CN-121982645-A - Pavement crack monitoring method and system

CN121982645ACN 121982645 ACN121982645 ACN 121982645ACN-121982645-A

Abstract

The invention relates to a pavement crack monitoring method and system, and belongs to the technical field of pavement state monitoring. The method solves the problems of insufficient identification reliability, lack of three-dimensional geometric information and incapability of dynamic prediction caused by dependence on single sensing data in the prior art. Synchronously acquiring multi-mode sensing data comprising a two-dimensional image and a three-dimensional point cloud, establishing a mapping relation between two-dimensional pixels and a three-dimensional space through space-time registration, extracting candidate crack areas based on a crack identification algorithm and projecting the candidate crack areas to the three-dimensional point cloud to generate candidate point clouds, calculating geometrical characteristic parameters including depth variance and normal vector consistency indexes, judging real cracks through comparison with a preset threshold value and outputting three-dimensional form information, finally associating the same crack through characteristic matching based on multi-period three-dimensional form information, calculating an expansion rate, and predicting the residual service life according to a material fatigue model. The method and the system are mainly used for crack detection and state evaluation in road maintenance management.

Inventors

ZHAO JIANBO
WU YANG
WANG FENGWEI
LI JIAXUN
HE LIHENG
DAI WEIWEI
LV DEPIN
LIU LIPENG
WANG JIAMING
Ji Yilei
ZHAI SEN
ZHANG DONGDONG
LI TAO
HU XIANBO
ZHANG XUYANG
HUA XIANG
CHEN BO
Xing Qingfang

Assignees

邢台路桥建设集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260129

Claims (10)

