Method and system for detecting paving segregation property of water-stable material

Abstract

The invention provides a method and a system for detecting paving segregation of a water-stable material. The method comprises the following steps: S1, synchronously collecting a three-dimensional point cloud data set and hyperspectral data of a paving layer at intervals along a paving belt through laser scanning and multispectral imaging; s2, noise reduction and registration are performed on the point cloud data, and then a combined feature matrix of elevation and intensity is constructed by combining the hyperspectral data; s3, extracting a particle size distribution probability matrix through a PointNet + + network, and obtaining a segregation probability matrix; s4, calculating a unit space variation coefficient and a unit segregation degree; and S5, dividing segregation grades according to a preset segregation degree threshold value, and triggering early warning. Laser scanning and multispectral imaging technologies are fused, and a PointNet + + network is combined to realize non-contact high-precision detection of paving and segregation of the water-stable material, so that the problems of high subjectivity and high missing judgment rate of a traditional visual inspection method are solved; the separation degree threshold value is dynamically adjusted to adapt to environmental parameter changes, the construction process adjustment is guided through a real-time early warning mechanism, and the paving quality control efficiency is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of segregation detection, and in particular to a method and system for detecting the segregation of water-stabilizing material paving. Background Art

[0002] In road base construction, inorganic binder-stabilized coarse-grained soil and unbound crushed stone are widely used due to their excellent mechanical properties and applicability. Paving quality, a key factor influencing base structural performance, is directly related to the overall bearing capacity and durability of the road. However, in actual construction, material segregation is a common phenomenon, whereby coarse and fine aggregates are unevenly distributed during paving, resulting in unstable material grading and significant local strength differences. This unevenness not only weakens the mechanical properties of the base material and reduces its overall stability, but can also trigger stress concentration under subsequent loads, leading to structural defects such as base cracking and settlement, and ultimately inducing premature pavement damage. The resulting shortened road service life and increased maintenance costs have become a major bottleneck restricting the improvement of road engineering quality.

[0003] At present, construction sites mainly rely on manual visual inspection to determine whether there is segregation in paving materials. This method is simple to operate and low-cost, but it has many limitations. First, the visual inspection method is highly dependent on the experience and subjective judgment of the inspectors, lacks a unified standard, and is easily interfered with by human factors, resulting in poor repeatability and accuracy of the test results; second, the visual inspection method can only identify more obvious segregation areas, and it is difficult to accurately identify minor or potential segregation problems, and there is a greater risk of missed judgment; third, this method usually requires on-site spot checks after paving is completed, which is not real-time and cannot provide timely feedback and adjust the paving process. It is easy to cause large-scale segregation areas to be discovered only after they are formed, increasing rework costs. Therefore, a method and system for detecting segregation in water-stabilizing material paving is proposed to solve the above problems. Summary of the Invention

[0004] The main purpose of the present invention is to provide a method and system for detecting the segregation of water-stabilizing material paving, so as to solve the problem that the existing visual inspection method is difficult to achieve a comprehensive and systematic evaluation of the entire paving surface, the detection coverage is low, and it is difficult to meet the requirements of modern road engineering quality management for refinement, informatization and intelligent development.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: a method and system for detecting the segregation of water-stabilizing material paving, the method comprising: S1. Collect 3D point cloud datasets and hyperspectral data of the paving layer at synchronous intervals along the paving belt through laser scanning and multispectral imaging; S2, denoise and register the point cloud data, and then construct a joint feature matrix of elevation and intensity based on the hyperspectral data; S3. Extract the particle size distribution probability matrix through the PointNet++ network and obtain the segregation probability matrix; S4. Calculate the unit spatial variation coefficient and unit segregation degree; S5. Segregation levels are determined according to a preset segregation threshold, and an early warning is triggered.

[0006] In the preferred solution, the point cloud dataset in step S1 is: ; Where, For a complete point cloud dataset; For the A three-dimensional space point; For the The horizontal coordinate of a point; For the The vertical coordinate of a point; For the The elevation value of a point; For the The reflection intensity of each point; is the unique index identifier of the point; For collection Contains from arrive common points; The specific method of step S2 is: perform voxel grid filtering to reduce noise on the point cloud data, and perform multi-frame registration using the ICP algorithm; based on the registered point cloud data, construct a joint feature matrix containing the normalized intensity of elevation and the normalized intensity of spectral dynamics: ; Where, is the elevation value; is the reflection intensity; is the global minimum intensity; dynamic range = .

