CN115655098A - High-density laser point cloud technology used in power grid engineering excavation and filling earthwork calculation method - Google Patents

Abstract

The invention discloses a method for measuring and calculating excavated and filled earthwork of a power grid project by using a high-density laser point cloud technology, and belongs to the technical field of actual calculation of earthwork of soil and water conservation monitoring work. The unmanned aerial vehicle carries on the laser acquisition equipment, gathers high density laser point cloud data, according to the region of taking a photograph by plane, applies for and handles the flight wholesale. The method comprises the steps of selecting a suitable field along the line as a take-off and landing base, signing a guarantee protocol with an air traffic control unit, coordinating an airspace, performing batch flight according to the related air traffic control unit, designing, self-checking and checking, ensuring the accuracy of data in the design process of an aerial zone, and being better applied to engineering application.

Description

High-density laser point cloud technology used in power grid engineering excavation and filling earthwork calculation method

technical field

The invention relates to the technical field of actual calculation of earthwork for soil and water conservation monitoring work, and more specifically, relates to a method for measuring and calculating earthwork for excavation and filling of power grid projects using high-density laser point cloud technology.

Background technique

The amount of earth and stone is a very important value in the water and soil conservation work of construction projects, and it is also very important in the process of the main project. Accurate measurement of earthwork volume is of great significance for the acceptance of water conservation, earthwork disposal and transportation of construction projects. However, the amount of earth and stone in the current construction project cannot be accurately measured, and can only be estimated by the number of earth transportation vehicles, and the results are often very different from the actual situation.

High-density laser point cloud (UAV Lidar) technology can extract elevation information very accurately, and is usually used for terrain survey, power transmission line selection, channel data extraction, etc. It is seldom used in the calculation of soil and water conservation in engineering construction.

In view of this, we propose high-density laser point cloud technology for the calculation method of earthwork excavation and filling in power grid projects.

Contents of the invention

technical problem to be solved

The purpose of the present invention is to provide a high-density laser point cloud technology for earthwork calculation method for excavation and filling of power grid projects, so as to solve the problems raised in the above-mentioned background technology.

Technical solutions

High-density laser point cloud technology is used in power grid engineering excavation and filling earthwork calculation method, including the following steps:

S1: The UAV is equipped with a laser radar to collect high-density laser point cloud data before and after the excavation and filling of the project;

S2: After the point cloud is automatically classified, manual review is carried out, and the data is processed in the office;

S3: Use GIS to delineate the calculation range;

S4: Input the data into the data system, use the data before and after excavation to calculate the volume change, and calculate the earth and stone volume.

As a preferred solution of the artificial intelligence-based material level sensing device described in the present invention, step S1 also includes the following steps: Step S1 also includes the following steps:

S101: UAV Lidar data processing removes noise points and transforms the coordinates of the data obtained by the airborne Lidar, then automatically classifies the point cloud, manually checks and corrects the classified results, and finally outputs the corresponding classification according to the requirements results.

Based on the above technical features: in the process of designing the flight belt, three steps of design, self-inspection and verification are required to ensure the accuracy of the data and better apply it to engineering applications.

As a preferred solution of the artificial intelligence-based material level sensing device described in the present invention, step S101 also includes the following steps:

S1011: Preprocessing point cloud data;

S1012: point cloud noise removal;

S1013: classification of point cloud data;

S1014: Mapping the point cloud data, processing the ground point model, and obtaining elevation information.

Based on the above technical features: the purpose of laser point cloud data preprocessing is to obtain the spatial coordinates of point cloud data and provide data for post-production DEM. The data required for processing includes: original laser point cloud data, POS solution data, etc. The processing content includes laser point cloud, laser point cloud coordinate conversion, etc.

As a preferred solution of the artificial intelligence-based material level sensing device described in the present invention, S1012: point cloud noise removal;

Based on the above technical features: Lidar will be affected by various factors during the process of collecting 3D point cloud data, so some noise will appear when acquiring data. In fact, in actual work, in addition to the error of self-measurement, it will also be affected by the external environment, such as the measured target being blocked, obstacles and the surface material of the measured target and other influencing factors; in addition, some local large-scale noise due to the distance from the target point cloud Farther away, it cannot be filtered using the same method.

Noise is a point that has no connection with the target information description, and is useless for the subsequent reconstruction of the entire 3D scene; the workflow of laser point cloud data processing is: use the POS solution result and the point cloud of the calibration field to eliminate the eccentricity angle Error, use the line shape measurement point of the calibration field to eliminate the torsion error, the elevation point to eliminate the distance error, and finally perform coordinate conversion on the point cloud data after calibration to generate point cloud data that meets the requirements;

In the actual point cloud data processing algorithm, it is not easy to distinguish the noise point from the target point with feature information. In the process of denoising, it is always accompanied by some feature information due to many external factors. lost. A good point cloud filtering algorithm not only requires high real-time performance, but also preserves the feature information of the model well while denoising. It is necessary to thoroughly apply the noise point features of point cloud data in order to propose a better denoising algorithm.

Point cloud data is an unstructured data format. The point cloud data obtained by lidar scanning is affected by the distance between the object and the radar, and the distribution is uneven. Object point cloud data is sparsely distributed. In addition, point cloud data is characterized by disorder and asymmetry, which leads to the lack of a clear and unified data structure for point cloud data in data representation, which exacerbates the difficulty of subsequent point cloud segmentation and recognition processing. As an end-to-end network structure, the neural network often processes conventional input data, such as sequences, images, videos, and 3D data. It cannot directly process unordered data such as point sets. Convolution When operating and processing point cloud data, convolution directly discards the shape information of the point cloud, and only retains the sequence information of the point cloud.

