Ground data processing method and system based on three-dimensional point cloud

Abstract

The application provides a kind of based on three-dimensional point cloud ground data processing method and system, the application relates to ground data processing technical field, the method includes determining the multiple division regions of ground based on three-dimensional point cloud data before ground excavation;Determine the cross-sectional shooting video of each division region based on the multiple division regions of ground and cross-sectional shooting video in the process of ground excavation;Generate the simulation three-dimensional ground image of each division region, the excavation information of each division region;Determine the earthwork correction coefficient of each division region based on the simulation three-dimensional ground image of each division region, the excavation information of each division region;Determine the corrected all earthwork result data based on the estimated earthwork data of each division region and the correction coefficient of each division region, the method can accurately determine the earthwork data of ground excavation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of foundation data processing, and particularly relates to a foundation data processing method and system based on three-dimensional point cloud. BACKGROUND

[0002] In the field of foundation construction, the accurate accounting of earthwork data directly affects the engineering cost and construction safety. Traditional foundation data processing relies on single-point measurement or empirical formula simulation, which has significant technical bottlenecks. Single-point measurement can only obtain discrete data and cannot reflect the overall physical characteristics of the foundation, such as soil moisture distribution and geological structure continuity, resulting in large errors in earthwork estimation. The finite element simulation based on experience requires the pre-setting of a large number of parameters, which has poor adaptability to complex geological conditions such as cavities and collapsible soil layers, and is prone to deviation due to long calculation period. In addition, dynamic monitoring information generated during construction, such as cross-sectional soil state changes, is not effectively integrated and utilized, making it difficult to correct the earthwork calculation results in real time according to the actual geological conditions, and often resulting in significant deviation between actual compaction and estimated results due to soil property differences.

[0003] Therefore, how to accurately determine the earthwork data of foundation excavation is a problem to be solved at present. SUMMARY

[0004] The technical problem solved by the present application is how to accurately determine the earthwork data of foundation excavation.

[0005] According to a first aspect, the present application provides a foundation data processing method based on three-dimensional point cloud, comprising: acquiring three-dimensional point cloud data before foundation excavation, three-dimensional point cloud data after foundation excavation, and cross-section shooting video during foundation excavation; determining a plurality of divided regions of the foundation based on the three-dimensional point cloud data before foundation excavation; determining the estimated earthwork data of each divided region based on the three-dimensional point cloud data before foundation excavation, the three-dimensional point cloud data after foundation excavation, and the plurality of divided regions of the foundation; determining the cross-section shooting video of each divided region based on the plurality of divided regions of the foundation and the cross-section shooting video during foundation excavation; generating the simulated three-dimensional foundation image of each divided region and the excavation information of each divided region based on the cross-section shooting video of each divided region; determining the earthwork correction coefficient of each divided region based on the simulated three-dimensional foundation image of each divided region and the excavation information of each divided region; and determining the corrected overall earthwork result data based on the estimated earthwork data of each divided region and the correction coefficient of each divided region.

[0006] In a possible implementation, the determining the multiple partition regions of the foundation based on the three-dimensional point cloud data before the foundation excavation comprises: determining multiple salient point cloud sets based on the three-dimensional point cloud data before the foundation excavation; determining multiple high saliency partition points, multiple medium saliency partition points, and multiple low saliency partition points based on the multiple salient point cloud sets; determining multiple sets of partition schemes based on the multiple high saliency partition points, the multiple medium saliency partition points, and the multiple low saliency partition points; and determining the multiple partition regions of the foundation based on the multiple sets of partition schemes using a neural network model.

[0007] In a possible implementation, the foundation excavation is a straight-line foundation excavation of an airplane runway.

[0008] In a possible implementation, the determining the multiple partition regions of the foundation based on the multiple sets of partition schemes using the neural network model comprises: constructing a scheme graph based on the multiple sets of partition schemes, the scheme graph comprising multiple partition scheme nodes and edges between the nodes, a node feature of a partition scheme node comprising three-dimensional point cloud data of each partition region corresponding to the partition scheme, and an edge between the partition scheme nodes being a difference degree of the partition schemes; processing the scheme graph based on a graph neural network to obtain a target partition scheme; and partitioning the foundation based on the target partition scheme to obtain the multiple partition regions of the foundation.

[0009] According to a second aspect, the present application provides a foundation data processing system based on three-dimensional point cloud, comprising: a data acquisition module configured to acquire three-dimensional point cloud data before foundation excavation, three-dimensional point cloud data after foundation excavation, and cross-section shooting video during foundation excavation; a region partitioning module configured to determine multiple partition regions of the foundation based on the three-dimensional point cloud data before foundation excavation; a earthwork estimation module configured to determine estimated earthwork data of each partition region based on the three-dimensional point cloud data before foundation excavation, the three-dimensional point cloud data after foundation excavation, and the multiple partition regions of the foundation; a video processing module configured to determine cross-section shooting video of each partition region based on the multiple partition regions of the foundation and the cross-section shooting video during foundation excavation; an image generation module configured to generate simulated three-dimensional foundation images of each partition region and excavation information of each partition region based on the cross-section shooting video of each partition region; a correction coefficient determination module configured to determine earthwork correction coefficients of each partition region based on the simulated three-dimensional foundation images of each partition region and the excavation information of each partition region; and an earthwork correction module configured to determine corrected overall earthwork result data based on the estimated earthwork data of each partition region and the correction coefficients of each partition region.

[0010] In a possible implementation, the region division module is further configured to: determine a plurality of salient point cloud sets based on the three-dimensional point cloud data before excavation of the foundation; determine a plurality of high saliency division points, a plurality of medium saliency division points, and a plurality of low saliency division points based on the plurality of salient point cloud sets; determine a plurality of division schemes based on the plurality of high saliency division points, the plurality of medium saliency division points, and the plurality of low saliency division points; and determine a plurality of division regions of the foundation based on the plurality of division schemes using a neural network model.

[0011] In a possible implementation, the excavation of the foundation is a straight-line excavation of an airplane runway.

