but there are still problems such as strong data noise and unclear texture parameter extraction methods, which bring challenges to this detection technology. Combined with the three-dimensional laser ranging technology, the present invention addresses this gap and proposes an intelligent detection method for steel slag content based on pavement laser point cloud and deep learning technology. Summary of the invention

[0007] In order to achieve the above-mentioned purpose, the present invention provides an intelligent detection method for steel slag asphalt steel slag content based on laser point cloud, which realizes the automated intelligent detection of steel slag content in steel slag asphalt pavement, improves the technical system in the field of steel slag asphalt pavement, and improves the technical level of the steel slag asphalt pavement detection end, which is beneficial to the quality control of steel slag asphalt pavement.

[0008] According to an invention of the present invention, a method for intelligent detection of steel slag asphalt steel slag content based on laser point cloud is provided, which specifically comprises the following steps:

[0009] (1) Point cloud collection of steel slag asphalt pavement: Integrate the 3D laser scanner, travel encoder, and acceleration sensor on the vehicle.

[0010] A bracket is extended backward from the top of the vehicle to set up a 3D laser scanner. The 3D laser scanner is perpendicular to the road surface. The distance between the 3D laser scanner and the road surface is determined according to the scanning range of the 3D laser scanner to ensure that the scanning width exceeds 2m. One scanning transmits a horizontal survey line to obtain the elevations of M points on the horizontal survey line.

[0011] Among them, the traveling encoder is installed on the left rear wheel to measure the vehicle's travel distance and periodically trigger the 3D laser scanner; every time the vehicle travels W meters, the traveling encoder triggers the 3D laser scanner N times to obtain N horizontal measurement lines, and saves the elevation scanning results of M×N points on the road surface in the form of a point cloud data file in dat format;

[0012] The acceleration sensor is attached to the three-dimensional laser scanner and records the vertical motion parameters of the three-dimensional laser scanner at the moment when the encoder triggers the three-dimensional laser scanner.

[0013] (2) Abnormal data cleaning method: Taking the dat file as input, the program decodes it into a matrix of size M×N and calculates the mean μ and variance σ of the elevation in a certain matrix. 2 , the elevation points outside the confidence interval [μ-2σ,μ+2σ] are considered as outliers, and the elevation of the outlier point is assigned the average elevation of the points around it, that is:

[0014]

[0015] In the formula, z′ m，,n After data cleaning, is the original data.

[0016] (3) Suppressing vibration noise based on empirical mode decomposition: The present invention develops a set of vibration noise suppression procedures for steel slag asphalt pavement point cloud based on empirical mode decomposition (EMD), including longitudinal measurement line extraction, vibration component decomposition, vibration component sorting, equidistant plane calculation, and transverse measurement line regression.

[0017] Among them, the method of extracting longitudinal lines is to use the cleaned file as input and decode it into a matrix of size N×M, combine the points with the same m coordinates into a one-dimensional array, and obtain a total of M longitudinal lines, which are numbered 1, 2, 3…j,…, M, and the shape of each longitudinal line is N×1;

[0018] Among them, the method of vibration component decomposition is to take out a longitudinal survey line as the original signal, make the upper and lower extreme points of the original signal to make the upper and lower envelopes, find the mean curve of the upper and lower envelopes, and use the original signal and the mean curve to make a difference to get the intermediate signal. If the number of extreme points and zero-crossing points of the intermediate signal in the entire data segment is equal or the difference is no more than one, and the average value of any upper envelope formed by the local maximum point and the lower envelope formed by the local minimum point is zero, then the vibration component decomposition is completed. If the intermediate signal does not meet the above conditions, it is recorded as vibration component No. 1, and the interpolation of the original signal and vibration component No. 1 is used as the new original signal. Repeat this step until the vibration component decomposition is completed, and the vibration components are numbered 1, 2, ..., i, ... I respectively. Repeat the above steps for N longitudinal survey lines to obtain M groups of I×M vibration components Where j represents the number of the longitudinal measurement line, and i represents the number of the vibration component obtained from the jth longitudinal measurement line;

