A corrugated pipe positioning detection method and device and electronic equipment

Abstract

The application discloses a corrugated pipe positioning detection method and device and electronic equipment, and relates to the technical field of data processing. The application adopts a well-established training data set to train a neural network model to obtain a prediction model, inputs point cloud data including color information and coordinate information after preprocessing into the prediction model to obtain segmented point cloud data, then, performs visual processing on the segmented point cloud data to obtain a visual result, corrects the segmented point cloud data based on the corrugated pipe color in the visual result, so that the segmentation accuracy of the corrugated pipe is improved, and finally, performs algorithm fitting on the segmented point cloud data after correction to obtain height data of the corrugated pipe, so that the purpose of improving the positioning accuracy of the corrugated pipe is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a corrugated pipe positioning detection method and device and electronic equipment. BACKGROUND

[0002] The installation and positioning of prestressed corrugated pipes is a key link in bridge construction, and the effective stress of the prefabricated beam is directly related to the positioning accuracy of the prestressed pipe. At present, the installation and positioning of prestressed corrugated pipes in China still has the following problems.

[0003] Firstly, the positioning data of the corrugated pipe is generally obtained through two-dimensional drawings, but the prestressed corrugated pipe is a spatial curve with complex horizontal and vertical bending changes in three dimensions, and the traditional two-dimensional design drawing cannot accurately express the positioning coordinates of the pipe.

[0004] Secondly, the traditional positioning method basically relies on manual pipe positioning, but in actual construction, due to the large number of prestressed corrugated pipe positioning points and the difficulty of positioning, there are problems such as repeated operation, large positioning error, low work efficiency, etc., which makes it difficult to ensure the positioning accuracy of the prestressed corrugated pipe.

[0005] Thirdly, the traditional detection method is to use a tape measure and a level to sample and measure some points of the prestressed corrugated pipe, but due to the large number of longitudinal and transverse steel bars in the beam body, the worker's detection space is narrow, the measurement is difficult, and the measurement results obtained also have a large deviation from the design values.

[0006] Fourthly, the existing technologies applied to the inspection of prestressed corrugated pipes are mostly in the design stage, and the collision detection of drawings and the height output of prestressed corrugated pipes are realized through BIM technology, such as the Tekla software + Navisworks technology disclosed in Chinese patent CN112347549A, the Revit + CAD technology disclosed in Chinese patent CN104715117A. Or in the real scene, a variety of measuring instruments are used comprehensively, such as the technology disclosed in Chinese patent CN109612429A, total station + manual or software coordinate calculation, the device of laser sensor + range finder disclosed in Chinese patent CN110514115A. The method of combining BIM + three-dimensional laser scanner + reconstructed model + deviation comparison in Chinese patent CN113221221A pays attention to using BIM technology and point cloud reconstruction to obtain height deviation, but does not directly segment and extract the real point cloud obtained by scanning;

[0007] Based on the above description, most of the prior art is still mainly artificial detection (measuring instrument), and lacks intelligence. Although BIM+three-dimensional laser scanning technology uses three-dimensional modeling and reconstruction technology, it needs to delete a large amount of non-corrugated pipe data before reconstruction, and does not use deep learning technology for more generalized and more intelligent application.

[0008] Even if the deep learning technology is used to locate the corrugated pipe, without considering the color characteristics of the corrugated pipe and the complex construction environment, there are still problems such as inaccurate semantic segmentation of the corrugated pipe, and thus accurate positioning of the corrugated pipe cannot be achieved. SUMMARY

[0009] To solve the above problems existing in the prior art, the present application provides a corrugated pipe positioning detection method, device and electronic equipment.

[0010] To achieve the above purpose, the present application provides the following solutions:

[0011] A corrugated pipe positioning detection method, comprising:

[0012] Establishing a training data set;

[0013] Training a neural network model using the training data set to obtain a prediction model;

[0014] Obtaining point cloud data of each measuring station, and preprocessing the point cloud data; the point cloud data includes color information and coordinate information;

[0015] Inputting the preprocessed point cloud data into the prediction model to obtain segmented point cloud data;

[0016] Visualizing the segmented point cloud data to obtain a visualization result;

[0017] Correcting the segmented point cloud data based on the corrugated pipe color in the visualization result;

[0018] Algorithm fitting the corrected segmented point cloud data to obtain height data of the corrugated pipe.

