A method for monitoring changes in the geometry of a vertical barrier wall at a contaminated site

Abstract

The application discloses a kind of contaminated site vertical barrier wall geometric shape change monitoring method, comprising: the basic data of contaminated site vertical barrier site is collected;Actual investigation is carried out to vertical barrier site;Basic data and actual investigation result are combined, and the flight line above vertical barrier site is planned to obtain;2D image data of site at vertical barrier wall is obtained using tilt measurement method;Several in-situ site phase control points are arranged around vertical barrier wall;Based on deep learning, 3D model is generated using 2D image data collected by unmanned aerial vehicle;Ground point data is obtained by combining 3D model, and rasterization is carried out on ground point data by natural neighbor algorithm to judge and monitor the area and size change of vertical barrier wall at different times before and after.The present application can solve the technical problems that the geometric shape change of vertical barrier wall cannot be monitored.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of vertical barrier wall, and particularly relates to a method for monitoring geometric shape change of a vertical barrier wall in a contaminated site. BACKGROUND

[0002] The vertical barrier wall in a contaminated site is widely used for pollution control, however, the geometric size of the vertical barrier wall often changes during service. Therefore, it is necessary to judge the service performance of the vertical barrier wall by the geometric size change thereof during service. However, the contaminated site often contains harmful gases to human body, which makes the staff in a dangerous state. Moreover, the geometric shape cannot be quantified by using the traditional measurement method. At present, there is no suitable method for detecting the geometric shape change of the vertical barrier wall. SUMMARY

[0003] The technical problem solved by the present application is that the present application provides a method for monitoring geometric shape change of a vertical barrier wall in a contaminated site, which can solve the technical problem that the geometric shape change of the vertical barrier wall cannot be monitored.

[0004] TECHNICAL SOLUTION

[0005] The present application provides a method for monitoring geometric shape change of a vertical barrier wall in a contaminated site, which comprises the following steps:

[0006] S1, collecting basic data of the vertical barrier wall in the contaminated site; actually surveying the vertical barrier wall in the contaminated site; and planning an air route above the vertical barrier wall in the contaminated site according to the basic data and the actual surveying result;

[0007] S2, controlling an unmanned aerial vehicle to fly along the planned air route, and acquiring 2D image data of the site at the vertical barrier wall by using an inclinometer method during the flight, wherein the 2D image data comprises one vertical 2D image and four inclined 2D images with five different angles;

[0008] S3, arranging a plurality of in-situ site phase control points around the vertical barrier wall to ensure that the four sides of the vertical barrier wall to be monitored are surrounded;

[0009] S4, generating a 3D model based on deep learning and using the 2D image data collected by the unmanned aerial vehicle; specifically, predicting the missing depth information and the object contour by using a deep learning algorithm, performing depth evaluation and prediction on all 2D image data based on a depth estimation method of binocular stereo matching of two picture cost volumes, and finally recovering the three-dimensional point cloud;

[0010] S5, obtaining ground point data based on the 3D model, and performing rasterization on the ground point data by using a natural neighbor algorithm to judge and monitor the area and size change of the vertical barrier wall at different times.

[0011] Further, in step S1, the basic data of the vertical barrier site of the contaminated site includes the proportioning of the vertical barrier wall material, the cracking condition of the vertical barrier wall surface and the size of the vertical barrier wall surface.

[0012] Further, in step S1, the actual survey of the vertical barrier site means that the inner and outer circumferences of the entire vertical barrier wall are respectively surveyed.

[0013] Further, in step S2, during the flight of the unmanned aerial vehicle, the Beidou system is used for real-time positioning, the laser three-dimensional system is used for real-time acquisition of point cloud information, the multispectral camera is used for image information acquisition, and the laser range finder is used for obtaining the height of the unmanned aerial vehicle relative to the base point.

[0014] Further, in step S4, the process of depth evaluation prediction on all 2D image data and final recovery of three-dimensional point cloud includes the following steps:

[0015] The depth extraction unit is used for perspective selection, and different 2D images are input as reference images: the neural network is used to extract the depth features Fi of 3D and output the feature map:

[0016] The feature map is subjected to cost accumulation, wherein the pixels in the cost accumulation are similar to the pixels in the reference image;

[0017] The feature map after cost accumulation is subjected to depth learning training of MVSNet and a 3D model is established.

[0018] Further, in step S5, the process of obtaining ground point data in combination with the 3D model includes the following steps:

[0019] Photogrammetry processing is performed using Erdas Imagine software, including internal / external orientation of stereo pairs, 3D point cloud extraction and DEM generation; for internal orientation, the ground point data contained in the original calibration certificate corresponding to the 2D image is used; for external orientation, the same position at different times is used as a ground control point, and after checking whether the position of the selected reference object changes over time, the ground control point is uniformly selected in the overlapping area of the photos;

[0020] After external orientation, the three-dimensional point cloud of the analysis area is extracted by automatic terrain extraction and the dense point matching module contained in Erdas Imagine.

[0021] Further, the monitoring method further includes the following steps:

[0022] S8, when the calculated vertical barrier geometric size change value exceeds the threshold value designed by the program, an alarm is displayed.