As a preferred solution of an artificial intelligence-based material level sensing device described in the present invention, wherein S1033: classification of point cloud data;

Based on the above technical features: In TerraSolid software, the point cloud data classification layers are as follows: 1) Default; 2) Ground; 3) vegetation; 4) Building; 5) Lowpoint; 6) Modelkeypoints. Among them, the Default layer is the point cloud data layer temporarily stored in the operation process; the Ground (ground point) mainly stores the real topography of the ground (artificially built embankments, earth embankments, ladder roads, naturally formed and large-scale pits and piles), and artificially constructed soil ridges, dams, dikes, sluices and other hydraulic structures connected to the ground are regarded as ground points; Vegetation (vegetation points) mainly store the ground Vegetation points, such as grassland, shrubs, bamboo forests, nurseries, young forests, gardens, and woodlands, are regarded as vegetation points; Building (building points) mainly store surface buildings, and houses and greenhouses are regarded as building points.

The optimal filter function is determined according to the terrain slope change. For a given height difference, as the distance between two points decreases, the possibility that the laser footpoint with a large elevation value belongs to the ground point becomes smaller.

The slope-based filtering algorithm has the characteristics of simple calculation and strong adaptability, but it needs to know the slope of the terrain in advance and determine the size of the window to be opened. The selected point must be compared with all other points to determine whether the point is a ground point or not. It is necessary to calculate the slope of each point in the entire data set, which will inevitably increase the amount of calculation and slow down the speed. The original point cloud data is divided into blocks according to the statistical characteristics of the terrain, and then each block is changed according to the slope. The filtering algorithm is used to process the ground point sets of each block of data, and finally the blocks are spliced according to the feature points of the overlapping area to obtain a complete set of ground points. In this way, different blocks get different filtering thresholds, which avoids the singleness of thresholds and reduces classification errors.

TIN is an important model to represent digital elevation, and it is often used to store the proximity relationship between spatial discrete points. For irregularly distributed elevation points, it can be formally described as an unordered point set on the plane. The common method to convert the point set into TIN is to construct the Delaunay triangulation network of the point set. The adjacent points in the Delaunay triangulation network are in the TIN Also adjacent in the model.

The filtering algorithm is to use the elevation mutation relationship of the adjacent point cloud of the ground object in the TIN model, and apply the parameters such as the height difference critical value condition and the number of adjacent points satisfying the condition to filter the ground object point;

As a preferred solution of the artificial intelligence-based material level sensing device in the present invention, S1014: map point cloud data, process ground point models, and obtain elevation information.

Based on the above technical characteristics: the solution to the above problems needs to be in the actual point cloud filtering operation, according to the specific terrain and features and their spatial distribution characteristics, select the appropriate combination of constraints and parameters, and adopt a progressive filtering method to achieve Satisfactory results, in quite a few cases, also need to divide the continuous ground into several blocks according to the law of ground undulations, and select the respective parameter combinations according to the characteristics of the blocks.

As a preferred solution of the artificial intelligence-based material level sensing device described in the present invention, step S1011 also includes the following steps:

S10111: Filter the point cloud preprocessing information, including slope-based filtering based on mathematical morphology, TIN-based UAV Lidar point cloud filtering, pseudo scan line filtering, and wavelet layered filtering.

Based on the above technical features: when the mathematical morphology theory is used in the UAV Lidar point cloud data filtering, in order to extract ground points, the traditional mathematical morphology "open" operator is improved, and processed according to the following process:

Discrete pit corrosion treatment. Traverse the UAV Lidar point cloud data, open a window of w×w size with any point as the center, compare the elevation of each point in the window, and take the minimum elevation value in the window as the elevation after corrosion;

Discrete point dilation processing. Traverse the UAV Lidar point cloud data again, and use the same size structural window to expand the corroded data. That is, open a window of w×w size with any point as the center. At this time, replace the original elevation value with the corroded elevation value, compare the elevations of each point in the window, and take the maximum elevation value in the window as the height after expansion;

Ground point extraction. Let Zp be the original elevation of point p, and t be the threshold value. At the end of each point expansion operation, a judgment is made whether the point is a ground point. If the absolute value of the difference between the inflated elevation value of point p and its original elevation value Zp is less than or equal to the threshold t, point p is considered to be a ground point, otherwise it is a non-ground point.

Determine the optimal filter function according to the terrain slope change. For a given height difference, as the distance between two points decreases, the possibility that the laser footpoint with a large elevation value belongs to the ground point is less likely; the filtering algorithm based on slope It has the characteristics of simple calculation and strong adaptability, but it needs to know the slope of the terrain in advance and determine the size of the opened window. The selected point must be compared with all other points to determine whether the point is a ground point. It also needs to be in the entire data set. , to calculate the slope of each point, which will inevitably increase the amount of calculation and slow down the speed. The original point cloud data is divided into blocks according to the statistical characteristics of the terrain, and then each block is processed according to the filtering algorithm based on the slope change. The ground point set of each block of data is obtained, and finally the blocks are spliced according to the feature points of the overlapping area to obtain a complete set of ground points. In this way, different blocks get different filtering thresholds, which avoids the singleness of thresholds and reduces classification errors;

TIN is an important model to represent digital elevation, and it is often used to store the proximity relationship between spatial discrete points. For irregularly distributed elevation points, it can be formally described as an unordered point set on the plane. The common method to convert the point set into TIN is to construct the Delaunay triangulation network of the point set. The adjacent points in the Delaunay triangulation network are in the TIN Also adjacent in the model.

The filtering algorithm is to use the elevation mutation relationship of the adjacent point cloud of the ground object in the TIN model, and apply the parameters such as the height difference critical value condition and the number of adjacent points satisfying the condition to filter the ground object points.

Suppose a non-empty point cloud Pt_Cloud is given, and two threshold conditions of height difference (threshold_h) and number of adjacent points (threshold_vn) are given according to the distribution of regional terrain, buildings, vegetation, etc., and elevation changes, and Filtered and Unfiltered are defined Two arrays record filtered points and unfiltered points respectively.