[0007] In the preferred solution, the specific method for extracting the separated features in step S3 is to construct a PointNet++ network structure, where the input layer receives point cloud data, the feature encoding layer embeds a local attention mechanism, and the output layer generates a particle size distribution probability matrix: ; Where, It is the three-dimensional probability matrix output by the PointNet++ network, representing the probability density of different particle sizes at each location in the detection area; is the spatial resolution of the detection area; is the particle size classification number; After Softmax normalization, the separation probability matrix is ​​obtained: ; Where, is the input point cloud feature; is the local attention module; Represents the distribution probability of m types of particle sizes.

[0008] In a preferred embodiment, the calculation formula for the spatial variation coefficient of each detection unit in step S4 is: ; Where, is the spatial variation coefficient of the jth detection unit, which quantifies the degree of dispersion of the particle size distribution; For the The standard deviation of the particle size of each detection unit measures the degree to which the data deviates from the mean; For the The mean particle size of each detection unit represents the average particle size of the unit; The calculation formula of unit segregation is: ; Where, For the Separation of each detection unit; is the material sensitivity coefficient; It is the mean value of the design benchmark particle size, which is determined by the construction mix ratio; is the base of natural logarithms.

[0009] The preferred solution further includes: S6, calculation of overall segregation, the formula of which is: ; Where, is the total number of detection units, For the The weighting coefficient of each unit; Segregation levels are determined based on the preset overall segregation threshold and an early warning is triggered.

[0010] In the preferred embodiment, the segregation threshold value is adjusted according to the environmental parameters of temperature and humidity to adjust the segregation alarm threshold value, and the formula is: ; Where, is the initial threshold; is the deviation of the real-time temperature from the standard temperature (25°C); For relative humidity, the threshold is directly weakened to compensate for the effect of humidity on material properties.

[0011] In the preferred solution, the method for synchronous interval acquisition along the paving belt in step S1 adopts the variable density scanning interval formula: ; Where, The scanning interval is dynamically adjusted according to the resolution; is the degree of segregation; In this process, Kalman filtering is used to predict missing point clouds.

[0012] In a preferred embodiment, the segregation threshold is determined by static testing before construction. The specific method includes: preparing three sets of standardized samples with mix proportions consistent with the construction design; spraying barium sulfate developer on the samples and sealing the edges with epoxy resin; placing the samples at the center of a rotating platform and rotating them to obtain a three-dimensional point cloud dataset and hyperspectral data; then calculating the segregation using the method of steps S2-S4, and determining the segregation threshold based on the average segregation value of the three sets of samples.

[0013] The system includes: A scanning module, which includes a laser scanning module and a multispectral imaging module, is used to obtain three-dimensional point cloud data and hyperspectral data of the paving layer; Environmental detection module, including temperature sensor and humidity sensor; A data processing module, which collects data from the scanning module and the environment detection module, and implements the steps of any one of the methods of claims 1-8; An evaluation and decision module divides the segregation level calculated by the data processing module according to a preset segregation threshold; Early warning feedback module: when the level classified by the evaluation and decision module is abnormal, an audible and visual alarm will be sounded and an early warning message will be issued; Terminal, used to receive warning information and input parameters into the data processing module; Wireless connection module, used for wireless signal connection between modules.

[0014] In the preferred solution, the scanning module and the environmental detection module are both arranged on a gantry at the rear of the paver, and the gantry can move along the paving direction.