[0012] In a possible implementation, the region division module is further configured to: construct a scheme graph based on the plurality of division schemes, the scheme graph including a plurality of division scheme nodes and edges between the nodes, a node feature of a division scheme node including three-dimensional point cloud data of each division region corresponding to the division scheme, and an edge between the division scheme nodes being a difference degree of the division schemes; process the scheme graph based on a graph neural network to obtain a target division scheme; and divide the foundation based on the target division scheme to obtain the plurality of division regions of the foundation.

[0013] According to a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method comprising: obtaining three-dimensional point cloud data before excavation of the foundation, three-dimensional point cloud data after excavation of the foundation, and a cross-section shooting video during the excavation of the foundation; determining a plurality of division regions of the foundation based on the three-dimensional point cloud data before excavation of the foundation; determining estimated earthwork data of each division region based on the three-dimensional point cloud data before excavation of the foundation, the three-dimensional point cloud data after excavation of the foundation, and the plurality of division regions of the foundation; determining a cross-section shooting video of each division region based on the plurality of division regions of the foundation and the cross-section shooting video during the excavation of the foundation; generating a simulated three-dimensional foundation image of each division region and excavation information of each division region based on the cross-section shooting video of each division region; determining an earthwork correction coefficient of each division region based on the simulated three-dimensional foundation image of each division region and the excavation information of each division region; and determining corrected overall earthwork result data based on the estimated earthwork data of each division region and the correction coefficient of each division region.

[0014] According to a fourth aspect, the embodiment provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the ground data processing method based on three-dimensional point cloud provided in the foregoing aspects, and the method comprises: obtaining three-dimensional point cloud data before ground excavation, three-dimensional point cloud data after ground excavation, and cross-section shooting video in the ground excavation process; determining a plurality of divided regions of the ground based on the three-dimensional point cloud data before ground excavation; determining estimated earthwork data of each divided region based on the three-dimensional point cloud data before ground excavation, the three-dimensional point cloud data after ground excavation, and the plurality of divided regions of the ground; determining cross-section shooting video of each divided region based on the plurality of divided regions of the ground and the cross-section shooting video in the ground excavation process; generating a simulated three-dimensional ground image of each divided region and excavation information of each divided region based on the cross-section shooting video of each divided region; determining an earthwork correction coefficient of each divided region based on the simulated three-dimensional ground image of each divided region and the excavation information of each divided region; and determining corrected overall earthwork result data based on the estimated earthwork data of each divided region and the correction coefficient of each divided region.

[0015] The application provides a ground data processing method and system based on three-dimensional point cloud, which comprises the following steps: obtaining three-dimensional point cloud data before ground excavation, three-dimensional point cloud data after ground excavation, and cross-section shooting video in the ground excavation process; determining a plurality of divided regions of the ground based on the three-dimensional point cloud data before ground excavation; determining estimated earthwork data of each divided region based on the three-dimensional point cloud data before ground excavation, the three-dimensional point cloud data after ground excavation, and the plurality of divided regions of the ground; determining cross-section shooting video of each divided region based on the plurality of divided regions of the ground and the cross-section shooting video in the ground excavation process; generating a simulated three-dimensional ground image of each divided region and excavation information of each divided region based on the cross-section shooting video of each divided region; determining an earthwork correction coefficient of each divided region based on the simulated three-dimensional ground image of each divided region and the excavation information of each divided region; and determining corrected overall earthwork result data based on the estimated earthwork data of each divided region and the correction coefficient of each divided region, so that the earthwork data of the ground excavation can be accurately determined. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A flowchart of a ground data processing method based on three-dimensional point cloud is provided for the embodiment of the application.

[0017] Figure 2 A schematic diagram of the effect before three-dimensional point cloud data denoising is provided for the embodiment of the application.

[0018] Figure 3A three-dimensional point cloud data denoising effect schematic diagram provided for an embodiment of the present application;

[0019] Figure 4 A colorless three-dimensional point cloud data schematic diagram provided for an embodiment of the present application before excavation of a foundation;

[0020] Figure 5 A color three-dimensional point cloud data schematic diagram provided for an embodiment of the present application before excavation of a foundation;

[0021] Figure 6 A flowchart schematic diagram of determining a plurality of division regions of a foundation based on the three-dimensional point cloud data before excavation of the foundation provided for an embodiment of the present application;

[0022] Figure 7 A flowchart schematic diagram of determining a plurality of division regions of a foundation based on the plurality of sets of division schemes using a neural network model provided for an embodiment of the present application;

[0023] Figure 8 A scheme atlas schematic diagram constructed provided for an embodiment of the present application;

[0024] Figure 9 A three-dimensional point cloud-based foundation data processing system schematic diagram provided for an embodiment of the present application; DETAILED DESCRIPTION

[0025] The present application will be further described in details through specific embodiments combined with the drawings. In different embodiments, similar elements are associated with similar element labels. In the following embodiments, many details are described in order to make the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted in different cases, or can be replaced by other elements, materials, methods. In some cases, some operations related to the present application are not shown or described in the specification in order to avoid the core part of the present application being overwhelmed by too much description, and it is not necessary to describe these related operations in detail for those skilled in the art according to the description in the specification and general technical knowledge in the art.

[0026] In an embodiment of the present application, a three-dimensional point cloud-based foundation data processing method is provided, as shown in Figure 1 The three-dimensional point cloud-based foundation data processing method includes steps S1-S7:

[0027] Step S1, obtaining three-dimensional point cloud data before excavation of a foundation, three-dimensional point cloud data after excavation of the foundation, and a cross-section shooting video during the excavation of the foundation.

[0028] In some embodiments, the foundation excavation is a straight-line foundation excavation of an airplane runway.

[0029] The three-dimensional point cloud data is a data set of multiple discrete points in a three-dimensional space that can be used to describe the surface features of an object or terrain. Each point in the three-dimensional point cloud data contains X, Y, and Z coordinates.

[0030] In some embodiments, the three-dimensional point cloud data can be collected by scanning with a three-dimensional laser scanner, and noise reduction data processing is integrated during the collection process to ultimately obtain usable three-dimensional point cloud data.

[0031] In some embodiments, noise reduction processing can be performed on the three-dimensional point cloud data during the three-dimensional point cloud data collection process, Figure 2 A schematic diagram of the effect of three-dimensional point cloud data before noise reduction for an embodiment of the present application, Figure 3 A schematic diagram of the effect of three-dimensional point cloud data after noise reduction for an embodiment of the present application.