The method based on the triangulated irregular network (TIN) is a method based on two-dimensional neighborhood search, and its calculation amount and algorithm complexity are relatively large. Generally speaking, due to the obvious elevation mutation between tall buildings and vegetation and its adjacent ground points, the algorithm can filter the elevation mutation objects better, but when filtering shrubs or low ground objects, there will be Excessive error. Mao Jianhua et al. used the TIN filter algorithm to filter the experimental data (data including buildings, roads, vegetation, dry land and reservoirs, etc.) located in undulating mountainous areas, and the results showed that vegetation, buildings, etc. The filtering effect of is better. In the obtained DEM, the terrain fluctuations are basically maintained, and the basic shape of the road can be clearly displayed, but the error is large in the dry land flat area.

The solution to the above problems needs to be in the actual point cloud filtering operation, according to the specific terrain and objects and their spatial distribution characteristics to select the appropriate combination of constraint parameters, and adopt a progressive filtering method to achieve satisfactory results. In quite a few cases, it is also necessary to divide the continuous ground into several blocks according to the law of ground undulations, and select the respective parameter combinations according to the characteristics of the blocks;

The pseudo-scan line method reorganizes the two-dimensional discretely distributed laser points on the horizontal plane into a data structure of a one-dimensional linear continuous distribution point sequence, which is called a pseudo-scan line.

Filtering based on pseudo-scanlines uses elevation mutation information to distinguish ground points from non-ground points. Its basic idea is: the height difference between two points is caused by the undulation of natural terrain and the height of ground objects. If the height difference between two adjacent points is larger, then the possibility that the height difference is caused by natural terrain is smaller, and it is more likely that the higher point is on the ground object and the lower point is on the ground, that is, : Suppose there are two adjacent laser foot points p ₁ and p ₂ , p ₁ is the ground point, and p ₂ is its adjacent point. If their height values h ₁ and h ₂ satisfy the condition: h ₂ -h ₁ ≤Δh _max ×d (Δh _max is the tolerance of height difference, d is the horizontal distance between them), then p ₂ is considered to be the ground point, otherwise p ₂ is considered to be a non-ground point.

Based on the pseudo scan line filtering algorithm, the two-dimensional filtering problem is simplified to one-dimensional filtering problem. The algorithm structure is simple, which effectively reduces the calculation amount of filtering and ensures the accuracy. At the same time, the algorithm only needs two filtering parameters, which is easier to realize automation. . However, because most of the two-dimensional filters in the local neighborhood assume that the lowest elevation point in the neighborhood is the ground point, when there are few ground points, this kind of filtering method often fails. However, the filtering algorithm based on pseudo-scanning lines can always ensure that each filtering window contains ground points, and can obtain a relatively small type of error and total error, and accurately extract topographic points. In flat areas, the effect of pseudo-scanline filtering is very good, and in areas with relatively steep terrain, its error is also controlled within a small range. However, in areas with steep slopes and dramatic elevation changes or when filtering large objects, in order to obtain reliable results, it is usually necessary to reduce the threshold value of the elevation and the size of the filtering window; in urban areas, in order to filter out all large buildings, It is necessary to increase the filtering window appropriately so that the size of the filtering window is not smaller than the maximum size of the building. At present, the selection of these two parameters cannot be fully automated, and the method needs to be further improved.

Based on the filtering method of multi-resolution direction prediction, the point cloud of UAV Lidar laser footpoints located on different ground objects will show different elevation differences, and the elevation mutation between adjacent laser footpoints can be used to distinguish ground points from non-ground points .

For a certain distance range, if the difference between the current point pi and the predicted values of all directions is greater than the maximum height difference tolerance under the distance condition, the point is a ground object point, otherwise it is a ground point.

Wavelet layering method; wavelet transform is introduced into UAV Lidar data filtering, and data filtering based on wavelet transform is realized. In general, the ground point of the point cloud data of UAV Lidar is the lowest point of elevation in the local area. Therefore, the original data can be divided by windows of a specific size, and then a point with the lowest elevation can be selected in each window to form a new data description. These ground points are networked to form a rough terrain surface. This rough terrain surface is then used as a reference surface for filtering in the next layer to get more ground points. It must be guaranteed that there is at least one ground point in each segmentation window, which requires the segmentation window to be large enough. A window slightly larger than the largest building area in the target area is used as the scale described by the top layer of data. Next, determine the window scale described by the first layer of data. Take the area of the smallest man-made building in the target area as the window scale described by the first layer of data.

The wavelet layered filtering algorithm needs to be layered first, and then pass this rough domain of interest to the next layer as the initial value of the domain of interest in the current layer, thereby reducing the calculation time and improving the accuracy of the processing results, but each layer The result of the judgment is affected by the upper layer. If there is an error in the processing of the upper layer, this error will lead to an error in the judgment of the data point type of the next layer. In addition, the selection of the scale of the split window is also very important. The scale described by the top layer of data should be slightly larger than the largest building area in the target area, and the scale of the window described by the first layer of data should be selected from the smallest artificial The size of the area of the building. In addition, the wavelet layered filtering algorithm also needs to consider more carefully in the selection of data initial values and discrimination rules, and eliminate the gross errors in the data.

Vegetation and building classification method application, using echo information for point cloud classification.

At present, methods for extracting vegetation and buildings based on LiDAR point clouds mainly include supervised learning classification methods and point cloud segmentation and hierarchical extraction methods. The former mainly uses machine learning methods to train feature samples to generate classifiers to extract buildings. The performance of depends on the selection of samples, and the cost is large. The latter is based on the auxiliary information provided by the LiDAR system or the three-dimensional topological relationship of the point cloud for feature classification.

As a preferred solution of the artificial intelligence-based material level sensing device described in the present invention, step S10111 also includes the following steps:

S101111: Ground point model processing, modeling the automatically classified ground points. If there is an unreasonable triangulation, manually divide the unseparated points into ground points until no unreasonable triangulation appears. It can handle the areas with sudden changes in elevation;

S101112: Obtain the elevation information, separate the key points of the contour line on the finely classified ground point model, and use the software to automatically generate the contour line. Will be removed to generate contours at the scale required for the project. After the contour lines are exported, the key points of the model are exported to the ENZ format, and the contour lines and elevation points are edited with CASS software to complete the collection of elevation elements of the topographic map.

