A rapid detection method for large deformation and settlement of highway surface based on 3D laser point cloud

Abstract

The present invention discloses a method for rapid detection of large deformation and settlement of a highway road surface based on a three-dimensional laser point cloud, comprising: obtaining three-dimensional point cloud data of the highway road surface based on a three-dimensional laser scanner; performing point cloud completion on the three-dimensional point cloud data; obtaining road surface settlement data based on the completed three-dimensional point cloud data; storing the three-dimensional point cloud data and road surface settlement data obtained from each detection; and predicting the development trend of a single settlement and the settlement distribution trend based on the stored historical detection data. The present invention obtains a three-dimensional point cloud map of the highway through three-dimensional laser scanning, uses point cloud completion technology to complete the missing point clouds in the point cloud map, uses a three-dimensional convolutional neural network to accurately mark the settlement position, and displays the development and changes of a single settlement point by constructing a coordinate system and the distribution characteristics of the point cloud, while achieving the prediction of the expansion of a single settlement and the settlement distribution trend.

Description

Technical Field

[0001] The present invention belongs to the field of road detection, and in particular relates to a method for quickly detecting large deformation and settlement of a highway surface based on three-dimensional laser point clouds. Background Art

[0002] During highway construction and renovation and expansion projects, the roadbed undergoes varying degrees of settlement during the construction phase due to the combined effects of increasing fill and compaction loads from construction machinery, as well as factors such as rainfall, groundwater, and temperature. During the operational phase, the roadbed also experiences a certain degree of settlement due to the combined effects of concentrated loads from the roadbed itself and other moving loads such as vehicles. Excessive settlement and settlement rate can lead to roadbed instability or even destruction, endangering the project's construction and operational safety, posing a serious threat to socioeconomic development and the safety of people's lives and property. Therefore, to ensure both construction safety and operational safety during the operational phase, it is necessary to monitor road surface settlement.

[0003] The existing technology for highway detection methods mainly uses the satellite positioning system (GPS) and the layout of the leveling network, as well as the establishment of a groundwater dynamic monitoring network. The settlement detection is carried out by investigating and analyzing the groundwater extraction volume, the groundwater level depth and elevation, and the groundwater quality changes. The above method does not select a certain number of settlement detection points during the detection, the detection section range is small, the detection data is not reliable enough, and the location, depth and width of the road surface settlement area cannot be fully detected. It is also not convenient to back up and store the data after detection, and data loss may occur. At the same time, it cannot effectively reflect the development and changes of settlement and cannot predict the distribution trend of settlement. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for rapid detection of large deformation and settlement of highway surface based on three-dimensional laser point cloud, so as to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above objectives, the present invention provides a method for rapid detection of large deformation and settlement of highway surface based on three-dimensional laser point cloud, comprising the following steps:

[0006] Scan the road surface to be inspected using a 3D laser scanner to obtain a 3D point cloud image of the road surface;

[0007] Due to the single or limited viewing angles of the 3D laser scanner and the unavoidable factors of the road surface, the scanned objects may be incomplete. Therefore, it is necessary to identify the missing points or unevenly distributed points in the 3D point cloud data obtained after scanning, and perform point cloud completion processing.

[0008] The specific steps of point cloud completion include: dividing the point cloud blocks in the three-dimensional point cloud data according to the components of the highway (the point cloud block is a collection of point clouds with different distribution types, for example, the point cloud distribution type of traffic signs on the highway is completely different from the point cloud distribution type of the road surface), obtaining multiple groups of point cloud blocks with different distribution types, and counting the point cloud distribution characteristics in each group of point cloud blocks; determining the point cloud block types included in the missing part of the point cloud, and completing the incomplete part of the point cloud based on the point cloud distribution characteristics corresponding to the point cloud block type.

[0009] After the point cloud is completed and the 3D point cloud data is obtained, the locations where the road subsidence occurs in the 3D point cloud data are marked:

[0010] Obtain the distribution characteristics of the road surface point cloud and the settlement point cloud according to the divided point cloud blocks; build a settlement recognition model based on the three-dimensional convolutional neural network;

[0011] Positive samples are samples that belong to a certain category, and negative samples are samples that do not belong to a certain category. Using positive and negative samples can effectively identify settlement points on the highway surface. Therefore, the point cloud distribution features corresponding to the settlement are used as positive samples, and the point cloud distribution features corresponding to the road surface are used as negative samples to input into the settlement recognition model for training.