[0019] Optionally, the training data set is established, specifically comprising:

[0020] Establishing a whole model of a steel corrugated pipe;

[0021] Converting the whole model of the steel corrugated pipe into point cloud data to obtain a point cloud file;

[0022] Data enhancement processing the point cloud data in the point cloud file to obtain a virtual data set;

[0023] adding different RGB colors to the point cloud data in the virtual data set to obtain a point cloud virtual data set;

[0024] obtaining real point cloud data of the bellows, and preprocessing the real point cloud data to obtain a real point cloud set;

[0025] adding a bellows label to the real point cloud set and the point cloud virtual data set respectively;

[0026] performing a hybrid splicing operation on the real point cloud set and the point cloud virtual data set to which the bellows label is added to obtain the training data set.

[0027] Optionally, the neural network model is constructed based on a RandLA-Net network.

[0028] Optionally, the segmented point cloud data is corrected based on the color of the bellows in the visualization result, and the correction specifically includes:

[0029] statistically counting the RGB value of each point cloud in the segmented point cloud data according to the point cloud sequence number, and generating a scatter plot based on the RGB value;

[0030] obtaining a color average threshold of the bellows, and obtaining a threshold range based on the color average threshold;

[0031] when the RGB value is outside the threshold range, determining that the point cloud corresponding to the RGB value is an abnormal segmented point;

[0032] locating the three-dimensional coordinate position of the abnormal segmented point based on the point cloud sequence number of the abnormal segmented point;

[0033] generating a point cloud density histogram with the long axis coordinate of the bellows as the horizontal axis based on the three-dimensional coordinate position;

[0034] based on the point cloud density histogram, regarding a region with a point cloud density lower than a preset value as an occlusion region;

[0035] adopting a three-dimensional rectangular frame to frame the occlusion region, reading the three-dimensional coordinates of the three-dimensional rectangular frame, and determining the center point of the three-dimensional rectangular frame;

[0036] constructing a spherical region with the center point as the center and a preset distance as the radius;

[0037] re-scanning the point cloud data in the spherical region, and returning to perform "statistically counting the RGB value of each point cloud in the segmented point cloud data according to the point cloud sequence number, and generating a scatter plot based on the RGB value" until the RGB value of all point clouds is within the threshold range, thereby completing the correction of the segmented point cloud data.

[0038] Optionally, the preset distance is half of the distance between the center point and a specific measuring point, and the specific measuring point is the closest measuring point to the center point among all the measuring points.

[0039] Optionally, the segmented point cloud data after correction is subjected to algorithm fitting to obtain height data of the corrugated pipe, and the height data specifically comprises:

[0040] Color information and coordinate information of the segmented point cloud data after correction are obtained, and a boundary line in the long axis direction of the corrugated pipe is obtained through an edge extraction algorithm, and a center line of the corrugated pipe is obtained by combining a center line fitting method.

[0041] The height value of the corrugated pipe is determined according to the boundary line and the center line, and the height data is obtained.

[0042] According to the specific embodiments provided by the present application, the following technical effects are provided:

[0043] The corrugated pipe positioning detection method provided by the present application can improve the segmentation accuracy of the corrugated pipe and further improve the positioning accuracy of the corrugated pipe on the basis of point cloud data including color information and coordinate information in combination with a deep learning method.

[0044] In addition, in order to implement the corrugated pipe positioning detection method provided above, the present application further provides the following device or equipment:

[0045] A corrugated pipe positioning detection device comprises an unmanned vehicle, a fixing module, a three-dimensional laser scanner, a data processing system and a communication module.

[0046] The three-dimensional laser scanner is fixed on the unmanned vehicle through the fixing module, and the three-dimensional laser scanner and the communication module are electrically connected with the data processing system.

[0047] The three-dimensional laser scanner is used to obtain real point cloud data of the corrugated pipe, the data processing system is implanted with a software program, and the software program is used to implement the corrugated pipe positioning detection method provided above based on the real point cloud data.

[0048] Optionally, the data processing system comprises:

[0049] A data set construction module is used to construct a training data set.

[0050] A model training module is used to train a neural network model to obtain a prediction model by using the training data set.