Step S2 also includes:

As a preferred solution of the artificial intelligence-based material level sensing device in the present invention, S201: Check the data source before construction.

Based on the above technical features: Before construction, the Yun-5 aircraft equipped with UCXp-WA aerial camera was used to obtain the aerial photography of the application area. The acquisition time was December 2016. Encrypt and extract the digital elevation model data with a grid interval of 0.5m within the scope of the aerial photo, and obtain the results of ground points (*. Fast tessellation of elevation models.

As a preferred solution of the artificial intelligence-based material level sensing device described in the present invention, S202: Check the data source of the completion period,

Based on the above technical features: during the completion of the project, the Pegasus UAV equipped with the SONYILCE-5100 aerial camera was used to obtain aerial images of the application area. The acquisition time was August 2021. The intelligent mosaic module of the human-machine housekeeper can realize the rapid stitching of orthophotos through the process of aerial triangulation, dem generation, and fast mosaic process.

As a preferred solution of an artificial intelligence-based material level sensing device in the present invention, S303: Summarize data sources.

Based on the above technical features: the data of the two are aggregated and processed.

Preferably, said step S4 also includes the following steps:

S401: Create a project, use software, create a new project, set project parameters, and build a unified data platform for subsequent calculations.

S402: Import point cloud data; right-click CreatePointClould on pointClouds in the Prospector to create point cloud data.

Define the basic point cloud information: specify the name, point cloud style (Single), and point cloud layer (V-SITE-SCAN) in the information dialog box;

In the SourceData dialog box, SourceData (source data) select "create new point cloud database", select "LAS" under "point cloud file format", add data and assign the corresponding coordinate system. Under New Point Cloud Database, enter a name under Specify New Point Cloud Database. Among them, the settings of "point cloud database coordinate system" and "coordinate system of current drawing" should match.

Point cloud database coordinate system modification: This value is obtained from the point cloud source file, and the coordinate system can be changed after the data is imported into the point cloud database.

Coordinate System Modification of Current Drawing: This value is obtained from the Units and Zones tab of the Drawing Settings dialog.

S403: Add point cloud data to the surface, specify source data, verify point cloud parameters, and create point cloud objects.

Beneficial effect

Compared with the prior art, the present invention has the advantages of:

1. It is possible to calculate the slope of each point, which will inevitably increase the amount of calculation and slow down the speed. The original point cloud data is divided into blocks according to the statistical characteristics of the terrain, and then each block is followed by a filtering algorithm based on slope changes The ground point set of each block of data is obtained by processing, and finally the blocks are spliced according to the feature points of the overlapping area to obtain a complete set of ground points. In this way, different blocks get different filtering thresholds, which avoids the singleness of thresholds and reduces classification errors.

2. Calculate the slope of each point, which will inevitably increase the amount of calculation and slow down the speed. The original point cloud data is divided into blocks according to the statistical characteristics of the terrain, and then each block is followed by a filtering algorithm based on slope changes. The ground point sets of each block of data are processed, and finally the blocks are spliced according to the feature points of the overlapping areas to obtain a complete set of ground points. In this way, different blocks get different filtering thresholds, which avoids the singleness of thresholds and reduces classification errors.

3. During the completion of the project, use Pegasus UAV equipped with SONYILCE-5100 aerial camera to obtain the aerial photography image of the application area. The acquisition time is August 2021. The original aerial image resolution is 0.10m, and the Pegasus UAV steward is used The intelligent mosaic module can realize the rapid stitching of orthophotos through the process of aerial triangulation, dem generation, and fast mosaic process.

4. In the earthwork balance, it is necessary to comprehensively consider the overall earthwork balance results in the construction of the site, slope, road, etc., and the earthwork volume of the site leveling occupies a relatively large proportion. This application compares the earth and rock volume calculated by the UAV Lidar point cloud in the project area with the earth and rock volume in the quarterly report (annual report) of soil and water conservation monitoring, and analyzes the actual application effect of the UAV Lidar data in the earth and rock volume estimation. During the calculation process, the calculation range is obtained from the water and soil conservation zoning map provided by the design institute, and the coordinates are unified.

Description of drawings

Fig. 1 is the flow chart of laser data preprocessing of the present invention;

Fig. 2 is a schematic diagram of the first and last echoes of plants and the first and last echoes of building edges in the present invention;

Fig. 3 is a schematic flow chart of the method for extracting building edges based on local multi-feature point cloud classification in the present invention;

Fig. 4 is that the present invention imports point cloud data figure;

Fig. 5 specifies the basic point cloud information figure for the present invention;

Fig. 6 is a schematic diagram of the operation of specifying source data in the present invention;

Fig. 7 is a schematic diagram of the image setting operation of the present invention;

Fig. 8 is a schematic diagram of verifying point cloud parameters in the present invention;

Fig. 9 is that the present invention adds point cloud figure to curved surface;

Fig. 10 is a schematic diagram of the scope operation of the specified area or the self-defined composition surface of the present invention;

Fig. 11 is a schematic diagram of information forming a curved surface in the present invention;

Fig. 12 creates a design surface diagram for the present invention;

Fig. 13 is that the present invention creates element lines from objects;

Fig. 14 is a display diagram of the construction of the design curved surface of the present invention;

Fig. 15 is a schematic diagram of the operation of adding a boundary to a designed curved surface according to the present invention.

Fig. 16 is a schematic diagram of the operation of creating a volumetric surface in the present invention.

Detailed ways

In describing the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", " Orientation indicated by rear, left, right, vertical, horizontal, top, bottom, inside, outside, clockwise, counterclockwise, etc. The positional relationship is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, Therefore, it should not be construed as limiting the invention.