[0012] Input the 3D point cloud data to be detected into the trained settlement recognition model to identify and output the settlement location of the road surface;

[0013] After obtaining the settlement position of the road surface, a spatial rectangular coordinate system is constructed with the center point of the highway as the origin, and the settlement position is expressed and marked in the form of coordinates.

[0014] The three-dimensional point cloud data and road surface settlement data obtained from each test are stored;

[0015] Based on the historically stored three-dimensional point cloud data and the road surface settlement data, the development trend of a single settlement is predicted, and the settlement distribution trend is predicted.

[0016] Optionally, the road surface settlement data includes settlement distribution quantity and settlement distribution; and the process of obtaining the road surface settlement data based on the pre-processed three-dimensional point cloud data includes:

[0017] Obtain the distribution location of road surface settlement based on the pre-processed three-dimensional point cloud;

[0018] Based on the road surface settlement distribution position, the settlement distribution quantity is determined and the highway settlement distribution situation is obtained.

[0019] Optionally, the process of obtaining the road surface settlement distribution position based on the pre-processed three-dimensional point cloud data includes:

[0020] Obtain the distribution characteristics of the road surface point cloud and the settlement point cloud according to the divided point cloud blocks;

[0021] Constructing a settlement recognition model based on a three-dimensional convolutional neural network;

[0022] The settlement point cloud distribution features are used as positive samples, and the road surface point cloud distribution features are used as negative samples to input into the settlement recognition model for training;

[0023] Inputting the three-dimensional point cloud data into the trained settlement recognition model to identify and output the settlement position of the road surface;

[0024] A spatial rectangular coordinate system is constructed with the center point of the highway as the origin, and the coordinates of the settlement position are obtained and marked.

[0025] Optionally, the process of counting the settlement distribution quantity based on the road surface settlement distribution position and obtaining the highway settlement distribution includes:

[0026] Dividing the spatial rectangular coordinate system in the three-dimensional point cloud data into quadrants, and labeling each divided quadrant;

[0027] The settlement distribution number in each quadrant is counted, and the highway settlement distribution situation is obtained based on each group of numbered quadrants and the corresponding settlement distribution number.

[0028] Optionally, the process of predicting the development trend of a single settlement based on the historically stored three-dimensional point cloud data and the road surface settlement data includes:

[0029] Overlapping the spatial rectangular coordinate systems in each set of the historically stored three-dimensional point cloud data based on the origin;

[0030] Based on the settlement point cloud distribution characteristics in each set of overlapping three-dimensional point cloud data, the settlement expansion point cloud coordinates are obtained;

[0031] Build a point cloud expansion prediction model based on 3D convolutional neural network;

[0032] Inputting each set of settlement electrical cloud distribution features corresponding to the settlement enlarged point cloud coordinates into the point cloud enlargement prediction model, and outputting the predicted distribution features of the settlement;

[0033] Calculating a settlement prediction value based on the predicted distribution characteristics, and setting a threshold value according to the settlement point cloud distribution characteristics when the road is damaged;

[0034] Repeatedly input the predicted distribution features into the point cloud expansion prediction model to perform prediction and calculate the settlement prediction value. When the settlement prediction value reaches a set threshold, the prediction is stopped and the number of repeated predictions is output.

[0035] Optionally, the process of predicting the settlement distribution trend includes:

[0036] Counting the highway settlement distribution in the historically stored road surface settlement data, and constructing a distribution trend prediction curve based on the settlement distribution quantity of each group of numbered quadrants in the highway settlement distribution;

[0037] The settlement distribution trend is predicted based on the distribution trend prediction curve.

[0038] The technical effects of the present invention are:

[0039] The present invention proposes a highway surface settlement detection method based on three-dimensional laser point cloud. The three-dimensional point cloud map of the highway is obtained through three-dimensional laser scanning, and the missing point clouds in the point cloud map are completed using point cloud completion technology. The settlement position is accurately marked using a three-dimensional convolutional neural network. The development and changes of a single settlement point are displayed by constructing a coordinate system and the distribution characteristics of the point cloud. At the same time, the expansion of a single settlement and the settlement distribution trend are predicted. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0041] Figure 1 This is a flow chart of a method for rapid detection of large deformation and settlement of a highway surface based on three-dimensional laser point cloud in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0043] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0044] Example 1

[0045] like Figure 1 As shown, this embodiment provides a method for rapid detection of large deformation and settlement of a highway surface based on a three-dimensional laser point cloud, comprising the following steps:

[0046] Scan the road surface to be inspected using a 3D laser scanner to obtain a 3D point cloud image of the road surface;

[0047] Due to the single or limited viewing angles of the 3D laser scanner and the unavoidable factors of the road surface, the scanned objects may be incomplete. Therefore, it is necessary to identify the missing points or unevenly distributed points in the 3D point cloud data obtained after scanning, and perform point cloud completion processing.