[0051] A preprocessing module is used to obtain point cloud data of each measuring point and to pre-process the point cloud data, and the point cloud data comprises color information and coordinate information.

[0052] a data segmentation module, configured to input the preprocessed point cloud data into the prediction model to obtain segmented point cloud data;

[0053] a visualization processing module, configured to perform visualization processing on the segmented point cloud data to obtain a visualization result;

[0054] a data correction module, configured to correct the segmented point cloud data based on a bellows color in the visualization result;

[0055] a height data determination module, configured to perform algorithm fitting on the corrected segmented point cloud data to obtain height data of the bellows.

[0056] Optionally, the method further comprises: a GPS positioning system, a terminal server, an unmanned vehicle remote control device, and a three-dimensional laser scanner control device.

[0057] The GPS positioning system and the terminal server are connected to the data processing system through the communication module.

[0058] The unmanned vehicle remote control device is wirelessly connected to the unmanned vehicle, and the three-dimensional laser scanner control device is wiredly or wirelessly connected to the three-dimensional laser scanner.

[0059] An electronic device comprises:

[0060] a memory, configured to store a software program;

[0061] a processor, connected to the memory, configured to call and execute the software program to implement the bellows positioning detection method provided above.

[0062] The technical effects achieved by the above device or equipment provided by the present application are the same as those achieved by the bellows positioning detection method provided by the present application, and thus will not be described here. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0064] Figure 1 The flowchart of the bellows positioning detection method provided by the present application. DETAILED DESCRIPTION

[0065] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0066] The present application aims to provide a bellows positioning detection method and device and electronic equipment, which can improve the segmentation accuracy of the bellows and further improve the positioning accuracy of the bellows.

[0067] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0068] As shown in the drawings, the present application provides a bellows positioning detection method, which comprises the following steps: Figure 1

[0069] Step 100: Establish a training data set. The implementation process of this step can be as follows:

[0070] A steel reinforced bellows overall model similar to the actual working environment of the bellows is established in batches through Revit software and its plug-in Dynamo.

[0071] The steel reinforced bellows overall model is converted into point cloud data to obtain a point cloud file.

[0072] In order to simulate the real environment, the point cloud data in the point cloud file is subjected to data enhancement processing to obtain a virtual data set. For example, noise is added to the point cloud in the point cloud data in the point cloud file, and data enhancement is performed through random flipping, down-sampling, scaling, etc.

[0073] Different RGB colors are added to the point cloud data in the virtual data set to obtain a point cloud virtual data set.

[0074] Real point cloud data of the bellows is obtained, and the real point cloud data is preprocessed to obtain a real point cloud set. The preprocessing method can be registration, denoising, thinning, etc.

[0075] Bellows labels are added to the real point cloud set and the point cloud virtual data set, respectively.

[0076] The real point cloud set and the point cloud virtual data set with the added bellows labels are subjected to a hybrid splicing operation to obtain a training data set.

[0077] ​Step 101: train the neural network model to obtain a prediction model using a training data set. In the present application, the color of the corrugated pipe is mostly silver, which is quite different from the color of the steel bar, so the color of the point cloud is also trained. For example, the RandLA-Net network is input with six-dimensional feature channel data for training, and the dimensions are X, Y, Z, R, G, and B. Finally, the prediction model is obtained.

[0078] Step 102: obtain the point cloud data of each measuring station and pre-process the point cloud data. The point cloud data includes color information and coordinate information. For example, the point cloud data with color information and coordinate information obtained at each measuring station is pre-processed by point cloud registration, denoising, and thinning. The point cloud registration can be realized by the scanner supporting software. The measuring station is designed after field investigation.

[0079] Step 103: input the pre-processed point cloud data into the prediction model to obtain segmented point cloud data.

[0080] Step 104: visualize the segmented point cloud data to obtain a visualization result. The point cloud is required to display the true color in the visualization.