In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

In the description of the present invention, it should be noted that, unless otherwise specified and limited, the terms "installed", "set with", "sleeved/connected", "connected", etc. should be understood in a broad sense, such as " Connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediary, and it can be an internal connection between two components. connectivity. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention in specific situations.

Please refer to Figure 1-16, the present invention provides a technical solution:

Example 1

High-density laser point cloud technology is used in power grid engineering excavation and filling earthwork calculation method, including the following steps:

S1: The UAV is equipped with laser radar to collect high-density laser point cloud data;

Step S1 also includes the following steps:

S101: According to the aerial photography area, apply for the approval document of the aerial flight. Select a suitable site along the route as the take-off and landing base, and sign a guarantee agreement with the air traffic control unit, coordinate the airspace and fly according to the approval of the relevant air traffic control unit;

S102: Three steps are required: design, self-inspection, and verification. The accuracy of the data can be guaranteed during the flight belt design process, and it can be better applied to engineering applications;

S103: UAV Lidar data processing removes noise points and coordinates conversion of the point cloud preprocessing results obtained by the airborne lidar, and then automatically classifies the point cloud, manually checks and corrects the classified results, and finally according to the requirements Output the corresponding classification results;

S2: After the point cloud is automatically classified, manual review is carried out, and the data is processed in the office;

S3: Use GIS to delineate the calculation range;

S4: Input the data into the data system, use the data of the two periods to calculate the volume change, and calculate the earth and stone volume.

Example 2

Point cloud data preprocessing, the purpose of laser point cloud data preprocessing is to obtain the spatial coordinates of point cloud data and provide data for post-production DEM. The data required for processing includes: original laser point cloud data, POS solution data, etc. The processing content includes laser point cloud, laser point cloud coordinate conversion, etc.

The workflow of laser point cloud data processing is as follows: use the POS solution result and the point cloud of the calibration field to eliminate the eccentric angle error, use the calibration field line shape measurement point to eliminate the torsion error, and the elevation point to eliminate the distance error, and finally the point cloud data after calibration Perform coordinate transformation to generate point cloud data that meets the requirements

Point cloud noise removal;

Lidar will be affected by various factors during the process of collecting 3D point cloud data, so some noise will appear when acquiring data. In fact, in actual work, in addition to the error of self-measurement, it will also be affected by the external environment, such as the measured target being blocked, obstacles and the surface material of the measured target and other influencing factors; in addition, some local large-scale noise due to the distance from the target point cloud Farther away, it cannot be filtered using the same method.

Noise is a point that has no relationship with the target information description, and is useless for the subsequent reconstruction of the entire 3D scene. However, in the actual point cloud data processing algorithm, it is not easy to distinguish the noise point from the target point with feature information. In the process of denoising, it is always accompanied by some feature information due to many external factors. lost. A good point cloud filtering algorithm not only requires high real-time performance, but also preserves the feature information of the model well while denoising. It is necessary to thoroughly apply the noise point features of point cloud data in order to propose a better denoising algorithm.

Point cloud data is an unstructured data format. The point cloud data obtained by lidar scanning is affected by the distance between the object and the radar, and the distribution is uneven. Object point cloud data is sparsely distributed. In addition, point cloud data is characterized by disorder and asymmetry, which leads to the lack of a clear and unified data structure for point cloud data in data representation, which exacerbates the difficulty of subsequent point cloud segmentation and recognition processing. As an end-to-end network structure, the neural network often processes conventional input data, such as sequences, images, videos, and 3D data. It cannot directly process unordered data such as point sets. Convolution When operating and processing point cloud data, convolution directly discards the shape information of the point cloud, and only retains the sequence information of the point cloud.

Point cloud data classification; in TerraSolid software, point cloud data classification layers are as follows: 1) Default; 2) Ground; 3) vegetation; 4) Building; 5) Lowpoint; 6) Modelkeypoints. Among them, the Default layer is the point cloud data layer temporarily stored in the operation process; the Ground (ground point) mainly stores the real topography of the ground (artificially built embankments, earth embankments, ladder roads, naturally formed and large-scale pits and piles), and artificially constructed soil ridges, dams, dikes, sluices and other hydraulic structures connected to the ground are regarded as ground points; Vegetation (vegetation points) mainly store the ground Vegetation points, such as grassland, shrubs, bamboo forests, nurseries, young forests, gardens, and woodlands, are regarded as vegetation points; Building (building points) mainly store surface buildings, and houses and greenhouses are regarded as building points.

The optimal filter function is determined according to the terrain slope change. For a given height difference, as the distance between two points decreases, the possibility that the laser footpoint with a large elevation value belongs to the ground point becomes smaller.

The slope-based filtering algorithm has the characteristics of simple calculation and strong adaptability, but it needs to know the slope of the terrain in advance and determine the size of the window to be opened. The selected point must be compared with all other points to determine whether the point is a ground point or not. It is necessary to calculate the slope of each point in the entire data set, which will inevitably increase the amount of calculation and slow down the speed. The original point cloud data is divided into blocks according to the statistical characteristics of the terrain, and then each block is changed according to the slope. The filtering algorithm is used to process the ground point sets of each block of data, and finally the blocks are spliced according to the feature points of the overlapping area to obtain a complete set of ground points. In this way, different blocks get different filtering thresholds, which avoids the singleness of thresholds and reduces classification errors.

TIN is an important model to represent digital elevation, and it is often used to store the proximity relationship between spatial discrete points. For irregularly distributed elevation points, it can be formally described as an unordered point set on the plane. The common method to convert the point set into TIN is to construct the Delaunay triangulation network of the point set. The adjacent points in the Delaunay triangulation network are in the TIN Also adjacent in the model.