[0019] The method of sorting the vibration components is to extract the motion parameter curve of the N×M matrix during the acquisition period and perform Fourier transform to obtain the measured vibration spectrum S 0 , the vibration component obtained in the previous step Perform Fourier transform to obtain a series of vibration spectra The meanings of i and j are the same as above, respectively calculating S 0 and each Cosine similarity of Extract the vibration component number corresponding to the maximum value of R in each measuring line and find the mode k of the component number. According to the corresponding Sort by, take the top 20

[0020] The method of calculating the isometric plane is to set the distance between the plane with the expression Ax+By+Cz+D=0 and all points of the N×M matrix to be the minimum, and randomly give a set of initial values ​​A=a 0 , B=b 0 , C=C 0 , D = d 0 , calculate the distance and sum according to the following formula 0 :

[0021]

[0022] Use the conjugate gradient method to update the initial value until sum 0 Deviation from 0 converges to 10 -6 Within, we get the isometric plane ax+by+cz+d=0;

[0023] The method of horizontal line regression is to extract the top 20 horizontal lines for each of the nth horizontal lines. The data of the nth point on the graph are used to calculate the equidistant straight line of these 20 points, which is recorded as the nth equidistant straight line. The method is the same as the calculation method of the equidistant plane. It is assumed that the relative positions of the nth equidistant straight line and all points on the nth transverse line remain unchanged before and after the vibration occurs. The deviation between the nth transverse line and the equidistant plane is the noise caused by the vertical vibration. The nth equidistant straight line and all points on the nth transverse line are synchronously regressed to the equidistant plane, that is:

[0024]

[0025] Where: is the elevation value after noise suppression, To obtain the solution after substituting the point cloud with coordinates (m,n) into the equidistant plane equation, This is the solution result after substituting the point cloud with coordinates (m,n) into the equation of the nth equidistant straight line.

[0026] (4) Texture parameter extraction procedure: This study proposes a method for extracting texture parameters of steel slag asphalt pavement based on local equidistant planes. For a large N×M point cloud matrix, it is divided into p small blocks (A 1 ,A 2 ,…,A p ), calculate the isometric plane of each small block separately.

[0027] Among them, taking the local equidistant plane as the standard, calculate a local area Various texture parameters, including mean cross-sectional depth MPD, arithmetic mean roughness R a , RMS height R q , skewness R sk , average structural depth MTD, etc., a certain parameter of the entire matrix is ​​the average value of the parameter in each local area.

[0028] (5) Establishing a steel slag content dataset: Use the testing equipment in (1) to conduct testing on steel slag asphalt pavement or steel slag asphalt concrete samples with known steel slag content, and link the texture parameters with the steel slag content to construct a dataset.

[0029] The texture parameter extraction method and types are the same as those described in (3).

[0030] Among them, if the steel slag content of the tested road surface ω∈[0%,20%], it is considered that the steel slag content at this location belongs to the "very little" range, and the label of the texture parameter obtained at this location is assigned to 0; if the steel slag content of the tested road surface ω∈(20%,40%], it is considered that the steel slag content at this location belongs to the "little" range, and the label of the texture parameter obtained at this location is assigned to 1; if the steel slag content of the tested road surface ω∈(40%,60%], it is considered that the steel slag content at this location belongs to the "medium" range, and the label of the texture parameter obtained at this location is assigned to 2; if the steel slag content of the tested road surface ω∈(60%,80%], it is considered that the steel slag content at this location belongs to the "much" range, and the label of the texture parameter obtained at this location is assigned to 3; if the steel slag content of the tested road surface ω∈(80%,100%], it is considered that the steel slag content at this location belongs to the "extremely much" range, and the label of the texture parameter obtained at this location is assigned to 4.

[0031] Among them, in order to enrich the collective data volume, two methods were used; first, the entire tested road surface was divided into small blocks, and each small block formed a piece of data; second, steel slag asphalt concrete specimens with specified steel slag content were prepared.

[0032] Among them, the data set is randomly divided into a training data set and a validation data set, and the ratio of the two is n 1 :n 2 .