[0032] The three-dimensional point cloud data before foundation excavation is three-dimensional spatial discrete point data obtained by scanning and collecting the construction area before foundation excavation with a three-dimensional laser scanner, and integrating noise reduction data processing during the collection process. The three-dimensional point cloud data before foundation excavation is composed of a large number of discrete points, each point containing X, Y, and Z coordinates.

[0033] In some embodiments, the three-dimensional point cloud data before foundation excavation can include colorless three-dimensional point cloud data and colored three-dimensional point cloud data. Figure 4 A schematic diagram of colorless three-dimensional point cloud data before foundation excavation for an embodiment of the present application, Figure 5 A schematic diagram of colored three-dimensional point cloud data before foundation excavation for an embodiment of the present application.

[0034] Each point in the colored three-dimensional point cloud data before foundation excavation contains X, Y, and Z coordinates, as well as reflectance intensity and color attributes.

[0035] The three-dimensional point cloud data before foundation excavation can accurately record the coordinates of the terrain surface and the reflectance characteristics of the terrain surface material.

[0036] The three-dimensional point cloud data after foundation excavation is three-dimensional spatial discrete point data obtained by re-scanning and collecting the foundation construction area with a three-dimensional laser scanner after the foundation excavation work is completed, and integrating noise reduction data processing during the collection process.

[0037] In some embodiments, the three-dimensional point cloud data after foundation excavation includes colorless three-dimensional point cloud data and colored three-dimensional point cloud data.

[0038] The cross-section shooting video in the foundation excavation process is a video obtained by shooting the cross-section of the foundation excavation area in real time through a high-resolution video acquisition device in the process of foundation excavation construction.

[0039] The cross-section shooting video in the foundation excavation process can record the dynamic information of the soil layer structure change of the cross-section of the foundation in the excavation process, the excavation depth progress, the operation track of the construction equipment, and the soil physical state.

[0040] In step S2, a plurality of divided regions of the foundation are determined based on the three-dimensional point cloud data before the foundation excavation.

[0041] In some embodiments, Figure 6 A flowchart for determining a plurality of divided regions of the foundation based on the three-dimensional point cloud data before the foundation excavation is provided for the embodiments of the present application, and the determination of the plurality of divided regions of the foundation based on the three-dimensional point cloud data before the foundation excavation includes steps S61-S64:

[0042] In step S61, a plurality of salient point cloud sets are determined based on the three-dimensional point cloud data before the foundation excavation.

[0043] In some embodiments, the plurality of salient point cloud sets can be determined based on the three-dimensional point cloud data before the foundation excavation through a point cloud analysis model. The point cloud analysis model is a convolutional neural network model, the input of the point cloud analysis model is the three-dimensional point cloud data before the foundation excavation, and the output of the point cloud analysis model is the plurality of salient point cloud sets.

[0044] The convolutional neural network model includes a convolutional neural network (CNN), which is a kind of feedforward neural network containing convolutional computation and having a deep structure. The convolutional neural network can extract features from input data through a convolutional layer, and then reduce network parameters by using local connection and weight sharing mechanism. The convolutional neural network can effectively process data with grid topology.

[0045] The plurality of salient point cloud sets are a plurality of discrete point sets with significant topographic or material features extracted from the three-dimensional point cloud data before the foundation excavation by the point cloud analysis model. Each point in the salient point cloud set contains X, Y, Z coordinates and attributes such as reflectance intensity and color. The points in the salient point cloud set can be located at places where the topographic structure changes significantly, such as slope mutations, topographic inflection points, or places where the material reflection characteristics are unique, such as soil humidity differences and roughness change areas. For example, in the foundation construction area, the vertex of the slope with large topographic undulations, the junction of different soil types, or the point with significantly abnormal reflectance intensity due to high soil humidity can all constitute a salient point cloud set.

[0046] By extracting a plurality of representative salient point clouds, the physical feature differences of different regions of the foundation can be more accurately identified.

[0047] The local receptive field characteristics of the convolutional layers of the convolutional neural network can extract local geometric features in the three-dimensional point cloud data before excavation of the foundation, such as the neighborhood distribution of points, curvature changes, etc. Then, through multi-layer convolution and pooling operations, points with significant geometric features or spatial distribution characteristics can be filtered from the original point cloud data. Then, based on a feature similarity threshold, the model can map points with significant features in a high-dimensional feature space to a three-dimensional space, thereby gathering into a plurality of salient point clouds. Each point cloud contains discrete points at topographic structure mutation points or abnormal material reflection feature points, such as slope vertexes, soil type junction points, etc.

[0048] In some embodiments, the point cloud analysis model includes a feature preliminary screening layer, a salient point quantization layer, and a point cloud set generation layer. The input of the feature preliminary screening layer is the three-dimensional point cloud data before excavation of the foundation, and the output of the feature preliminary screening layer is a topographic mutation point set and a color and reflection intensity abnormal point set. The input of the salient point quantization layer is the topographic mutation point set and the color and reflection intensity abnormal point set, and the output of the salient point quantization layer is a plurality of preliminary selected salient points, a feature difference degree of each point with neighboring points, a neighborhood point cloud density of each point, a position importance, and a surrounding humidity gradient change. The input of the point cloud set generation layer is the plurality of preliminary selected salient points, the feature difference degree of each point with neighboring points, the neighborhood point cloud density of each point, the position importance, and the surrounding humidity gradient change, and the output of the point cloud set generation layer is a plurality of salient point clouds.

[0049] The topographic mutation point set is a set of points with significant changes in topographic structure extracted from the three-dimensional point cloud data. The topographic mutation points include topographic slope mutation points and elevation abnormal points, such as steep slope inflection points, protruding rocks, or concave pits. The topographic mutation point set can reflect the undulating morphology and vertical abnormalities of the foundation surface.

[0050] The color and reflection intensity abnormal point set is a set of points in the three-dimensional point cloud data whose color attribute or laser reflection intensity significantly deviates from the surrounding area. The color abnormal points correspond to material color mutation regions in the foundation, such as wet areas, and the reflection intensity abnormal points correspond to soil humidity abnormal regions, such as low reflection intensity of high humidity points.