[0081] Step 105: correct the segmented point cloud data based on the color of the corrugated pipe in the visualization result. In this step, considering the interference of a large number of hoops, longitudinal steel bars, positioning steel bars, and reinforcing bars during scanning of the corrugated pipe, and the narrow and complex overall construction environment, there is a possibility of semantic segmentation error and deficiency of the corrugated pipe. Based on this, the implementation process of this step can be:

[0082] Analyze the color of the corrugated pipe in the visualization, count the RGB values of the segmented point cloud according to the point cloud serial number, and represent them in a scatter plot. Since the segmentation target is only the corrugated pipe, the corresponding RGB value is relatively uniform. The color average threshold of the corrugated pipe is set in advance, and the scatter points outside a certain threshold range are classified as abnormal segmentation points. By reading the serial number of the point cloud, the three-dimensional coordinate position of the abnormal segmentation point can be located. Taking the long axis coordinate of the corrugated pipe as the horizontal axis, a point cloud density histogram of the same interval in this direction is drawn, and the area with sparse point cloud density is classified as an occluded area. The long axis coordinate can quickly locate the position where the model is discontinuous or the scanning is defective.

[0083] The area of the model discontinuity and the non-corrugated pipe color is framed with the smallest three-dimensional box, the three-dimensional coordinates of the six points of the rectangular frame are read, and the center point of the three-dimensional rectangular frame is calculated. Through the calculation of the relative coordinates of the point cloud and the GPS geodetic coordinate system, the nearest survey station is found, the distance between the survey station and the center point of the frame is calculated, and the spherical area for re-scanning is taken with the center point as the center and half the distance between the two points as the radius. According to the re-issued survey range and the scanning instruction, the position adjustment of the three-dimensional laser scanner unmanned vehicle is carried out. For the point cloud generated by re-scanning, the above operation is repeated, and when the segmented corrugated pipe is correct, the point cloud data is spliced and fused with the previous point cloud data, the above operation is repeated, and the final complete and correct point cloud data (i.e. the corrected segmented point cloud data) is obtained.

[0084] Step 106: The height data of the corrugated pipe is obtained by algorithm fitting based on the corrected segmented point cloud data. For example, according to the obtained results with color information and coordinate values, the boundary line in the long axis direction of the corrugated pipe is obtained by edge extraction algorithm, and the center line of the corrugated pipe is obtained by center line fitting method, so as to obtain the height value of the corrugated pipe at any point in the long axis direction.

[0085] In the above implementation process, the color characteristics of the corrugated pipe and the complex construction environment are considered because the colors of the prestressed corrugated pipe, the reinforcing steel and the positioning jig frame are quite different, and the color characteristics can be used to improve the accuracy of semantic segmentation. In the process of using the detection instrument, the complex construction environment will have certain influence on the manual detection space and reading accuracy. For the detection scene of the prestressed corrugated pipe in the prefabrication plant, the deep learning technology must fully consider the problems that may occur in the real scene, such as the shielding interference of the closely arranged reinforcing steel to the acquisition of the corrugated pipe data by the three-dimensional laser scanner, which will affect the subsequent detection accuracy.

[0086] In order to implement the above-mentioned corrugated pipe positioning detection method, the present application further provides a corrugated pipe positioning detection device. The device comprises: an unmanned vehicle, a fixing module, a three-dimensional laser scanner, a data processing system and a communication module.

[0087] The three-dimensional laser scanner is fixed on the unmanned vehicle through the fixing module. The three-dimensional laser scanner and the communication module are electrically connected with the data processing system.

[0088] The three-dimensional laser scanner is used to acquire real point cloud data of the corrugated pipe. The software program is implanted in the data processing system. The software program is used to implement the above-mentioned corrugated pipe positioning detection method based on the real point cloud data.

[0089] In the present application, the communication module can be a wireless communication network.

[0090] Further, the above-mentioned data processing system can include:

[0091] a data set construction module configured to establish a training data set.

[0092] a model training module configured to train a neural network model using the training data set to obtain a prediction model.

[0093] a preprocessing module configured to obtain point cloud data of each measuring station and to preprocess the point cloud data, wherein the point cloud data comprises color information and coordinate information.

[0094] a data segmentation module configured to input the preprocessed point cloud data into the prediction model to obtain segmented point cloud data.

[0095] a visualization processing module configured to perform visualization processing on the segmented point cloud data to obtain a visualization result.

[0096] a data correction module configured to correct the segmented point cloud data based on a bellows color in the visualization result.