[0033] (6) Detection model construction and training: Load the training data set, convert the data set to "one-hot" label format, build a convolutional neural network based on tensorflow-gpu, set the training parameters: BATCH_SIZE = b, EPOCHS = p, LEARNING_RATE = α, BUFFER_SIZE = t, model_path = str1, and set the loss function to categorical cross entropy (CategoricalCrossentropy)

[0034]

[0035] After that, compile the model and start training. The output training results will be automatically saved in the model_path path.

[0036] (7) Detection model verification: Load the verification data set and the trained network. If the accuracy of the training data set is

[0037]

[0038] If it is over 90%, the model can be applied. If it is less than 70%, it is necessary to increase the richness of the data set, repeat step (5) and appropriately adjust the parameters in (6) until the verification accuracy reaches over 90%.

[0039] According to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the steps of the intelligent detection method of steel slag asphalt steel slag content based on laser point cloud of the present invention are implemented.

[0040] According to another aspect of the present invention, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for intelligently detecting steel slag asphalt content in steel slag based on laser point cloud of the present invention are implemented.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The present invention uses a three-dimensional laser scanner to scan the road lane, which can quickly and with high resolution obtain three-dimensional scanning data of the road surface, providing reliable data support for establishing steel slag content detection technology;

[0043] 2. Aiming at the abnormal data and vertical elevation errors that occur during the data collection process, the present invention has developed a set of vibration noise suppression procedures for steel slag asphalt pavement point cloud based on empirical mode decomposition, providing method support for high-precision extraction of pavement three-dimensional texture parameters;

[0044] 3. Aiming at the problem of selecting the reference plane for texture parameter calculation, the present invention proposes a method for extracting texture parameters of steel slag asphalt pavement based on local equidistant planes, which can extract 30 different texture parameters;

[0045] 4. The present invention establishes an end-to-end neural network that inputs multi-dimensional texture parameters and outputs the amount of surface slag, uses classification cross entropy as the loss function, improves analysis efficiency and ensures detection accuracy.

[0046] In summary, the present invention is used to realize the detection of steel slag content in steel slag asphalt pavement based on laser point cloud data. The present invention realizes the automation and intelligence from laser point cloud raw data to steel slag content, improves the technical system in the field of steel slag asphalt pavement, and improves the technical level of the steel slag asphalt pavement detection end, which is conducive to the efficient quality control of steel slag asphalt pavement. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments will be briefly introduced below. Obviously, the drawings in the following description only relate to some embodiments of the present invention, but are not intended to limit the present invention.

[0048] Figure 1 A schematic diagram of a scale used for calibration of a collection device in a preferred embodiment of the present invention;

[0049] Figure 2 This is a flow chart of suppressing vibration noise by empirical mode decomposition in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0050] To make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the embodiments of the present invention will be described clearly and completely in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them.

[0051] Unless otherwise defined, technical or scientific terms used herein shall have the same meaning as commonly understood by one of ordinary skill in the art to which the invention belongs.

[0052] The following is combined with Figure 1-2 The specific embodiments of the present invention are described in detail with examples.

[0053] An intelligent detection method for steel slag asphalt steel slag content based on laser point cloud, comprising the following steps:

[0054] (1) Steel slag asphalt pavement point cloud collection: Integrate the 3D laser scanner and the travel encoder on the vehicle; use a sawtooth ruler to calibrate the 3D laser scanner; drive the inspection vehicle on a steel slag asphalt pavement test section in a certain province for 2 km, and obtain 400 valid dat point cloud data files. The ruler is as shown in the attached figure. Figure 1 shown.

[0055] (2) Point cloud quality improvement: Taking the dat file in (1) as input, the program decodes it into a matrix of size 3521×2163×Z′. The vehicle travel length covered by a single point cloud data file is 3521 pixels (i.e., 3521 measurement lines), and the scan width covered by a single point cloud data file is 2163 pixels (i.e., there are 2163 measurement points on a measurement line). i It is a certain elevation measured by the laser point cloud within the range of 3521×2163; the mean value μ of the elevation of all points in the matrix is ​​2.011, and the standard deviation σ is 0.139. Therefore, the elevation of all points with elevations lower than 1.594 or higher than 2.428 needs to be replaced by the average value of the elevations of the surrounding points.