[0051] The plurality of preliminary selected salient points is a set of candidate points of salient points selected by preliminary screening.

[0052] The feature difference degree of each point with neighboring points is the difference degree of each preliminary selected salient point with neighboring preliminary selected salient points in terms of topographic features and reflection attributes.

[0053] The neighborhood point cloud density of each point is the number of point clouds containing other preliminary selected salient points per unit area in the local space centered on a single point.

[0054] The position importance can represent the representativeness of the preliminary salient point to the surrounding area features.

[0055] The surrounding humidity gradient change is the humidity change rate around the preliminary salient point, and the surrounding humidity gradient change can reflect the humidity transition speed per unit distance. The surrounding humidity gradient change can be used to identify the humidity transition zone or the abrupt boundary.

[0056] Different layers are responsible for different levels of information abstraction and processing, wherein the feature preliminary screening layer is responsible for extracting the basic feature points of terrain mutation and color and reflectance intensity anomaly from the three-dimensional point cloud data, the salient point quantification layer can calculate the preliminary salient points according to the basic feature points, and the feature difference degree, point cloud density, position importance and humidity gradient change of these points, and the point cloud set generation layer is responsible for determining the final salient point cloud set according to the quantification indexes. Through such hierarchical processing, the point cloud data with large quantity and complexity can be processed more effectively. By constructing multiple layers, the modularization degree of the system can be improved, and the accuracy and efficiency of the feature recognition of the three-dimensional point cloud data can be improved.

[0057] In step S62, a plurality of high salient division points, a plurality of medium salient division points, and a plurality of low salient division points are determined based on the plurality of salient point cloud sets.

[0058] In some embodiments, a division model can be used to determine a plurality of high salient division points, a plurality of medium salient division points, and a plurality of low salient division points based on the plurality of salient point cloud sets, the division model is a Transformer model, the input of the division model is the plurality of salient point cloud sets, and the output of the division model is the plurality of high salient division points, the plurality of medium salient division points, and the plurality of low salient division points.

[0059] The Transformer model is a deep learning model based on self-attention mechanism (Self-Attention), which can capture long-distance semantic association by calculating the global dependency of each element in the input sequence, and can realize feature extraction and semantic understanding of sequence data through multi-head attention mechanism and feedforward neural network.

[0060] The plurality of high salient division points is a discrete point set extracted from the plurality of salient point cloud sets by the division model, which has a decisive influence on the feature difference of the ground surface soil humidity. The reflectance intensity attribute of the region where the high salient division point is located can be significantly abnormal due to extremely high or extremely low soil humidity.

[0061] For example, in the construction area of the aircraft runway foundation, the local saturated soil area caused by water penetration, or the point with extremely low humidity and abnormal decrease in reflectivity caused by solar evaporation, are all high-remarkable division points.

[0062] As a humidity mutation point, the high-remarkable division point can be used to accurately define the boundary of different humidity regions.

[0063] The multiple medium-remarkable division points are a set of discrete points with medium influence on the soil humidity characteristics of the foundation surface by the division model. The reflectivity attribute change of the medium-remarkable division point reflects the gradient transition characteristics of soil humidity. The medium-remarkable division points are distributed in the humidity gradual change region between the high-remarkable division points, such as the transition zone from humid soil to semi-arid soil, or the continuous change segment where the soil humidity slowly rises from 60% to 80%.

[0064] For example, in the construction area of the aircraft runway foundation, the local saturated soil area caused by water penetration, or the point with extremely low humidity and abnormal decrease in reflectivity caused by solar evaporation, are all high-remarkable division points.

[0065] The multiple low-remarkable division points are a set of discrete points with low influence on the soil humidity characteristics of the foundation surface by the division model. The low-remarkable division points are distributed in the region with relatively uniform soil humidity, and the reflectivity attribute values of these points are close, indicating that the soil humidity in the region is consistent, such as the homogeneous clay layer with humidity maintained at 50%±5%, and the sand soil concentration area with humidity fluctuation less than 3%.

[0066] The multiple remarkable point clouds contain key points at the mutation of topographic structure, such as slope inflection points, topographic boundary points, and material reflection characteristic abnormal points such as soil humidity difference points, where the coordinates and attribute data can accurately reflect the physical characteristic differences of different regions of the foundation. For example, the point cloud set of the region with dramatic topographic relief can represent the complexity of geological structure, and the point cloud set with abnormal reflectivity corresponds to the mutation zone of soil humidity. Through multiple remarkable point cloud sets, the key feature points that play a decisive role in regional division can be directly located, and high-remarkable division points, medium-remarkable division points, and low-remarkable division points can be selected according to the degree of feature remarkable.

[0067] The transformer model can calculate the weight of each point in the global point cloud through the multi-head attention mechanism, and the statistical correlation between the intensity of reflection and the soil moisture is converted into the feature weight. For example, the abnormal value of the reflection intensity of the moisture mutation point will obtain a high weight in the self-attention calculation of the transformer model, so as to be identified as a high-significant division point. The points in the moisture gradual change area have a lower weight than the high-significant division point because the change of the reflection intensity is gentle, so the points can be classified as medium-significant division points. The points in the area with relatively uniform moisture have the lowest weight, and are classified as low-significant division points.

[0068] In step S63, a plurality of sets of division schemes are determined based on the plurality of high-significant division points, the plurality of medium-significant division points, and the plurality of low-significant division points.

[0069] In some embodiments, a plurality of sets of division schemes can be determined based on the plurality of high-significant division points, the plurality of medium-significant division points, and the plurality of low-significant division points using a scheme output model. The scheme output model is a deep neural network model, the input of the scheme output model is the plurality of high-significant division points, the plurality of medium-significant division points, and the plurality of low-significant division points, and the output of the scheme output model is a plurality of sets of division schemes.

[0070] The deep neural network model includes a deep neural network, which is a neural network model containing multiple layers of structures. The deep neural network can form a complex computing network by combining a large number of neurons, thereby being able to automatically learn high-level abstract features from input data. The deep neural network includes an input layer, a hidden layer, and an output layer, wherein the hidden layer can include multiple non-linear transformation layers such as fully connected layers, convolutional layers, etc., and the network parameters can be optimized through a back propagation algorithm.