[0015] The present invention provides a method and system for detecting the segregation of water-stabilizing material paving. By integrating laser scanning and multispectral imaging technology and combining with the PointNet++ network, non-contact, high-precision detection of water-stabilizing material paving segregation can be achieved, solving the problems of strong subjectivity and high missed judgment rate of traditional visual inspection methods. The segregation threshold is dynamically adjusted to adapt to changes in environmental parameters, and a real-time early warning mechanism is used to guide construction process adjustments, significantly improving the efficiency of paving quality control and reducing rework costs. The system has intelligent, precise and practical engineering value. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below with reference to the accompanying drawings and examples: Figure 1 It is a schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0017] Example 1 A method for detecting segregation of water-stabilizing material paving, the method comprising: S1. Data acquisition parameters: 3D point cloud data sets and hyperspectral data of the paving layer are collected at intervals along the paving belt using laser scanning and multispectral imaging. The point cloud data sets are: ; Where, For a complete point cloud dataset; For the A three-dimensional space point; For the The horizontal coordinate of a point; For the The vertical coordinate of a point; For the The elevation value of a point; For the The reflection intensity of each point; is the unique index identifier of the point; For collection Contains from arrive common points.

[0018] It should be noted that the laser scanning and multispectral imaging are time synchronized (error < 1ms), the spatial registration error is < 5cm, and the synchronization mechanism is to achieve multi-device synchronization via CAN bus. Level-one time synchronization is achieved by using a calibration target sphere to align the field of view of the laser scanner and the multispectral imager.

[0019] S2. Data preprocessing and feature fusion: De-noise and register the point cloud data. Specifically, isolated noise points are removed by voxel grid filtering, and the grid size is 2-5mm, preferably 2mm. Multi-frame registration is performed by ICP algorithm, and the registration error is <2mm. Then, a joint feature matrix of elevation and intensity is constructed based on the registered point cloud data and hyperspectral data to form Resolution space-material feature matrix , the joint feature matrix of the elevation normalized intensity and the spectral dynamic normalized intensity is: ; Where, is the elevation value; is the reflection intensity; is the global minimum intensity; dynamic range = .

[0020] S3. Segregation feature extraction: Extract the particle size distribution probability matrix through the PointNet++ network and obtain the segregation probability matrix.

[0021] The specific method is to build the PointNet++ network structure, the input layer: receiving Point cloud data Feature encoding layer: Three-layer Set Abstraction module, extracting local features layer by layer (number of sampling points: 512→128→32), while embedding the local attention mechanism (LA-Module) to enhance sensitivity to isolated areas. The output layer generates a three-dimensional particle size distribution probability matrix: ; Where, It is the three-dimensional probability matrix output by the PointNet++ network, representing the probability density of different particle sizes at each location in the detection area; is the spatial resolution of the detection area; is the particle size classification number ( =50 particle size classification); After Softmax normalization, the separation probability matrix is ​​obtained. Mapped into a segregation probability matrix, The larger the value, the higher the segregation risk at that location. The matrix model is: ; Where, is the input point cloud feature; is the local attention module; Represents the distribution probability of m types of particle sizes.

[0022] S4. Segregation calculation, calculate the unit spatial variation coefficient and unit segregation.

[0023] The calculation formula of the spatial variation coefficient of each detection unit is: ; Where, is the spatial variation coefficient of the jth detection unit, which quantifies the degree of dispersion of the particle size distribution; For the The standard deviation of the particle size of each detection unit measures the degree to which the data deviates from the mean; For the The mean particle size of each detection unit represents the average particle size of the unit.

[0024] Among them, the mean particle size is: ; Particle size standard deviation: ; The calculation formula of unit segregation is: ; Where, For the Separation of each detection unit; is the material sensitivity coefficient; It is the mean value of the design benchmark particle size, which is determined by the construction mix ratio; is the base of natural logarithms.

[0025] S5. Segregation levels are determined according to a preset segregation threshold, and an early warning is triggered.

[0026] In this embodiment, the segregation threshold is determined by a static test before construction. The specific method includes: preparing three groups of standardized samples with a mix ratio consistent with the construction design; spraying a barium sulfate developer with a concentration of 0.5% (drying time < 2 minutes) on the samples, and sealing the edges with epoxy resin (to prevent edge effects from interfering with the scan); placing the samples at the center of a rotating platform and rotating them to obtain a three-dimensional point cloud dataset and hyperspectral data; then calculating the segregation using the method of steps S2-S4, and determining the segregation threshold based on the average segregation of the three groups of samples. The threshold technical formula is: ; Where, is the initial segregation threshold; is the sum of the segregation degrees of the three groups of static test samples; For the The segregation of a group of static samples is calculated through laboratory calibration tests.