[0056] (3) Vibration noise suppression program based on empirical mode decomposition: Taking the matrix in (2) as input, the points with the same m coordinates are combined into a one-dimensional array, and a total of 2163 longitudinal measurement lines are obtained, and the shape of each longitudinal measurement line is 3521×1; empirical mode decomposition is performed on each longitudinal measurement line to obtain 2163 groups of 17304 vibration components, that is, each longitudinal measurement line is decomposed into 8 vibration components; using Fourier decomposition, 2163 groups of 17304 vibration spectra are obtained, and the cosine similarity of each vibration spectrum and the measured vibration spectrum is calculated respectively, and 2163 groups of 17304 cosine similarities are obtained. The vibration component number corresponding to the maximum cosine similarity in each group is counted, and the mode of the number is 4. All According to the corresponding Sort, take the top 20 Calculate the equidistant plane of the matrix; take the 10th horizontal line as an example, extract the top 20 The data of the 10th point on the horizontal line is obtained and the equidistant straight line of these 20 points is calculated, which is recorded as the 10th equidistant straight line. It is assumed that the relative positions of the 10th equidistant straight line and all points on the 10th horizontal line remain unchanged before and after the vibration. The deviation between the 10th horizontal line and the equidistant plane is the noise caused by the vertical vibration. The 10th equidistant straight line and all points on the 10th horizontal line are synchronously regressed to the reference plane. The above process is shown in the attached Figure 2 .

[0057] (3) Texture parameter extraction: The 3521×2163×Z′ matrix is ​​divided into blocks and local equidistant planes are taken as an example. The MTD of each block is calculated, which are 0.8707, 0.7897, 0.8848, 0.8161, 0.7782, 0.8074, 0.8897, 0.7619, 0.7539, 0.7566, 0.8761, 0.7589, 0.8442, 0.8313, 0.7835,0.7867,0.8039,0.7931,0.7973,0.8291,0.7850,0.8084,0.7696,0.8304,0.8021, calculate the average value, and the MTD of the matrix is ​​0.8084; use this method to calculate the remaining 29 texture parameters and save them as a dictionary. The dictionary format is {NULL:(0.8084, parameter 2, ... parameter 30)}.

[0058] (4) Establishing a slag content dataset: Annotate the dictionary generated by each matrix. For example, if the slag content of the matrix obtained in (1) is 75%, the dictionary is annotated as {'more': 0.8084, parameter 2, ... parameter 30)}; annotate the 400 matrices obtained in (1) in blocks, and each block forms a piece of data. The data contains the slag content of the block and the corresponding 30 texture parameters, and finally 6000 pieces of data are obtained.

[0059] (5) Detection model construction and training: Set the training parameters in the convolutional neural network: BATCH_SIZE = 32, EPOCHS = 300, LEARNING_RATE = 0.0001, BUFFER_SIZE = 30, model_path = '. / model.h5', set the loss function to categorical cross entropy (CategoricalCrossentropy), then compile the detection model and start training. The output training results will be automatically saved in the working path and named model.h5.

[0060] (6) Detection model verification: The accuracy rate is 92.67%.

[0061] Embodiment 2:

[0062] The computer-readable storage medium of this embodiment stores a computer program thereon, which, when executed by a processor, implements the steps in the intelligent detection method of steel slag asphalt and steel slag dosage based on laser point cloud of Example 1.

[0063] The computer-readable storage medium of this embodiment may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal; the computer-readable storage medium of this embodiment may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card, a secure digital card, a flash memory card, etc. equipped on the terminal; further, the computer-readable storage medium may also include both an internal storage unit of the terminal and an external storage device.

[0064] The computer-readable storage medium of this embodiment is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or is to be output.

[0065] Embodiment 3:

[0066] The computer device of this embodiment includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method for intelligent detection of steel slag asphalt and steel slag content based on laser point cloud of Example 1 are implemented.