[0048] The specific steps of point cloud completion include: dividing the point cloud blocks in the three-dimensional point cloud data according to the components of the highway (the point cloud block is a collection of point clouds with different distribution types, for example, the point cloud distribution type of traffic signs on the highway is completely different from the point cloud distribution type of the road surface), obtaining multiple groups of point cloud blocks with different distribution types, and counting the point cloud distribution characteristics in each group of point cloud blocks; determining the point cloud block types included in the missing part of the point cloud, and completing the incomplete part of the point cloud based on the point cloud distribution characteristics corresponding to the point cloud block type.

[0049] After the point cloud is completed and the 3D point cloud data is obtained, the locations where the road subsidence occurs in the 3D point cloud data are marked:

[0050] Obtain the distribution characteristics of the road surface point cloud and the settlement point cloud according to the divided point cloud blocks; build a settlement recognition model based on the three-dimensional convolutional neural network;

[0051] Positive samples are samples that belong to a certain category, and negative samples are samples that do not belong to a certain category. Using positive and negative samples can effectively identify settlement points on the highway surface. Therefore, the point cloud distribution features corresponding to the settlement are used as positive samples, and the point cloud distribution features corresponding to the road surface are used as negative samples to input into the settlement recognition model for training.

[0052] Input the 3D point cloud data to be detected into the trained settlement recognition model to identify and output the settlement location of the road surface;

[0053] After obtaining the settlement position of the road surface, a spatial rectangular coordinate system is constructed with the center point of the highway as the origin, and the settlement position is expressed and marked in the form of coordinates.

[0054] After obtaining the settlement position, the spatial rectangular coordinate system in the three-dimensional point cloud data is divided into quadrants, with the first quadrant marked as 1, the second quadrant marked as 2, the third quadrant marked as 3, and so on, until the eighth quadrant;

[0055] Count the number of settlements in each quadrant. By observing the corresponding number of settlements in each numbered quadrant, the distribution of highway settlements in each quadrant can be obtained.

[0056] After obtaining the settlement location and distribution, the preliminary detection process is completed. At this time, the three-dimensional point cloud data and settlement location obtained by the detection are stored;

[0057] The present invention can predict the future development trend of settlement based on historical detection data, specifically including:

[0058] Based on the historically stored 3D point cloud data, settlement quantity, and settlement distribution trend, the development trend of a single settlement is predicted, and the settlement distribution trend is also predicted:

[0059] The spatial rectangular coordinate systems in each set of historically stored three-dimensional point cloud data are overlapped based on the origin. For example, when the development of a single settlement point needs to be predicted on a monthly basis, the spatial rectangular coordinate systems obtained by detection every other month are overlapped.

[0060] By observing the distribution characteristics of the settlement point cloud in each set of overlapping 3D point cloud data, the developed and expanded settlement point can be obtained, and the point cloud coordinates of the point can be obtained;

[0061] Since 3D neural networks can effectively display the distribution of 3D objects, a 3D convolutional neural network is used to construct a point cloud expansion prediction model.

[0062] Input each set of settlement electric cloud distribution features corresponding to the expanded point cloud coordinates of the settlement point into the point cloud expansion prediction model for prediction, and output the predicted distribution features of the settlement;

[0063] After obtaining the predicted electric cloud distribution characteristics, the distribution is converted into a digital form through calculation to obtain the predicted value;

[0064] When the distribution characteristics reach a certain level, road damage will occur. Therefore, a threshold is set according to the distribution characteristics of the settlement point cloud when the road is damaged.

[0065] The predicted distribution features obtained from each output are repeatedly input into the point cloud expansion prediction model for prediction and calculation of the settlement prediction value. When the settlement prediction value reaches the set threshold, the prediction is stopped and the road damage is judged at this time. By obtaining the number of repeated inputs, it can be known whether the settlement point is damaged at some time in the future.