1. The pavement crack monitoring method is characterized by comprising the following steps of: Synchronously acquiring multi-mode sensing data of a road surface to be monitored, wherein the multi-mode sensing data comprises two-dimensional image data acquired through an image sensor and three-dimensional point cloud data acquired through a laser radar; Performing space-time registration on the two-dimensional image data and the three-dimensional point cloud data to establish a mapping relation between two-dimensional pixel coordinates and three-dimensional space coordinates; extracting candidate crack areas from the two-dimensional image data based on a preset crack identification algorithm; Projecting the candidate crack areas into the three-dimensional point cloud data according to the mapping relation, so as to generate candidate three-dimensional crack point clouds; calculating geometrical characteristic parameters of the candidate three-dimensional crack point cloud set, wherein the geometrical characteristic parameters comprise depth variance and normal vector consistency indexes; Comparing the geometric characteristic parameter with a preset crack discrimination threshold, and when the depth variance is larger than the depth threshold and the normal vector consistency index is lower than the consistency threshold, judging the candidate crack area as a real crack and outputting three-dimensional morphological information of the real crack; Based on the three-dimensional form information of the real crack obtained in different monitoring periods, correlating the same crack through characteristic matching, calculating the expansion rate of the same crack according to the three-dimensional form information of the same crack at different time points, and predicting the residual service life of the same crack according to a material fatigue model and the expansion rate.
2. The method for monitoring a pavement crack according to claim 1, wherein, The step of performing space-time registration on the two-dimensional image data and the three-dimensional point cloud data specifically comprises the following steps: At least three reference targets with known accurate three-dimensional coordinates are distributed in the common field of view of the image sensor and the laser radar; Synchronously triggering the image sensor and the laser radar to respectively acquire a two-dimensional calibration image and three-dimensional point cloud data containing the reference target; Extracting a two-dimensional image pixel center coordinate of the reference target from the two-dimensional calibration image, and extracting a three-dimensional point cloud center coordinate of the reference target from the three-dimensional point cloud data; calculating a space transformation matrix between the image sensor and the laser radar by solving a perspective n-point positioning problem based on the two-dimensional image pixel center coordinates and the corresponding three-dimensional point cloud center coordinates; and taking the space transformation matrix as the mapping relation for subsequently projecting the candidate crack region from the two-dimensional image data to the three-dimensional point cloud data.
3. The method for monitoring a pavement crack according to claim 1, wherein, The setting method of the depth threshold and the consistency threshold comprises the following steps: After the synchronous acquisition step, extracting three-dimensional point clouds of a plurality of lossless areas of the pavement to be monitored from the three-dimensional point cloud data; Calculating depth variance and normal vector consistency indexes of three-dimensional point clouds of the plurality of lossless areas as background noise statistics; And determining the depth threshold and the consistency threshold according to a preset confidence level based on the background noise statistic, wherein the depth threshold is set as a coefficient which is obtained by multiplying a value of a depth variance in the background noise statistic by more than 1, and the consistency threshold is set as a coefficient which is obtained by multiplying a value of a normal vector consistency index in the background noise statistic by more than 1.
4. The method for monitoring a pavement crack according to claim 1, wherein, The step of associating the same crack through feature matching specifically comprises the following steps: Extracting a first feature vector of any real crack acquired in the current monitoring period, wherein the first feature vector at least comprises a global geometric feature and a spatial context feature of the crack, the global geometric feature comprises the length, the trend and the minimum circumscribed rectangle of the crack, and the spatial context feature comprises the relative position relation between the crack and a preset road surface datum line or other adjacent fixed road surface features; Performing traversal search on all real cracks acquired in the previous monitoring period, and extracting a second feature vector corresponding to each crack; calculating the similarity between the first feature vector and each second feature vector; And when a second feature vector with the similarity with the first feature vector being higher than a preset matching threshold and the spatial position deviation being smaller than a preset distance threshold exists, judging that the crack in the current monitoring period and the crack in the last monitoring period are the same crack.
5. The method for monitoring a pavement crack according to claim 1, wherein, In the step of predicting the remaining service life according to the material fatigue model, the material fatigue model is subjected to field calibration by the following modes: Acquiring a sample test piece of the same material as the pavement in an adjacent area of the pavement to be monitored; Carrying out indoor fatigue tests on the sample test piece under different stress levels, and recording the cycle times of the sample test piece under each stress level until the sample test piece is destroyed so as to obtain actual fatigue performance data of the pavement material; performing inversion analysis and calibration on key parameters in the material fatigue model based on the actual fatigue performance data; And using the calibrated material fatigue model for predicting the residual service life.
6. The method for monitoring a pavement crack according to claim 1, wherein, Before the step of projecting the candidate crack region into the three-dimensional point cloud data, preprocessing the three-dimensional point cloud data specifically includes: performing downsampling processing on the three-dimensional point cloud data through a voxel grid downsampling algorithm, wherein the size of the voxel grid is set to be not smaller than the size of the maximum aggregate particle size on the surface of the pavement material; And executing the step of generating candidate three-dimensional crack point clouds and subsequent geometric feature calculation on the down-sampled three-dimensional point cloud data.
7. The method for monitoring a pavement crack according to claim 2, wherein, Synchronously acquiring inertial measurement unit data for monitoring the pose change of the sensor in the step of synchronously triggering the image sensor and the laser radar; The method further comprises an online calibration step: based on the inertial measurement unit data, calculating the pose change quantity of the mobile monitoring platform between two adjacent acquisition moments; And dynamically correcting the relative pose error between the image sensor and the laser radar caused by platform vibration according to the pose variation and the space transformation matrix so as to update the mapping relation.
8. The method for monitoring a pavement crack according to claim 3, wherein, The step of extracting the plurality of nondestructive areas of the pavement to be monitored from the three-dimensional point cloud data specifically comprises the following steps: Based on the two-dimensional image data, utilizing a pre-trained pavement material segmentation model to identify a region which has uniform materials and no visual texture abnormality as a candidate lossless region; projecting the candidate lossless region onto the three-dimensional point cloud data according to the mapping relation; Calculating the geometric flatness of the three-dimensional point clouds corresponding to each candidate lossless region, wherein the geometric flatness is the root mean square value of the deviation between the elevation of all the point clouds in the region and the elevation of the fitting reference plane; and screening out candidate lossless regions with the geometric flatness lower than a preset flatness threshold value, and determining the candidate lossless regions as the lossless regions for calculating background noise statistics.
9. The method for monitoring a pavement crack according to claim 4, wherein, The first feature vector and the second feature vector further comprise topological features; The method for extracting the topological structure features comprises the following steps: Converting a two-dimensional skeleton line of the crack into a topological graph structure, wherein nodes of the topological graph structure represent intersection points and end points of a skeleton of the crack, and edges represent skeleton branches connecting the nodes; and calculating a graph isomorphism invariant of the topological graph structure as the topological structure characteristic, wherein the graph isomorphism invariant at least comprises a degree value sequence of each node.
10. A pavement crack monitoring system, comprising: The data acquisition module is used for synchronously acquiring multi-mode sensing data of the road surface to be monitored, wherein the multi-mode sensing data comprises two-dimensional image data acquired through an image sensor and three-dimensional point cloud data acquired through a laser radar; The data registration module is used for carrying out space-time registration on the two-dimensional image data and the three-dimensional point cloud data so as to establish a mapping relation between two-dimensional pixel coordinates and three-dimensional space coordinates; the preliminary identification module is used for extracting candidate crack areas from the two-dimensional image data based on a preset crack identification algorithm; the three-dimensional mapping module is used for projecting the candidate crack areas into the three-dimensional point cloud data according to the mapping relation so as to generate candidate three-dimensional crack point clouds; The geometric verification module is used for calculating geometric feature parameters of the candidate three-dimensional crack point cloud set, wherein the geometric feature parameters comprise depth variance and normal vector consistency indexes, comparing the geometric feature parameters with preset crack discrimination thresholds, judging the candidate crack region as a real crack when the depth variance is larger than the depth threshold and the normal vector consistency indexes are lower than the consistency threshold, and outputting three-dimensional morphological information of the real crack; The prediction analysis module is used for associating the same crack through characteristic matching based on the three-dimensional form information of the real crack acquired in different monitoring periods, calculating the expansion rate of the same crack according to the three-dimensional form information of the same crack at different time points, and predicting the residual service life of the same crack according to a material fatigue model and the expansion rate.