[0023] Further, when the alarm duration reaches a preset duration threshold, the alarm information is remotely sent to a designated user terminal.

[0024] Advantages:

[0025] The pollution site vertical barrier wall geometric shape change monitoring method has the advantages of high efficiency, wide monitoring range and safety and reliability. The pollution site vertical barrier wall geometric shape change monitoring method does not require measurement personnel to perform field measurement at the pollution site, and the tilt photography of the unmanned aerial vehicle can accurately measure and calculate the geometric shape change of the vertical barrier wall. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 The pollution site vertical barrier wall geometric shape change monitoring method flowchart of the embodiment of the application is shown in the figure.

[0027] Figure 2 The tilt photography process schematic diagram of the unmanned aerial vehicle is shown in the figure.

[0028] Figure 3 The vertical barrier wall geometric size change result schematic diagram is shown in the figure. DETAILED DESCRIPTION

[0029] The following embodiments enable a professional in the technical field to more comprehensively understand the application, but do not limit the application in any way.

[0030] The embodiment discloses a pollution site vertical barrier wall geometric shape change monitoring method, which comprises the following steps:

[0031] S1, collecting basic data of the pollution site vertical barrier site; performing actual survey on the vertical barrier site; combining the basic data and the actual survey result to plan an air route above the vertical barrier site;

[0032] S2, controlling the unmanned aerial vehicle to fly along the planned air route, and acquiring 2D image data of the site at the vertical barrier wall by using a tilt measurement method during the flight, the 2D image data comprising one vertical 2D image and four tilted 2D images at five different angles;

[0033] S3, arranging a plurality of in-situ site phase control points around the vertical barrier wall to ensure that the four sides of the vertical barrier wall to be monitored are surrounded;

[0034] S4, generating a 3D model based on deep learning and using the 2D image data collected by the unmanned aerial vehicle; specifically, using a deep learning algorithm to predict missing depth information and object contours, performing depth evaluation and prediction on all 2D image data based on a binocular stereo matching depth estimation method of two picture cost volume, and finally recovering three-dimensional point cloud;

[0035] S5, combined with the 3D model, the ground point data is rasterized by the natural neighbor algorithm to judge and monitor the area and size changes of the vertical barrier wall at different times.

[0036] In combination Figure 1 A geometric shape transformation of a vertical barrier wall of a contaminated site based on unmanned aerial vehicle oblique photography monitoring includes the following steps:

[0037] Step 1: vertical barrier site information collection;

[0038] The information collection includes the main material of the vertical barrier wall, the cracking condition exposed on the ground of the vertical barrier wall, and the basic range of the vertical barrier wall.

[0039] Step 2: on-site investigation of the vertical barrier site; the on-site investigation requires the investigator to circle the entire vertical barrier wall.

[0040] The main object of the investigation is the ground environment, and the basic geometric shape is observed to lay the foundation for the planning of the flight route above the vertical barrier site. Due to the surrounding characteristics of the vertical barrier wall, the entire vertical barrier wall needs to be investigated during the on-site investigation, and the inner and outer circumferences need to be investigated.

[0041] Step 3: planning of the flight route above the vertical barrier site:

[0042] The planning process of the flight route above the vertical barrier wall of the contaminated site mainly depends on the geometric shape of the range of the vertical barrier wall of the contaminated site, which is often after the on-site investigation. Since the geometric size of the vertical barrier wall site is often smaller than other engineering scenes, the flight height meets the requirements in most cases, so the accuracy of the oblique photography image is important. Figure 2 The oblique photography process of the unmanned aerial vehicle is shown in the figure.

[0043] Step 4: correction of the flight route of the unmanned aerial vehicle;

[0044] The Beidou system performs real-time positioning, the laser three-dimensional system collects point cloud information in real time, the multispectral camera collects image information, and the laser range finder obtains the height of the unmanned aerial vehicle relative to the base point.

[0045] Step 5: layout of in-situ site control points:

[0046] The layout of the in-situ site control points needs to ensure that the control points are distributed on the center line of the barrier wall and around it to ensure the accuracy of the geometric shape range data, and to ensure that the ground data points obtained before and after are covered.

[0047] Step 6: establish a 3D model and generate corresponding images;

[0048] The deep learning algorithm is used to predict the missing depth information and the object contour, so as to perform depth evaluation prediction on all 2D images, and finally restore the three-dimensional point cloud. The basic process and principle are as follows: first, a depth extraction unit is performed, that is, the view angle is selected first, and then different 2D images are input as references. Among them, the neural network is used to extract the depth feature Fi of 3D and output the feature map. And cost accumulation is carried out, in which the pixels in the cost accumulation are similar to the pixels in the reference image. Finally, the depth learning training of MVSNet is carried out and the 3D model is established.

[0049] Step 7: data acquisition and processing, geometric size change calculation. The obtained three-dimensional model is imported into the software for data acquisition at each stage to obtain ground point data; the natural neighbor algorithm is used for gridding to judge the area and size change before and after.