[0071] The plurality of sets of division schemes are generated by the division model and are used to define the boundaries of different soil moisture characteristic regions before foundation excavation. Each set of division schemes passes through all high-significant division points, part of medium-significant division points, and part of low-significant division points. The difference between different division schemes is reflected in the granularity and boundary direction of the moisture partition.

[0072] For example, one set of division schemes takes all high-significant division points as the boundary core, selects medium-significant division points with soil moisture increasing from 70% to 80% and low-significant division points with uniform moisture, and finally divides the area before foundation excavation into a high-moisture area (moisture ≥ 80%) and a medium-moisture area (moisture < 80%); another set of division schemes also contains all high-significant division points, but selects medium-significant division points with moisture of 60%, 70%, and 80% and low-significant division points in different uniform moisture areas, and further divides the area before foundation excavation into a saturated area, a high-moisture area, a medium-moisture area, and a dry area.

[0073] The high-significant division point as a boundary core can quickly determine a large range of humidity partition, and avoid unnecessary subdivision. In a complex construction area such as an aircraft runway, the high-significant point can mark the inflection point of the steep slope, the junction point of the landform, and other structural mutations, providing a basic framework for subsequent division.

[0074] In the gradual change region between the high-significant division points, the medium-significant division point can refine the boundary trend of the humidity partition, so that the division is more in line with the actual humidity distribution. By introducing the medium-significant point, the accuracy of the division can be ensured while avoiding the region redundancy caused by excessive subdivision. The low-significant division point is located in a region with relatively uniform soil humidity (such as a homogeneous clay layer or a sand soil concentration area), and the attribute values of the reflection intensity are close. In the uniform humidity region, the low-significant point can be used as a "noise filter" to avoid misjudging the uniform region as a humidity change region. The core purpose of dividing multiple high-significant division points, multiple medium-significant division points and multiple low-significant division points is to extract the key feature points of the ground surface, combine the influence degree of points with different significance levels on the soil humidity characteristics, and realize the fine division and efficient modeling of the ground region.

[0075] The deep neural network can automatically learn the spatial distribution pattern of multiple high-significant division points, multiple medium-significant division points and multiple low-significant division points, and the correlation with the soil humidity characteristics through multiple nonlinear transformations. The deep neural network can take the three-dimensional coordinates and reflection intensity attributes of these division points as input, extract the geometric relationship between the division points and the similarity of the humidity characteristics through the neuron combination of the hidden layer, so as to semantically fuse the humidity mutation determining features of the high-significant division points and the humidity gradual change auxiliary features of the medium and low-significant division points. For example, the model can learn that the high-significant division point must be used as the core anchor point of the humidity region boundary, and the connection mode of the medium and low-significant division points can form different boundary trends. Finally, the deep neural network can determine multiple sets of division schemes through different combinations of hidden layer weights.

[0076] Step S64, determining multiple division regions of the ground based on the multiple sets of division schemes using a neural network model.

[0077] In some embodiments, Figure 7 A flowchart for determining multiple division regions of the ground based on the multiple sets of division schemes using a neural network model is provided for the embodiments of the present application, and the determination of the multiple division regions of the ground based on the multiple sets of division schemes using a neural network model comprises steps S71-S73:

[0078] Step S71, constructing a scheme graph based on the multiple sets of division schemes, the scheme graph comprising multiple division scheme nodes and edges between the nodes, the node features of the division scheme nodes comprising the three-dimensional point cloud data of each division region corresponding to a division scheme, and the edges between the division scheme nodes being the difference degrees of the division schemes.

[0079] The scheme graph is a data structure composed of multiple division scheme nodes and edges between the nodes. The division scheme node contains node features, wherein the node features are three-dimensional point cloud data of each division region corresponding to a division scheme, and the node features can describe the spatial distribution and soil moisture characteristics of the division scheme. The edge is used to connect different division scheme nodes and can describe the association relationship between two division schemes, and the edge is the difference degree of the division schemes. Figure 8 A schematic diagram of a constructed scheme graph is provided for an embodiment of the present application. As shown in Figure 8 Figure 8 includes division scheme node A, division scheme node B, division scheme node C, and division scheme node D. The edge between two division scheme nodes is the difference degree of the division schemes contained by the two division scheme nodes.

[0080] The three-dimensional point cloud data of each division region corresponding to a division scheme is the three-dimensional point cloud data corresponding to each humidity division region in the division scheme. The three-dimensional point cloud data of each division region corresponding to a division scheme is extracted from the three-dimensional point cloud original data before ground excavation, and then the three-dimensional point cloud data corresponding to each division region is filtered out through spatial matching according to the region boundary coordinates of the division scheme.

[0081] The difference degree of the division scheme can be used to quantify the difference degree of different division schemes in terms of humidity partition boundary, soil moisture characteristic coverage, etc. In some embodiments, the difference degree between division schemes can be calculated by a deep neural network.

[0082] In step S72, the scheme graph is processed based on a graph neural network to obtain a target division scheme.

[0083] A graph neural network (GNN) is a deep learning model, and the graph neural network can be used to process graph data. The graph neural network can pass information between nodes in the graph through a message passing mechanism, thereby being able to learn the representation of the nodes in the graph data.

[0084] The target division scheme is an optimal division scheme of a region before ground excavation output by processing the scheme graph by the graph neural network. The target division scheme can control the humidity difference in each division region within a small fluctuation range, while avoiding excessive subdivision leading to redundant partition numbers.

[0085] ​The scheme graph can clearly show the relationship between each division scheme and the adjacent scheme, and this association information has a direct impact on screening the optimal ground division scheme. By taking the three-dimensional point cloud data of each division area corresponding to the division scheme node as the node feature, and taking the difference degree of the division scheme as the feature of the edge, the potential association information of each scheme can be fully utilized. This helps the graph neural network better understand the interactive influence of different schemes on the continuity of the humidity partition boundary and the uniformity of the single area, thereby improving the accuracy of screening the target division scheme.