The filtering algorithm is to use the elevation mutation relationship of the adjacent point cloud of the ground object in the TIN model, and apply the parameters such as the height difference critical value condition and the number of adjacent points satisfying the condition to filter the ground object point;

The solution to the above problems needs to be in the actual point cloud filtering operation, according to the specific terrain and features and their spatial distribution characteristics to select the appropriate constraint parameter combination, and adopt a progressive filtering method to achieve satisfactory results. In quite a few cases, it is also necessary to divide the continuous ground into several blocks according to the law of ground undulation, and select the respective parameter combinations according to the characteristics of the blocks.

Mapping point cloud data, processing ground point models, and obtaining elevation information.

Example 3

Ground point filtering method application, including filtering algorithm based on mathematical morphology, filtering method based on slope, UAV Lidar point cloud filtering algorithm based on TIN, pseudo scan line algorithm, wavelet layering, etc., filtering algorithm based on mathematical morphology, In order to extract ground points when mathematical morphology theory is used in the UAV Lidar point cloud data filtering, the traditional mathematical morphology "open" operator is improved and processed according to the following process:

Discrete pit corrosion treatment. Traverse the UAV Lidar point cloud data, open a window of w×w size with any point as the center, compare the elevation of each point in the window, and take the minimum elevation value in the window as the elevation after corrosion;

Discrete point dilation processing. Traverse the UAV Lidar point cloud data again, and use the same size structural window to expand the corroded data. That is, open a window of w×w size with any point as the center. At this time, replace the original elevation value with the corroded elevation value, compare the elevations of each point in the window, and take the maximum elevation value in the window as the height after expansion;

Ground point extraction. Let Zp be the original elevation of point p, and t be the threshold value. At the end of each point expansion operation, a judgment is made whether the point is a ground point. If the absolute value of the difference between the inflated elevation value of point p and its original elevation value Zp is less than or equal to the threshold t, point p is considered to be a ground point, otherwise it is a non-ground point.

The filtering algorithm based on slope change; it is to determine the optimal filter function according to the terrain slope change. For a given height difference, as the distance between two points decreases, the possibility that the laser foot point with a large elevation value belongs to the ground point is just The smaller the value; the slope-based filtering algorithm has the characteristics of simple calculation and strong adaptability, but it needs to know the slope of the terrain and determine the size of the window in advance. The selected point must be compared with all other points to determine whether the point is the ground point, it is also necessary to calculate the slope of each point in the entire data set, which will inevitably increase the amount of calculation and slow down the speed. The filtering algorithm based on the slope change is processed to obtain the ground point set of each block of data, and finally the blocks are spliced according to the feature points of the overlapping area to obtain a complete ground point set. In this way, different blocks get different filtering thresholds, which avoids the singleness of thresholds and reduces classification errors.

Filtering based on Triangulated Irregular Network (TIN); TIN is an important model to represent digital elevation, and is often used to store the proximity relationship between spatial discrete points. For irregularly distributed elevation points, it can be formally described as an unordered point set on the plane. The common method to convert the point set into TIN is to construct the Delaunay triangulation network of the point set. The adjacent points in the Delaunay triangulation network are in the TIN Also adjacent in the model.

The filtering algorithm is to use the elevation mutation relationship of the adjacent point cloud of the ground object in the TIN model, and apply the parameters such as the height difference critical value condition and the number of adjacent points satisfying the condition to filter the ground object points.

Suppose a non-empty point cloud Pt_Cloud is given, and two threshold conditions of height difference (threshold_h) and number of adjacent points (threshold_vn) are given according to the distribution of regional terrain, buildings, vegetation, etc., and elevation changes, and Filtered and Unfiltered are defined Two arrays record filtered points and unfiltered points respectively.

The method based on the triangulated irregular network (TIN) is a method based on two-dimensional neighborhood search, and its calculation amount and algorithm complexity are relatively large. Generally speaking, due to the obvious elevation mutation between tall buildings and vegetation and its adjacent ground points, the algorithm can filter the elevation mutation objects better, but when filtering shrubs or low ground objects, there will be Excessive error. Mao Jianhua et al. used the TIN filter algorithm to filter the experimental data (data including buildings, roads, vegetation, dry land and reservoirs, etc.) located in undulating mountainous areas, and the results showed that vegetation, buildings, etc. The filtering effect of is better. In the obtained DEM, the terrain fluctuations are basically maintained, and the basic shape of the road can be clearly displayed, but the error is large in the dry land flat area.

The solution to the above problems needs to be in the actual point cloud filtering operation, according to the specific terrain and objects and their spatial distribution characteristics to select the appropriate combination of constraint parameters, and adopt a progressive filtering method to achieve satisfactory results. In quite a few cases, it is also necessary to divide the continuous ground into several blocks according to the law of ground undulation, and select the respective parameter combinations according to the characteristics of the blocks.

The pseudo-scan line method reorganizes the two-dimensional discretely distributed laser points on the horizontal plane into a data structure of a one-dimensional linear continuous distribution point sequence, which is called a pseudo-scan line. Wu Jianyao et al. proposed a filtering method based on pseudo-scanning lines based on the one-dimensional neighborhood search method based on pseudo-scanning lines.

Filtering based on pseudo-scanlines uses elevation mutation information to distinguish ground points from non-ground points. Its basic idea is: the height difference between two points is caused by the undulation of natural terrain and the height of ground objects. If the height difference between two adjacent points is larger, then the possibility that the height difference is caused by natural terrain is smaller, and it is more likely that the higher point is on the ground object and the lower point is on the ground, that is, : Suppose there are two adjacent laser foot points p ₁ and p ₂ , p ₁ is the ground point, and p ₂ is its adjacent point. If their height values h ₁ and h ₂ satisfy the condition: h ₂ -h ₁ ≤Δh _max ×d (Δh _max is the tolerance of height difference, d is the horizontal distance between them), then p ₂ is considered to be the ground point, otherwise p ₂ is considered to be a non-ground point.