[0067] In this embodiment, the processor may be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, readily available programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0068] Those skilled in the art will appreciate that the disclosed content of the embodiments may be provided as methods, systems, or computer program products. Therefore, the present solution may adopt the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Moreover, the present solution may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program codes.

[0069] The present solution is described with reference to the method according to the embodiment of the present solution and the flowchart and / or block diagram of the computer program product. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions; these computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 one or more processes and / or methods Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0070] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 one or more processes and / or methods Figure 1 A function specified in one or more boxes.

[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 one or more processes and / or methods Figure 1The steps for the functions specified in one or more boxes.

[0072] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0073] The examples described in the present invention are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the concept and scope of the present invention. Without departing from the design concept of the present invention, various modifications and improvements made to the technical solutions of the present invention by engineers and technicians in this field should all fall within the protection scope of the present invention.

Claims

1. An intelligent detection method for steel slag asphalt steel slag content based on laser point cloud, characterized by: The method comprises the following steps: Point cloud collection of steel slag asphalt pavement; Abnormal data cleaning method; Vibration noise suppression based on empirical mode decomposition; Texture parameter extraction; Establish a slag dosage data set; Detection model construction and training; Detection model validation.

2. The method according to claim 1, characterized in that The specific steps of collecting point cloud of steel slag asphalt pavement are as follows: Integrate 3D laser scanners, travel encoders and acceleration sensors on vehicles; A bracket is extended backward from the top of the vehicle to set up a 3D laser scanner. The 3D laser scanner is perpendicular to the road surface. The distance between the 3D laser scanner and the road surface is determined according to the scanning range of the 3D laser scanner. One scanning transmits a horizontal survey line to obtain the elevations of M points on the horizontal survey line. Among them, the traveling encoder is installed on the left rear wheel to measure the vehicle's travel distance and periodically trigger the 3D laser scanner; every time the vehicle travels W meters, the traveling encoder triggers the 3D laser scanner N times to obtain N horizontal measurement lines, and saves the elevation scanning results of M×N points on the road surface in the form of a point cloud data file in dat format; The acceleration sensor is attached to the three-dimensional laser scanner and records the vertical motion parameters of the three-dimensional laser scanner at the moment when the encoder triggers the three-dimensional laser scanner.

3. The method according to claim 2, characterized in that The abnormal data cleaning method includes the following steps: taking the dat file as input, the program decodes it into a matrix of size M×N, and calculates the mean μ and variance σ of the elevation in a certain matrix. 2 , the elevation points outside the confidence interval [μ-2σ,μ+2σ] are considered as outliers, and the elevation of the outlier point is assigned the average elevation of the points around it, that is: In the formula, z′ m,n After data cleaning, is the original data.