[0086] The graph neural network can pass through the message passing mechanism to enable each division scheme node to combine the three-dimensional point cloud data features and difference degree information of adjacent nodes, and then analyze the optimization effect of different medium-low salient division point combinations on the actual humidity distribution of the partition boundary through graph convolution operation, such as whether to form a continuous transition humidity gradient area, and the control ability on the humidity uniformity within the partition, and finally screen out the target division scheme which can form a reasonable humidity partition boundary to avoid the redundancy of the number of division areas, and can control the soil humidity difference within each division area within an acceptable small fluctuation range.

[0087] Step S73, dividing the foundation based on the target division scheme to obtain a plurality of division areas of the foundation.

[0088] The plurality of division areas of the foundation are a plurality of division area units obtained by dividing the area of the foundation before excavation based on the target division scheme, and the soil within each division area has similar humidity characteristics.

[0089] Step S3, determining the estimated earthwork data of each division area based on the three-dimensional point cloud data before the foundation is excavated, the three-dimensional point cloud data after the foundation is excavated, and the plurality of division areas of the foundation.

[0090] In some embodiments, the estimated earthwork data of each division area can be determined based on the three-dimensional point cloud data before the foundation is excavated, the three-dimensional point cloud data after the foundation is excavated, and the plurality of division areas of the foundation using an earthwork analysis model, the earthwork analysis model is a Transformer model, the input of the earthwork analysis model is the three-dimensional point cloud data before the foundation is excavated, the three-dimensional point cloud data after the foundation is excavated, and the plurality of division areas of the foundation, and the output of the earthwork analysis model is the estimated earthwork data of each division area.

[0091] The estimated earthwork data of each division area is data representing the estimated excavation earthwork volume of each division area output by the earthwork analysis model. The estimated earthwork data is in cubic meters and can represent the volume of removed soil in the corresponding division area after the foundation is excavated.

[0092] The three-dimensional point cloud data before foundation excavation and the three-dimensional point cloud data after foundation excavation respectively contain the original terrain of the construction area before foundation excavation and the terrain information after foundation excavation, as well as physical characteristics such as color and reflectivity, thereby being able to provide a spatial data basis representing the change in terrain. The plurality of divided regions of the foundation define spatial units with different soil humidity characteristics through three-dimensional spatial coordinate ranges, so that the model can process differentiated soil characteristics according to the granularity of the regions. The Transformer model can capture the spatial correlation between the coordinate distribution and the boundary of the divided regions in the three-dimensional point cloud data before and after foundation excavation through the self-attention mechanism, and automatically learn the mapping rules between the terrain change characteristics, the humidity change characteristics and the earthwork data. Furthermore, the Transformer model uses the feedforward neural network to extract the implicit correlation between the point cloud reflectivity and the soil humidity, and thus is able to output the estimated earthwork data of each divided region.

[0093] Step S4, determining a cross-section shooting video of each divided region based on the plurality of divided regions of the foundation and the cross-section shooting video during the foundation excavation process.

[0094] In some embodiments, a gated recurrent unit can be used to determine the cross-section shooting video of each divided region based on the plurality of divided regions of the foundation and the cross-section shooting video during the foundation excavation process, the input of the gated recurrent unit being the plurality of divided regions of the foundation and the cross-section shooting video during the foundation excavation process, and the output of the gated recurrent unit being the cross-section shooting video of each divided region.

[0095] The gated recurrent unit (GRU) can be used to process sequence data and time series information. The gated recurrent unit includes three components: a memory unit, an update gate and a reset gate.

[0096] The cross-section shooting video of each divided region is divided from the cross-section shooting video during the foundation excavation process according to the plurality of divided regions of the foundation by the gated recurrent unit, and each divided region corresponds to a dedicated cross-section shooting video.

[0097] The gated recurrent unit models the time series characteristics of the video frame sequence through a gating mechanism, and then uses the update gate to remember the dynamic change of the boundary of the divided region in the continuous frames, while combining the spatial matching of each frame with the coordinate range of the region, so that the global video can be accurately divided into the cross-section shooting video corresponding to each divided region according to the spatial region.

[0098] Step S5, generating a simulation three-dimensional foundation image of each divided region and excavation information of each divided region based on the cross-section shooting video of each divided region.

[0099] In some embodiments, the simulation three-dimensional foundation image of each division region, the excavation information of each division region can be generated based on the cross-section shooting video of each division region using a generative adversarial network, an input of the generative adversarial network being the cross-section shooting video of each division region, an output of the generative adversarial network being the simulation three-dimensional foundation image of each division region, the excavation information of each division region.

[0100] The generative adversarial network is a deep learning model composed of a generator and a discriminator. The generative adversarial network can realize data generation through an adversarial training mechanism. The generator can try to generate samples similar to the real data distribution from random noise, while the discriminator is responsible for distinguishing between generated samples and real samples. Through a game process, both are constantly optimized, and finally the generator can generate realistic samples that are difficult to distinguish by the discriminator.

[0101] The simulation three-dimensional foundation image of each division region is a three-dimensional visualization image that can reflect the internal structure of the region soil, which is generated by the generative adversarial network based on the cross-section shooting video corresponding to the division region. The simulation three-dimensional foundation image contains soil layering information, soil internal humidity distribution, and geological anomaly body (such as cavity, fault) position, etc. physical characteristics. For example, the simulation three-dimensional foundation image of a high-humidity soil area can clearly show the depth range of the saturated soil layer, the water content gradient change, and the boundary relationship with the surrounding area. The color value of each pixel in the image corresponds to a specific soil humidity value.

[0102] The excavation information of each division region is the quantitative information of the foundation soil physical properties and the excavation process in each division region output by the generative adversarial network. The excavation information includes soil structure information, soil physical properties, and material content information.

[0103] The soil structure information includes the soil cavity area, cavity distribution density, and cavity connectivity in the division region.

[0104] The soil physical properties include soil looseness, soil humidity gradient, and soil internal friction angle.

[0105] The material content information includes stone content ratio, clay content ratio, and organic matter content.

[0106] The cross-section video records the continuous dynamic information of the soil structure and humidity change in the range of the divided area during the whole foundation excavation process. By analyzing the changes of soil color and texture in the video, the soil internal humidity gradient and material content distribution can be determined. The model can use stereo matching between consecutive frames to construct the three-dimensional geometry of each divided area, including the location, size, and connectivity of the voids. These data can accurately support the geometric modeling and physical property assignment of the simulated three-dimensional foundation image for each divided area, as well as the dynamic quantification of the excavation information for each divided area.