[0027] In this embodiment, three groups of samples The values ​​are 0.63, 0.65, and 0.67 respectively, and the threshold is 0.65. Therefore, in this embodiment, when Normal construction when <0.3; when 0.3≤ When <0.65, adjust the paving speed or vibration frequency; when When ≥0.65, immediate partial rework is required.

[0028] S6. Calculation of overall segregation: The formula is: ; Where, is the total number of detection units, For the The weighting coefficient of each unit.

[0029] Segregation levels are divided according to the preset overall segregation threshold, and an early warning is triggered. The threshold is 0.65.

[0030] This allows analysis of the overall segregation compliance.

[0031] In the preferred embodiment, the segregation threshold value is adjusted according to the environmental parameters of temperature and humidity to adjust the segregation alarm threshold value, and the formula is: ; Where, is the initial threshold; is the deviation of the real-time temperature from the standard temperature (25°C); For relative humidity, the threshold is directly weakened to compensate for the effect of humidity on material properties.

[0032] In the preferred solution, the method for synchronous interval acquisition along the paving belt in step S1 adopts the variable density scanning interval formula: ; Where, The scanning interval is dynamically adjusted according to the resolution; is the degree of segregation; In this process, Kalman filtering is used to predict missing point clouds, with a prediction error of <3 cm (verified by test data).

[0033] The following table is the verification data: Compared with the screening method:

[0034] Comparative Experimental Design: Comparison with Traditional Methods Efficiency test:

[0035] Extreme working condition testing Environmental adaptability:

[0036] Example 2 Further illustrate with reference to Example 1, Figure 1 The structure shown is a water-stabilizing material paving segregation detection system, comprising: The scanning module includes a laser scanning module and a multispectral imaging module, which are used to obtain three-dimensional point cloud data and hyperspectral data of the paving layer. In this embodiment, the laser scanning module is a line laser scanner, and the multispectral imaging module is a hyperspectral camera. The module has a built-in annular LED fill light and a black light-absorbing inner wall. The scanning module is set on the gantry behind the paver. The gantry can move along the paving direction to ensure the continuity of data collection. A shock-proof rubber base is set at the installation location.

[0037] The environmental detection module includes a temperature sensor and a humidity sensor. The environmental detection module is also set on the gantry behind the paver.

[0038] The data processing module collects data from the scanning module and the environmental detection module and implements the steps in the method of Example 1. The data processing module is equipped with a high-performance GPU computing unit to support real-time point cloud processing and deep learning reasoning.

[0039] The evaluation and decision-making module divides the segregation levels calculated by the data processing module according to a preset segregation threshold.

[0040] The early warning feedback module, when the level divided by the evaluation and decision module is an abnormal level, will sound an audible and visual alarm and issue an early warning message. It should be noted that the early warning feedback module includes an audible and visual alarm device.

[0041] The terminal is used to receive warning information and input parameters into the data processing module.

[0042] Wireless connection module, used for wireless signal connection between modules.

[0043] The above embodiments are merely preferred technical solutions of the present invention and should not be construed as limiting the present invention. The scope of protection of the present invention shall be the technical solutions set forth in the claims, including equivalent alternatives to the technical features of the technical solutions set forth in the claims. In other words, equivalent alternatives and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for detecting the segregation of water-stabilizing material paving, characterized by: The method includes: S1. Collect 3D point cloud datasets and hyperspectral data of the paving layer at synchronous intervals along the paving belt through laser scanning and multispectral imaging; S2, denoise and register the point cloud data, and then construct a joint feature matrix of elevation and intensity based on the hyperspectral data; S3. Extract the particle size distribution probability matrix through the PointNet++ network and obtain the segregation probability matrix; S4. Calculate the unit spatial variation coefficient and unit segregation degree; S5. Segregation levels are determined according to a preset segregation threshold, and an early warning is triggered.