4. The method according to claim 3, characterized in that ,The vibration noise suppression based on empirical mode decomposition specifically includes the following steps: Including longitudinal survey line extraction, vibration component decomposition, vibration component sorting, equidistant plane calculation and transverse survey line regression; The method of extracting longitudinal lines is as follows: take the cleaned file as input and decode it into a matrix of size N×M, combine the points with the same m coordinates into a one-dimensional array, and obtain a total of M longitudinal lines, which are numbered 1, 2, 3…j,…, M, and the shape of each longitudinal line is N×1; The method of vibration component decomposition is: take a longitudinal survey line as the original signal, make the upper and lower extreme points of the original signal to make the upper and lower envelopes, find the mean curve of the upper and lower envelopes, and use the original signal and the mean curve to make a difference to get the intermediate signal. If the number of extreme points and zero-crossing points of the intermediate signal in the entire data segment is equal or the difference is no more than one, and the average value of any upper envelope formed by the local maximum point and the lower envelope formed by the local minimum point is zero, then the vibration component decomposition is completed. If the intermediate signal does not meet the above conditions, it is recorded as vibration component No. 1, and the interpolation of the original signal and vibration component No. 1 is used as the new original signal. Repeat this step until the vibration component decomposition is completed. The vibration components are numbered 1, 2, ..., i, ... I respectively. Repeat the above steps for N longitudinal survey lines to obtain M groups of I×M vibration components Where j represents the number of the longitudinal measurement line, and i represents the number of the vibration component obtained from the jth longitudinal measurement line; The method of sorting vibration components is to extract the motion parameter curve of the N×M matrix during the acquisition period and perform Fourier transform to obtain the measured vibration spectrum S0, and then sort the vibration components obtained in the previous step. Perform Fourier transform to obtain a series of vibration spectra The meanings of i and j are the same as above, respectively calculating S0 and each Cosine similarity of Extract the vibration component number corresponding to the maximum value of R in each measuring line and find the mode k of the component number. According to the corresponding Sort by, take the top 20 The method for calculating the equidistant plane is to set the plane with the expression Ax+By+Cz+D=0 to have the minimum distance from all points in the N×M matrix, randomly give a set of initial values ​​A=a0, B=b0, C=C0, D=d0, and calculate the distance sum0 according to the following formula: Use the conjugate gradient method to update the initial value until the deviation of sum0 from 0 converges to 10 -6 Within, we get the isometric plane ax+by+cz+d=0; The method of horizontal line regression is to extract the top 20 horizontal lines for each of the nth horizontal lines. The data of the nth point on the graph are used to calculate the equidistant straight line of these 20 points, which is recorded as the nth equidistant straight line. The method is the same as the calculation method of the equidistant plane. It is assumed that the relative positions of the nth equidistant straight line and all points on the nth transverse line remain unchanged before and after the vibration occurs. The deviation between the nth transverse line and the equidistant plane is the noise caused by the vertical vibration. The nth equidistant straight line and all points on the nth transverse line are synchronously regressed to the equidistant plane, that is: Where: is the elevation value after noise suppression, To obtain the solution after substituting the point cloud with coordinates (m,n) into the equidistant plane equation, This is the solution result after substituting the point cloud with coordinates (m,n) into the equation of the nth equidistant straight line.

5. The method according to claim 4, characterized in that The specific steps of texture parameter extraction are as follows: for a larger N×M point cloud matrix, divide it into p small blocks (A1, A2, …, A p ), calculate the isometric plane of each small block respectively; Among them, taking the local equidistant plane as the standard, calculate a local area Various texture parameters, including mean cross-sectional depth MPD, arithmetic mean roughness R a , RMS height R q , skewness R sk , average structural depth MTD, etc., a certain parameter of the entire matrix is ​​the average value of the parameter in each local area.

6. The method according to claim 5, characterized in that The specific steps to establish a steel slag dosage dataset are as follows: use the detection equipment in the steel slag asphalt pavement point cloud acquisition to conduct detection on steel slag asphalt pavement or steel slag asphalt concrete samples with known steel slag dosage, and link the texture parameters with the steel slag dosage to construct data.

7. The method according to claim 6, characterized in that The specific steps of building and training the detection model are as follows: load the training data set, convert the data set to "one-hot" label format, build a convolutional neural network based on tensorflow-gpu, set the training parameters: BATCH_SIZE = b, EPOCHS = p, LEARNING_RATE = α, BUFFER_SIZE = t, model_path = str1, and set the loss function to categorical cross entropy (CategoricalCrossentropy) After that, compile the model and start training. The output training results will be automatically saved in the model_path path.

8. The method according to claim 7, characterized in that The specific steps of the detection model verification are: load the verification data set and the trained network, if the accuracy of the training data set is If it exceeds 90%, the model can be applied. If it is less than 70%, it is necessary to increase the richness of the data set, repeat the steps of establishing the slag dosage data set and appropriately adjust the parameters in the detection model construction and training steps until the verification accuracy reaches more than 90%.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the intelligent detection method for steel slag asphalt and steel slag content based on laser point cloud and random sample consistency as described in any one of claims 1 to 8 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the intelligent detection method for steel slag asphalt and steel slag content based on laser point cloud and random sample consistency are implemented as described in any one of claims 1 to 8.

Images