[0107] The generative adversarial network (GAN) can efficiently convert the two-dimensional time series information in the cross-section video into three-dimensional spatial data and physical parameters based on its unique adversarial training mechanism. The generator learns the visual features such as soil texture, color, and geometric relationship in the video frames to generate three-dimensional foundation structures that conform to physical laws, such as the layered interface of clay and sand. The discriminator continuously compares the generated data with the real geological structure to ensure the accuracy and reliability of the output simulated three-dimensional foundation image. At the same time, the generative adversarial network can establish a mapping relationship between the visual features in the cross-section video and the physical parameters such as soil void area and looseness to output the excavation information for each divided area.

[0108] Step S6, determining the earthwork correction coefficient of each divided area based on the simulated three-dimensional foundation image of each divided area and the excavation information of each divided area.

[0109] In some embodiments, the earthwork correction coefficient of each divided area can be determined using a correction model based on the simulated three-dimensional foundation image of each divided area and the excavation information of each divided area. The correction model is a convolutional neural network, the input of the correction model is the simulated three-dimensional foundation image of each divided area and the excavation information of each divided area, and the output of the correction model is the earthwork correction coefficient of each divided area.

[0110] The earthwork correction coefficient of each divided area is a numerical value that accurately corrects the soil volume after compaction under ideal conditions.

[0111] For example, in a certain divided area, the soil is high-humidity clay. Due to the high water content of the soil, "rubber soil" is formed, and the water film between soil particles hinders the close arrangement of particles, resulting in a volume that exceeds the expected volume under normal compaction conditions. The correction model determines that the earthwork correction coefficient of this area is 0.95, i.e., the volume after compaction is 95% of the estimated soil data for this divided area.

[0112] For another divided region, the soil composition is a sandy soil containing a large number of cavities, the volume of the excavated soil after compaction is 82% of the excavated amount, and the soil correction coefficient of the region is set to 0.82 accordingly.

[0113] The convolutional neural network can slide on the simulated three-dimensional foundation image through the convolution kernel of the convolutional layer, and then automatically extract spatial features such as soil layering structure and cavity distribution, and reduce the data dimension while retaining key information through the pooling layer. For the soil structure indicators, soil physical properties, and material content indicators in the excavated information, the convolutional neural network can establish an associated mapping of the three-dimensional foundation image features through the fully connected layer. Through this multi-level feature extraction and fusion mechanism, the convolutional neural network can capture the nonlinear relationship between the foundation physical properties and the soil correction coefficient, for example, identifying that high humidity areas correspond to larger correction coefficients and multiple cavity areas correspond to smaller correction coefficients, and finally outputting the soil correction coefficient of each divided region that fits the actual situation.

[0114] Step S7, determining the corrected total earthwork result data based on the estimated earthwork data of each divided region and the correction coefficient of each divided region.

[0115] The corrected total earthwork result data refers to the soil volume after the excavated soil is removed and compacted as a whole after the foundation excavation is completed.

[0116] The calculation method of the corrected total earthwork result data is to multiply and couple the estimated earthwork data of each divided region with the correction coefficient of the corresponding divided region to obtain the compacted earthwork volume of each divided region, then sum up the compacted earthwork volumes of all divided regions, and finally obtain the compacted earthwork volume covering the entire foundation excavation construction area, which is the corrected total earthwork result data.

[0117] Based on the same inventive concept, Figure 9 A three-dimensional point cloud-based foundation data processing system schematic diagram is provided for the embodiments of the present application, which comprises:

[0118] The data acquisition module 91 is configured to acquire three-dimensional point cloud data before foundation excavation, three-dimensional point cloud data after foundation excavation, and cross-section shooting video during foundation excavation.

[0119] The region division module 92 is configured to determine a plurality of divided regions of the foundation based on the three-dimensional point cloud data before foundation excavation.

[0120] The earthwork estimation module 93 is configured to determine estimated earthwork data of each divided region based on the three-dimensional point cloud data before the foundation is excavated, the three-dimensional point cloud data after the foundation is excavated, and the plurality of divided regions of the foundation.

[0121] The video processing module 94 is configured to determine cross-section shooting videos of each divided region based on the plurality of divided regions of the foundation and the cross-section shooting videos during the excavation of the foundation.

[0122] The image generation module 95 is configured to generate a simulated three-dimensional foundation image of each divided region and excavation information of each divided region based on the cross-section shooting videos of each divided region.

[0123] The correction coefficient determination module 96 is configured to determine an earthwork correction coefficient of each divided region based on the simulated three-dimensional foundation image of each divided region and the excavation information of each divided region.

[0124] The earthwork correction module 97 is configured to determine corrected overall earthwork result data based on the estimated earthwork data of each divided region and the correction coefficient of each divided region.

[0125] In addition, unless the claims recite otherwise, the order of the processing elements and sequences, the use of numerals, or the use of other designations herein are not intended to limit the scope of the processes and methods described in this specification. Although the above disclosure discusses some presently preferred embodiments of the application by way of various examples, it is to be understood that such disclosure of details is not intended to limit the scope of the application, which is defined in the appended claims only. On the contrary, it is believed that the scope of the application encompasses all modifications and equivalents consistent with the spirit and scope of the described embodiments as defined by the appended claims. For example, although the system components described above can be implemented by hardware devices, they can also be implemented by software solutions, such as installing the described system on existing servers or mobile devices.

[0126] Similarly, it should be noted that the description of the embodiments of this specification, in order to simplify the expression of this disclosure and to help the understanding of one or more embodiments of the application, sometimes combines various features into one embodiment, figure or description thereof. However, this method of disclosure does not mean that the features required by the subject of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0127] Finally, it should be understood that the embodiments described herein are only given by way of example and that other modifications can occur to persons skilled in the art. Therefore, the scope of the present description is not intended to be limited to the embodiments described herein but is only limited by the claims that follow.