Based on the pseudo scan line filtering algorithm, the two-dimensional filtering problem is simplified to one-dimensional filtering problem. The algorithm structure is simple, which effectively reduces the calculation amount of filtering and ensures the accuracy. At the same time, the algorithm only needs two filtering parameters, which is easier to realize automation. . However, because most of the two-dimensional filters in the local neighborhood assume that the lowest elevation point in the neighborhood is the ground point, when there are few ground points, this kind of filtering method often fails. However, the filtering algorithm based on pseudo-scanning lines can always ensure that each filtering window contains ground points, and can obtain a relatively small type of error and total error, and accurately extract topographic points. In flat areas, the effect of pseudo-scanline filtering is very good, and in areas with relatively steep terrain, its error is also controlled within a small range. However, in areas with steep slopes and dramatic elevation changes or when filtering large objects, in order to obtain reliable results, it is usually necessary to reduce the threshold value of the elevation and the size of the filtering window; in urban areas, in order to filter out all large buildings, It is necessary to increase the filtering window appropriately so that the size of the filtering window is not smaller than the maximum size of the building. At present, the selection of these two parameters cannot be fully automated, and the method needs to be further improved.

Based on the filtering method of multi-resolution direction prediction, the point cloud of UAV Lidar laser footpoints located on different ground objects will show different elevation differences, and the elevation mutation between adjacent laser footpoints can be used to distinguish ground points from non-ground points .

The idea of the direction prediction method: For a certain distance range, if the difference between the current point pi and all direction prediction values is greater than the maximum height difference tolerance under the distance condition, the point is a ground object point, otherwise it is a ground point.

Wavelet layering method; wavelet transform is introduced into UAV Lidar data filtering, and data filtering based on wavelet transform is realized. In general, the ground point of the point cloud data of UAV Lidar is the lowest point of elevation in the local area. Therefore, the original data can be divided by windows of a specific size, and then a point with the lowest elevation can be selected in each window to form a new data description. These ground points are networked to form a rough terrain surface. This rough terrain surface is then used as a reference surface for filtering in the next layer to get more ground points. It must be guaranteed that there is at least one ground point in each segmentation window, which requires the segmentation window to be large enough. A window slightly larger than the largest building area in the target area is used as the scale described by the top layer of data. Next, determine the window scale described by the first layer of data. Take the area of the smallest man-made building in the target area as the window scale described by the first layer of data.

Example 4

The wavelet layered filtering algorithm needs to be layered first, and then pass this rough domain of interest to the next layer as the initial value of the domain of interest in the current layer, thereby reducing the calculation time and improving the accuracy of the processing results, but each layer The result of the judgment is affected by the upper layer. If there is an error in the processing of the upper layer, this error will lead to an error in the judgment of the data point type of the next layer. In addition, the selection of the scale of the split window is also very important. The scale described by the top layer of data should be slightly larger than the largest building area in the target area, and the scale of the window described by the first layer of data should be selected from the smallest artificial The size of the area of the building. In addition, the wavelet layered filtering algorithm also needs to consider more carefully in the selection of data initial values and discrimination rules, and eliminate the gross errors in the data.

Vegetation and building classification method application, using echo information to classify point clouds. At present, methods for extracting vegetation and buildings based on LiDAR point clouds mainly include supervised learning classification methods and point cloud segmentation hierarchical extraction methods. The former mainly uses machine learning. The method uses feature samples to train and learn to generate a classifier to extract buildings. The performance of this type of method depends on the selection of samples, and the cost is relatively high. The latter is based on the auxiliary information provided by the LiDAR system or the three-dimensional topological relationship of the point cloud for feature classification.

Example 5

Ground point model processing, automatically classified ground point modeling (DEM), observation model for manual intervention. If there is an unreasonable triangulation, manually divide the unseparated points into ground points until no unreasonable triangulation appears. Adjust parameters or algorithms for areas with sudden changes in elevation, and re-classify small areas automatically;

Obtain the elevation information, separate the key points of the contour line on the finely classified ground point model, and use the software to automatically generate the contour line. Removed to generate contours at the scale required for the project. After the contour lines are exported, the key points of the model are exported to the ENZ format, and the contour lines and elevation points are edited with CASS software to complete the collection of elevation elements of the topographic map.

Example 6

Check the data source before the construction. Before the construction, use the Yun-5 aircraft to carry the UCXp-WA aerial camera to obtain the aerial photography of the application area. The acquisition time is December 2016. The resolution of the original aerial photograph is 0.20m. The aerial three-dimensional encryption extracts the digital elevation model data with a grid interval of 0.5m within the scope of the aerial photograph, and obtains the ground point (*. The digital elevation model above is quickly mosaiced.

Check the data source of the completion period. During the completion period of the project, use the Pegasus drone equipped with the SONYILCE-5100 aerial camera to obtain the aerial photography of the application area. The acquisition time is August 2021. The resolution of the original aerial image is 0.10m. The smart puzzle module of the horse drone steward can realize the rapid splicing of orthophotos through aerial triangulation processing, dem generation, and fast mosaic process.

Summarize the data sources and aggregate the data of the two.

Example 7

Create a project, use software, create a new project, set project parameters, and build a unified data platform for subsequent calculations.

Import point cloud data; right-click CreatePointClould on pointClouds in Prospector to create point cloud data.

Define the basic point cloud information: specify the name, point cloud style (Single), and point cloud layer (V-SITE-SCAN) in the information dialog box;

In the SourceData dialog box, SourceData (source data) select "create new point cloud database", select "LAS" under "point cloud file format", add data and assign the corresponding coordinate system. Under New Point Cloud Database, enter a name under Specify New Point Cloud Database. Among them, the settings of "point cloud database coordinate system" and "coordinate system of current drawing" should match.

Point cloud database coordinate system modification: This value is obtained from the point cloud source file, and the coordinate system can be changed after the data is imported into the point cloud database.

Coordinate System Modification of Current Drawing: This value is obtained from the Units and Zones tab of the Drawing Settings dialog.