2. The method for detecting the segregation of water-stabilizing material paving according to claim 1, wherein: The point cloud dataset in step S1 is: ； Where, For a complete point cloud dataset; For the A three-dimensional space point; For the The horizontal coordinate of a point; For the The vertical coordinate of a point; For the The elevation value of a point; For the The reflection intensity of each point; Small standard is the unique index identifier of the point; For collection Contains from arrive common points; The specific method of step S2 is: perform voxel grid filtering to reduce noise on the point cloud data, and perform multi-frame registration using the ICP algorithm; based on the registered point cloud data, construct a joint feature matrix containing the normalized intensity of elevation and the normalized intensity of spectral dynamics: ； Where, is the elevation value; is the reflection intensity; is the global minimum intensity; dynamic range = .

3. The method for detecting the segregation of water-stabilizing material paving according to claim 1, wherein: The specific method for extracting the separated features in step S3 is to construct a PointNet++ network structure. The input layer receives point cloud data, the feature encoding layer embeds the local attention mechanism, and the output layer generates the particle size distribution probability matrix: ； Where, It is the three-dimensional probability matrix output by the PointNet++ network, representing the probability density of different particle sizes at each location in the detection area; is the spatial resolution of the detection area; is the particle size classification number; After Softmax normalization, the separation probability matrix is ​​obtained: ； Where, is the input point cloud feature; is the local attention module; Represents the distribution probability of m types of particle sizes.

4. The method for detecting the segregation of water-stabilizing material paving according to claim 1, wherein: The calculation formula of the spatial variation coefficient of each detection unit in step S4 is: ； Where, is the spatial variation coefficient of the jth detection unit, which quantifies the degree of dispersion of the particle size distribution; For the The standard deviation of the particle size of each detection unit measures the degree to which the data deviates from the mean; For the The mean particle size of each detection unit represents the average particle size of the unit; The calculation formula of unit segregation is: ； Where, For the Separation of each detection unit; is the material sensitivity coefficient; It is the mean value of the design benchmark particle size, which is determined by the construction mix ratio; is the base of natural logarithms.

5. A method for detecting segregation of water-stabilizing material paving according to claim 4, characterized in that: Also includes: S6. Calculation of overall segregation: The formula is: ； Where, is the total number of detection units, For the The weighting coefficient of each unit; Segregation levels are determined based on the preset overall segregation threshold and an early warning is triggered.

6. A method for detecting segregation of water-stabilizing material paving according to claim 5, characterized in that: The segregation threshold adjusts the segregation alarm threshold according to the environmental parameters of temperature and humidity. The formula is: ； Where, is the initial threshold; is the deviation of the real-time temperature from the standard temperature (25°C); For relative humidity, the threshold is directly weakened to compensate for the effect of humidity on material properties.

7. A method for detecting the segregation of water-stable material paving according to claim 1, characterized in that: The method of synchronous interval acquisition along the paving belt in S1 adopts the variable density scanning interval formula: ； Where, The scanning interval is dynamically adjusted according to the resolution; is the degree of segregation; In this process, Kalman filtering is used to predict missing point clouds.

8. The method for detecting the segregation of water-stabilizing material paving according to claim 1, wherein: The segregation threshold is determined through static testing before construction. The specific method includes: preparing three sets of standardized samples with mix proportions consistent with the construction design; spraying barium sulfate developer on them and sealing the edges with epoxy resin; placing the samples at the center of a rotating platform and rotating them to obtain a three-dimensional point cloud dataset and hyperspectral data; then, calculating the segregation using the method in steps S2-S4, and determining the segregation threshold based on the average segregation value of the three sets of samples.

9. A water-stabilizing material paving segregation detection system, characterized by: include: A scanning module, which includes a laser scanning module and a multispectral imaging module, is used to obtain three-dimensional point cloud data and hyperspectral data of the paving layer; Environmental detection module, including temperature sensor and humidity sensor; A data processing module, which collects data from the scanning module and the environment detection module, and implements the steps of any one of the methods of claims 1-8; An evaluation and decision module divides the segregation level calculated by the data processing module according to a preset segregation threshold; Early warning feedback module: when the level classified by the evaluation and decision module is abnormal, an audible and visual alarm will be sounded and an early warning message will be issued; Terminal, used to receive warning information and input parameters into the data processing module; Wireless connection module, used for wireless signal connection between modules.

10. A water-stabilizing material paving segregation detection system according to claim 9, characterized in that: The scanning module and the environmental detection module are both installed on the gantry at the rear of the paver, and the gantry can move along the paving direction.

Images