Claims

1. A ground data processing method based on three-dimensional point cloud, characterized in that: include: Obtaining 3D point cloud data before and after foundation excavation, and cross-sectional video during foundation excavation; Determining a plurality of divided areas of the foundation based on the three-dimensional point cloud data before the foundation excavation, wherein determining a plurality of divided areas of the foundation based on the three-dimensional point cloud data before the foundation excavation comprises: Determining a plurality of significant point cloud sets based on the three-dimensional point cloud data before foundation excavation, wherein the plurality of significant point cloud sets are a plurality of discrete point sets having significant terrain or material features; Determining, based on the plurality of significant point clouds, a plurality of highly significant dividing points, a plurality of moderately significant dividing points, and a plurality of lowly significant dividing points, wherein the plurality of highly significant dividing points are a set of discrete points that have a decisive influence on the characteristic difference of soil moisture on the foundation surface, the plurality of moderately significant dividing points are a set of discrete points that have a moderate degree of influence on the characteristic difference of soil moisture on the foundation surface, and the plurality of lowly significant dividing points are a set of discrete points that have a low degree of influence on the characteristic difference of soil moisture on the foundation surface; Determining multiple sets of partitioning schemes based on the multiple high-saliency partitioning points, the multiple medium-saliency partitioning points, and the multiple low-saliency partitioning points; Determining a plurality of partitioned areas of the foundation using a neural network model based on the plurality of partitioning schemes; Determining estimated earthwork data for each divided area based on the three-dimensional point cloud data before foundation excavation, the three-dimensional point cloud data after foundation excavation, and the multiple divided areas of the foundation; determining a cross-sectional captured video of each divided area based on the plurality of divided areas of the foundation and the cross-sectional captured video during the excavation of the foundation; generating a simulated three-dimensional foundation image of each divided area and excavation information of each divided area based on the cross-sectional shot video of each divided area; determining an earthwork correction coefficient for each divided area based on the simulated three-dimensional foundation image of each divided area and the excavation information of each divided area; Corrected total earthwork result data is determined based on the estimated earthwork data of each divided area and the correction coefficient of each divided area.

2. The ground-based data processing method based on three-dimensional point cloud according to claim 1, characterized in that: The foundation excavation is the straight foundation excavation of the runway.

3. The ground-based data processing method based on three-dimensional point cloud according to claim 1, characterized in that: The method of determining the multiple division areas of the foundation using a neural network model based on the multiple division schemes includes: Constructing a scheme graph based on the multiple sets of partitioning schemes, the scheme graph includes multiple partitioning scheme nodes and edges between the nodes, the node features of the partitioning scheme nodes include three-dimensional point cloud data of each partitioned area corresponding to a partitioning scheme, and the edges between the partitioning scheme nodes represent the differences between the partitioning schemes; The target partitioning scheme is obtained by processing the scheme graph based on the graph neural network; The foundation is divided based on the target division scheme to obtain multiple division areas of the foundation.

4. A ground-based data processing system based on a three-dimensional point cloud, for implementing the ground-based data processing method based on a three-dimensional point cloud according to any one of claims 1 to 3, characterized in that: include: A data acquisition module is used to acquire three-dimensional point cloud data before foundation excavation, three-dimensional point cloud data after foundation excavation, and cross-sectional video during the foundation excavation process; A region division module is used to determine a plurality of divided regions of the foundation based on the three-dimensional point cloud data before the foundation excavation, and the region division module is further used to: Determining a plurality of significant point cloud sets based on the three-dimensional point cloud data before foundation excavation, wherein the plurality of significant point cloud sets are a plurality of discrete point sets having significant terrain or material features; Determining, based on the plurality of significant point clouds, a plurality of highly significant dividing points, a plurality of moderately significant dividing points, and a plurality of lowly significant dividing points, wherein the plurality of highly significant dividing points are a set of discrete points that have a decisive influence on the characteristic difference of soil moisture on the foundation surface, the plurality of moderately significant dividing points are a set of discrete points that have a moderate degree of influence on the characteristic difference of soil moisture on the foundation surface, and the plurality of lowly significant dividing points are a set of discrete points that have a low degree of influence on the characteristic difference of soil moisture on the foundation surface; Determining multiple sets of partitioning schemes based on the multiple high-saliency partitioning points, the multiple medium-saliency partitioning points, and the multiple low-saliency partitioning points; Determining a plurality of partitioned areas of the foundation using a neural network model based on the plurality of partitioning schemes; an earthwork estimation module, configured to determine estimated earthwork data for each divided area based on the three-dimensional point cloud data before foundation excavation, the three-dimensional point cloud data after foundation excavation, and the multiple divided areas of the foundation; A video processing module, configured to determine a cross-sectional video of each divided area based on the plurality of divided areas of the foundation and the cross-sectional video during the foundation excavation process; An image generation module is used to generate a simulated three-dimensional foundation image of each divided area and excavation information of each divided area based on the cross-sectional video of each divided area; a correction coefficient determination module, configured to determine an earthwork correction coefficient for each divided area based on the simulated three-dimensional foundation image of each divided area and the excavation information of each divided area; The earthwork correction module is used to determine the corrected total earthwork result data based on the estimated earthwork data of each divided area and the correction coefficient of each divided area.

5. The ground-based data processing system based on three-dimensional point cloud according to claim 4, characterized in that: The foundation excavation is the straight foundation excavation of the runway.

6. The ground-based data processing system based on three-dimensional point cloud according to claim 4, characterized in that: The area division module is also used for: Constructing a scheme graph based on the multiple sets of partitioning schemes, the scheme graph includes multiple partitioning scheme nodes and edges between the nodes, the node features of the partitioning scheme nodes include three-dimensional point cloud data of each partitioned area corresponding to a partitioning scheme, and the edges between the partitioning scheme nodes represent the differences between the partitioning schemes; The target partitioning scheme is obtained by processing the scheme graph based on the graph neural network; The foundation is divided based on the target division scheme to obtain multiple division areas of the foundation.

7. An electronic device, characterized in that: include: processor; Memory; And a computer program; wherein, the computer program is stored in the memory and is configured to be executed by the processor to implement the ground-based data processing method based on three-dimensional point cloud according to any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the ground-based data processing method based on three-dimensional point cloud as described in any one of claims 1 to 3 is implemented.

Images