Add point cloud data to surfaces, specify source data, validate point cloud parameters, and create point cloud objects,

In the Summary (Summary) dialog box, expand the collection in the Properties table and make sure the properties match those specified earlier. If the property values do not match, use the links on the left side of the dialog to return to the previous page. The status bar in the lower right corner will display the notification being processed. If there is a processing completed notification, it means that the point cloud database has been processed and the point cloud object has been created.

Click the "Point Cloud" tab, right click "AddPointstoSurface" to open the wizard window.

On the "SurfaceOptions" page, you can enter the name of the surface, select the style of the surface, and click Next. Here, you can also add point cloud points to existing surfaces in the current drawing.

On the "RegionOptions" (area options) page, you can choose to specify a region or customize the range that makes up the surface, and click "Next".

On the Summary page, expand the collection in the Properties table, making sure that the properties match those specified earlier. If the property values do not match, use the links on the left side of the dialog to return to previous pages. Displays surface contours and TINs in the drawing, and displays surface objects on the Prospector tab in Toolspace.

Select the surface, right-click - Object Viewer, you can view the 3D effect of the surface.

Design Surface Construction

①Create an empty surface-design surface

② Create a feature line from an object

Select the edge of the design site, create a feature line from the object, and specify the elevation of the feature line, where the elevation can be given as an estimate or based on the design elevation.

③ Add feature lines to the design surface

Select the feature line and add it to the design surface as a feature line. Other parameters can be defaulted. Click OK to add the feature line to the surface. Select the surface and you can see that the design surface is a plane in the object viewer.

④ Add boundaries to the design surface, select the element line, and fix the design range within the design red line Example 8

Result analysis

In the earthwork balance, it is necessary to comprehensively consider the overall earthwork balance results in the construction of the site, slope, road, etc., and the amount of earthwork for the leveling of the site occupies a relatively large proportion. This application compares the earth and rock volume calculated by the UAV Lidar point cloud in the project area with the earth and rock volume in the quarterly report (annual report) of soil and water conservation monitoring, and analyzes the actual application effect of the UAV Lidar data in the earth and rock volume estimation. During the calculation process, the calculation range is obtained from the water and soil conservation zoning map provided by the design institute, and the coordinates are unified.

In the earthwork balance, it is necessary to comprehensively consider the overall earthwork balance results in the construction of the site, slope, road, etc., and the amount of earthwork for the leveling of the site occupies a relatively large proportion. This application compares the earth and rock volume calculated by the UAV Lidar point cloud in the project area with the earth and rock volume in the quarterly report (annual report) of soil and water conservation monitoring, and analyzes the actual application effect of the UAV Lidar data in the earth and rock volume estimation. During the calculation process, the calculation range is obtained from the water and soil conservation zoning map provided by the design institute, and the coordinates are unified.

The basic principles, main features and advantages of the present invention have been shown and described above. Those skilled in the art should understand that the present invention is not limited by the above-mentioned embodiments, and those described in the above-mentioned embodiments and description are only preferred examples of the present invention, and are not intended to limit the present invention, without departing from the spirit and scope of the present invention. Under the premise, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the claimed invention. The protection scope of the present invention is defined by the appended claims and their equivalents.

Claims (7)

1. The high-density laser point cloud technology is used for the calculation method of excavating and filling earthwork in power grid engineering, which is characterized in that it includes the following steps: S1: The UAV is equipped with a laser radar to collect high-density laser point cloud data before and after the excavation and filling of the project; S2: After the point cloud is automatically classified, manual review is carried out, and the data is processed in the office; S3: Use GIS to delineate the calculation range; S4: Input the data into the data system, use the data before and after excavation to calculate the volume change, and calculate the earth and stone volume. 2. The high-density laser point cloud technology according to claim 1 is used for the earthwork calculation method of power grid engineering excavation and filling, characterized in that: said step S1 also includes the following steps: S101: UAV Lidar data processing removes noise points and transforms the coordinates of the data obtained by the airborne Lidar, then automatically classifies the point cloud, manually checks and corrects the classified results, and finally outputs the corresponding classification according to the requirements results. 3. The high-density laser point cloud technology according to claim 2 is used for power grid engineering excavation and filling earthwork calculation method, characterized in that: said step S101 also includes the following steps: S1011: Preprocessing point cloud data; S1012: point cloud noise removal; S1013: classification of point cloud data; S1014: Mapping the point cloud data, processing the ground point model, and obtaining elevation information. 4. The high-density laser point cloud technology according to claim 3 is used for power grid engineering excavation and filling earthwork calculation method, characterized in that: said step S1033 also includes the following steps: S10111: Filter the point cloud preprocessing information, including slope-based filtering based on mathematical morphology, TIN-based UAV Lidar point cloud filtering, pseudo scan line filtering, and wavelet layered filtering. 5. The high-density laser point cloud technology used in power grid engineering excavation and filling earthwork calculation method according to claim 4, characterized in that: said step S10111 also includes the following steps: S101111: Ground point model processing, modeling the automatically classified ground points. If there is an unreasonable triangulation, manually divide the unseparated points into ground points until no unreasonable triangulation appears. It can handle the areas with sudden changes in elevation; S101112: Obtain the elevation information, separate the key points of the contour line on the finely classified ground point model, and use the software to automatically generate the contour line. Will be removed to generate contours at the scale required for the project. After the contour lines are exported, the key points of the model are exported to the ENZ format, and the contour lines and elevation points are edited with CASS software to complete the collection of elevation elements of the topographic map. 6. The high-density laser point cloud technology used in power grid engineering excavation and filling earthwork calculation method according to claim 1, characterized in that: said step S2 also includes: S201: Check the data source before construction; S202: Check the data source of the completion period; S303: Summarize data sources. 7. The high-density laser point cloud technology according to claim 1 is used for power grid engineering excavation and filling earthwork calculation method, characterized in that: said step S4 also includes the following steps: S401: Create a project; S402: Import point cloud data; S403: Add point cloud data to the surface, specify source data, verify point cloud parameters, and create point cloud objects.

Images