[0066] The process of predicting the settlement distribution trend includes:

[0067] Statistically analyzing the highway settlement distribution in the historically stored road surface settlement data, and constructing a distribution trend prediction curve based on the settlement distribution quantity of each group of numbered quadrants in the highway settlement distribution;

[0068] Based on the distribution trend prediction curve, the future settlement development distribution trend of the quadrant is predicted.

[0069] The above description is merely a preferred embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for rapid detection of large deformation and settlement of highway surface based on three-dimensional laser point cloud, characterized in that: The following steps are involved: Acquire 3D point cloud data of highway surface based on 3D laser scanner; Preprocessing the three-dimensional point cloud data; Obtain road surface settlement data based on pre-processed 3D point cloud data; Storing the three-dimensional point cloud data and the road surface settlement data obtained in each detection; Predicting the development trend of a single settlement based on the historically stored three-dimensional point cloud data and the road surface settlement data, and predicting the settlement distribution trend; The process of preprocessing the three-dimensional point cloud data includes: Identifying incomplete point cloud parts in the three-dimensional point cloud data, and completing the incomplete point cloud parts; The incomplete part of the point cloud is the part where the point cloud is missing or unevenly distributed; The process of completing the incomplete part of the point cloud includes: Dividing the point cloud blocks in the three-dimensional point cloud data according to components of the highway, obtaining multiple groups of point cloud blocks with different distribution types, and counting the point cloud distribution features in each group of point cloud blocks; Determining the point cloud block type included in the missing portion of the point cloud, and completing the missing portion of the point cloud based on point cloud distribution characteristics corresponding to the point cloud block type; The road surface settlement data includes the settlement distribution quantity and settlement distribution. The process of obtaining the road surface settlement data based on the pre-processed three-dimensional point cloud data includes: Obtain the distribution location of road surface settlement based on the pre-processed three-dimensional point cloud; Determine the settlement distribution quantity based on the road surface settlement distribution position and obtain the highway settlement distribution situation; The process of obtaining the distribution location of road surface settlement based on pre-processed 3D point cloud data includes: Obtain the distribution characteristics of the road surface point cloud and the settlement point cloud according to the divided point cloud blocks; Constructing a settlement recognition model based on a three-dimensional convolutional neural network; The settlement point cloud distribution features are used as positive samples, and the road surface point cloud distribution features are used as negative samples to input into the settlement recognition model for training; Inputting the three-dimensional point cloud data into the trained settlement recognition model to identify and output the settlement position of the road surface; Constructing a spatial rectangular coordinate system with the center point of the highway as the origin, and obtaining and marking the coordinates of the settlement position; The process of counting the settlement distribution quantity based on the road surface settlement distribution position and obtaining the highway settlement distribution situation includes: Dividing the spatial rectangular coordinate system in the three-dimensional point cloud data into quadrants, and labeling each divided quadrant; Count the number of settlement distributions in each quadrant, and obtain the highway settlement distribution based on each group of numbered quadrants and the corresponding settlement distribution numbers; The process of predicting the development trend of a single settlement based on the historically stored three-dimensional point cloud data and the road surface settlement data includes: Overlapping the spatial rectangular coordinate systems in each set of the historically stored three-dimensional point cloud data based on the origin; Based on the settlement point cloud distribution characteristics in each set of overlapping three-dimensional point cloud data, the settlement expansion point cloud coordinates are obtained; Build a point cloud expansion prediction model based on 3D convolutional neural network; Inputting each set of settlement electrical cloud distribution features corresponding to the settlement enlarged point cloud coordinates into the point cloud enlargement prediction model, and outputting the predicted distribution features of the settlement; Calculating a settlement prediction value based on the predicted distribution characteristics, and setting a threshold value according to the settlement point cloud distribution characteristics when the road is damaged; Repeatedly inputting the predicted distribution features into the point cloud expansion prediction model to perform prediction and calculate the settlement prediction value, stopping the prediction when the settlement prediction value reaches a set threshold, and outputting the number of repeated predictions; The damage time of the corresponding settlement point is predicted based on the monitoring cycle of a single settlement point and the number of repeated inputs.

2. The method for rapid detection of large deformation and settlement of highway surface based on three-dimensional laser point cloud according to claim 1 is characterized in that: The process of predicting the settlement distribution trend includes: Counting the highway settlement distribution in the historically stored road surface settlement data, and constructing a distribution trend prediction curve based on the settlement distribution quantity of each group of numbered quadrants in the highway settlement distribution; The settlement distribution trend is predicted based on the distribution trend prediction curve.

Images