Description

Pavement crack monitoring method and system Technical Field The invention belongs to the technical field of pavement condition monitoring, and particularly relates to a pavement crack monitoring method and system. Background In the field of pavement engineering maintenance, regular and accurate monitoring of pavement cracks is a basic task for evaluating pavement health conditions and making scientific maintenance strategies. The existing monitoring methods mainly depend on manual inspection and automatic detection technologies, but the methods still have a plurality of problems to be solved in terms of accuracy, efficiency and prediction capability. Manual inspection is a traditional method which is still widely adopted at present. The method mainly relies on naked eye observation and experience judgment of maintenance personnel, and the found cracks are recorded manually and simply measured. This approach is subject to significant subjectivity and uncertainty, and there may be differences in the criteria for different people to judge the same fracture, resulting in poor consistency and comparability of the data. In addition, manual inspection efficiency is low, traffic needs to be sealed, cost is high, potential safety hazards exist, and the method is difficult to be suitable for road network general inspection in a large range and at high frequency. The root cause is that the method lacks objective quantification standard and efficient detection means. Automated detection techniques based on digital images are a current research hotspot. Such techniques typically acquire two-dimensional images of the road surface with an onboard camera and automatically identify the crack using image processing algorithms. However, this approach presents a serious challenge in practical applications. Firstly, the two-dimensional image is easily interfered by the change of ambient light, road surface stains, oil spots, shadows and complex texture background, so that a large number of false positives and false negatives are generated by the recognition algorithm. For example, linear stains of similar color to the cracks may be misjudged as cracks, while micro cracks in the case of insufficient illumination are easily missed. Secondly, two-dimensional image-based analysis methods have difficulty in acquiring key three-dimensional geometric properties of the fracture, especially depth information. And the depth of the fracture is an important indicator for assessing its severity and structural integrity of the pavement. The lack of three-dimensional information causes the assessment result to stay on the surface, and cannot provide sufficient basis for structural maintenance decisions. In order to overcome the limitation of two-dimensional vision, there are studies on the introduction of three-dimensional sensing technologies, such as laser radars, to acquire three-dimensional topography data of a road surface. However, how to effectively fuse the two-dimensional image information with the three-dimensional point cloud data and accurately extract the crack features therefrom is a technical difficulty. The simple data superposition can not automatically improve the recognition precision, and if the two-dimensional recognition result has errors, the subsequent three-dimensional analysis can be directly caused to be established on the basis of the errors. In addition, the three-dimensional point cloud data volume is huge, the direct processing can bring huge calculation load, the detection efficiency is affected, and how to perform high-efficiency data processing on the premise of not losing key characteristics is a practical difficulty in engineering application. Furthermore, most of the prior art focuses on the immediate identification and static parameter measurement of the crack, and lacks the capability of analyzing the dynamic evolution trend of the crack. Pavement cracks are a disease which continuously develops along with time, and the state of the pavement cracks at a certain moment is only known, so that the pavement cracks are insufficient for supporting predictive maintenance. To achieve crack tracking across time periods, the difficult problem of how to accurately correlate and match cracks detected at different times needs to be solved. The length and shape of the crack may change during the propagation process, which makes stable and reliable matching difficult. In addition, correlating crack propagation with fatigue performance of pavement materials and predicting residual service life thereof requires localization calibration of material models, and there is often a difference between actual performance of general models and specific road section materials, so how to obtain localization parameters with low cost and high efficiency is a challenge in practice. In summary, the existing pavement crack monitoring technology has technical bottlenecks of different degrees in the aspects of accuracy of multi-source inf