[0050] Specifically, the three-dimensional model is divided into blocks by corresponding algorithms to have corresponding spatial information. For example, Erdas Imagine software is used for photogrammetric processing, including internal / external orientation of stereo pairs, 3D point cloud extraction and DEM generation. For internal orientation, data contained in the original calibration certificate provided with the image is used. For external orientation, we use the same location at different times as ground control points. After checking whether the position of the selected reference object changes over time, the ground control points are evenly selected in the overlapping area of the photos. Then, they are manually fixed to the image. After external orientation, the three-dimensional point cloud of the analysis area is extracted by automatic terrain extraction and dense point matching module contained in Erdas Imagine. And the point cloud is manually filtered and imported into the software, and the natural neighbor algorithm is used for gridding to judge the area and size change before and after. Figure 3 The vertical barrier geometric size change result diagram.

[0051] Step 8: monitoring and early warning of vertical barrier geometric shape and area.

[0052] When the calculated degree of change of the geometric size of the vertical barrier exceeds the threshold value designed by the program, the computer will issue a warning. At the same time, the corresponding warning will be sent to the app of the mobile terminal of the operator.

[0053] The above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical solution falling within the scope of the present application is within the protection scope of the present application. It should be noted that, for ordinary skilled persons in the technical field, some improvements and refinements without departing from the principles of the present application should be considered as the protection scope of the present application.

Claims

1. A method for monitoring geometric changes in a vertical barrier wall at a contaminated site, characterized in that, The monitoring method includes the following steps: S1. Collect basic data on the vertical barrier site of the contaminated site; conduct an actual survey of the vertical barrier site; and plan the flight path above the vertical barrier site based on the basic data and the actual survey results. S2, control the drone to fly along the planned route. During the flight, use the tilt measurement method to obtain 2D image data of the site at the vertical barrier wall. The 2D image data includes one vertical and four tilted 2D images at five different angles. S3, set up several in-situ site phase control points around the vertical barrier wall to ensure that the geometry of the vertical barrier wall to be monitored is surrounded. S4, based on deep learning, generates 3D models using 2D image data collected by drones; specifically, it uses deep learning algorithms to predict missing depth information and object contours, uses a depth estimation method based on binocular stereo matching of cost volume of two images to evaluate and predict the depth of all 2D image data, and finally restores the 3D point cloud. S5 combines 3D models to obtain ground point data, and uses the natural nearest neighbor algorithm to rasterize the ground point data in order to judge and monitor the changes in area and size of the vertical barrier wall before and after different times.

2. The method for monitoring geometric changes of vertical barrier walls in contaminated sites according to claim 1, characterized in that, In step S1, the basic information of the vertical barrier site of the contaminated site includes the proportion of vertical barrier wall materials, the cracking condition of the vertical barrier wall surface, and the dimensions of the vertical barrier wall surface.

3. The method for monitoring geometric changes of vertical barrier walls in contaminated sites according to claim 1, characterized in that, In step S1, conducting an actual survey of the vertical barrier site means surveying the site along both the inner and outer perimeters of the entire vertical barrier wall.

4. The method for monitoring geometric changes of vertical barrier walls in contaminated sites according to claim 1, characterized in that, In step S2, during the flight of the UAV, the BeiDou system is used for real-time positioning, the laser 3D system collects point cloud information in real time, the multispectral camera collects image information, and the laser rangefinder obtains the altitude of the UAV relative to the base point.

5. The method for monitoring geometric changes of vertical barrier walls in contaminated sites according to claim 1, characterized in that, Step S4, which involves performing depth evaluation and prediction on all 2D image data and ultimately reconstructing the 3D point cloud, includes the following steps: A depth extraction unit is used for viewpoint selection, with different 2D images as reference images as input; a neural network is used to extract 3D depth features Fi and output a feature map; Cost accumulation is performed on the feature map, where the pixels in the cost accumulation are similar to the pixels in the reference image; The feature maps after cost accumulation are used to train MVSNet deep learning and build a 3D model.

6. The method for monitoring geometric changes of vertical barrier walls in contaminated sites according to claim 1, characterized in that, Step S5, the process of obtaining ground point data by combining the 3D model, includes the following steps: Photogrammetric processing was performed using Erdas Imagine software, including interior / exterior orientation of stereo pairs, 3D point cloud extraction, and DEM generation. For interior orientation, ground point data contained in the original calibration certificate corresponding to the 2D images was used. For exterior orientation, ground control points were used from the same location at different times, and ground control points were uniformly selected in the overlapping area of ​​the photographs after checking whether the position of the selected reference object changed over time. After external orientation, the 3D point cloud of the analysis area is extracted using automatic terrain extraction and the dense point matching module included in Erdas Imagine.

7. The method for monitoring geometric changes of vertical barrier walls in contaminated sites according to claim 1, characterized in that, The monitoring method further includes the following steps: S8: When the calculated change in vertical barrier geometry exceeds the threshold value designed in the program, an alarm is displayed.

8. The method for monitoring geometric changes of vertical barrier walls in contaminated sites according to claim 7, characterized in that, Once the alarm duration reaches the preset duration threshold, the alarm information will be remotely sent to the designated user terminal.

Images