[0097] a height data determination module configured to perform algorithm fitting on the corrected segmented point cloud data to obtain height data of the bellows.

[0098] Furthermore, the bellows positioning detection device provided by the above-mentioned application can further comprise a GPS positioning system, a battery, a terminal server, an unmanned vehicle remote control device and a three-dimensional laser scanner control device.

[0099] The GPS positioning system and the terminal server are connected to the data processing system through a communication module.

[0100] The unmanned vehicle remote control device is wirelessly connected to the unmanned vehicle, and the three-dimensional laser scanner control device is wiredly or wirelessly connected to the three-dimensional laser scanner.

[0101] The GPS positioning system records the absolute coordinates of the measuring station. The battery provides the data processing system and the communication module with endurance capability. The terminal server generally refers to a user computer. The unmanned vehicle remote control device and the three-dimensional laser scanner control device generally refer to iPads corresponding to different control software.

[0102] Based on this, the specific implementation principle of the bellows positioning detection device provided by the application is as follows:

[0103] First, the RandLA-Net neural network applied to the bellows semantic segmentation application in the data processing system is trained, and the training steps are as follows:

[0104] The steel reinforced corrugated pipe model similar to the actual working environment of the corrugated pipe is established by Revit software and its plug-in Dynamo, and the model is converted into a point cloud file. In order to simulate the real environment, noise is added to the point cloud in the virtual data set, and the data is enhanced by random flipping, downsampling, scaling and other methods.

[0105] The point cloud data obtained by real scanning has coordinate geometric information and color information. The scanned data is preprocessed by point cloud registration, denoising and thinning to form a real point cloud set.

[0106] The corrugated pipe label is added to the virtual data set and the real point cloud set, and different RGB colors are added to different point cloud targets in the virtual data set, and saved as a complete point cloud virtual data set. The two are spliced to form a training data set.

[0107] Considering that the color of the corrugated pipe is mostly silver iron pipe, which is quite different from the color of the steel bar, the color of the point cloud can also be trained. The six-dimensional feature channel data is input into the RandLA-Net network for training, and the dimensions are X, Y, Z, R, G, and B. Finally, the prediction model is obtained.

[0108] Secondly, through field investigation, the measurement points to be scanned are designed. During the scanning process, the three laser scanners are connected to the unmanned vehicle through the fixing module. Since the inside of the prestressed box girder and the bottom plate are all steel bars, it is not convenient for manual operation. Through the remote control equipment of the unmanned vehicle, the unmanned vehicle can quickly realize the shooting and space scanning of each measurement point.

[0109] Through the control equipment of the three-dimensional laser scanner, the scanning parameters are selected, the scanning is started, and the data is checked. During the scanning process, through the wireless communication network connection, the three-dimensional laser scanner transmits the point cloud data with color information obtained at each measurement point to the data processing system, and the GPS positioning system sends the position of each measurement point to the data processing system through the wireless communication network. In the data system, the point cloud data is automatically and real-time registered by the scanner supporting software. For the complete point cloud data that has been registered, the point cloud denoising and thinning operations are continued in the data processing system.

[0110] Then, the preprocessed point cloud data is put into the prediction model for semantic segmentation, and the segmented point cloud is visualized. In the visualization, the point cloud is required to display the real color.

[0111] Then, considering the interference of a large number of hoops, longitudinal steel bars, positioning steel bars and reinforcing bars of the corrugated pipe during scanning, and the narrow and complex overall construction environment, the corrugated pipe has the possibility of semantic segmentation error and deficiency. The shape and color of the corrugated pipe visualized by the data processing system are used to frame the discontinuous and non-corrugated pipe color area of the model with the smallest three-dimensional box, the three-dimensional coordinates of the six points of the rectangular box are read, and the center point of the three-dimensional rectangular box is calculated. Through the calculation of the relative coordinates of the point cloud and the GPS geodetic coordinates, the nearest survey station is found, the distance between the survey station and the center point of the box is calculated, and the spherical area of the re-scanning is taken as the center point and the half of the distance between two points as the radius. Through the wireless communication network, the user's remote control device sends a prompt, and the user adjusts the position of the device according to the survey station range of the re-scanning. The above operation is repeated for the point cloud generated by the re-scanning, and when the segmented corrugated pipe is correct, the point cloud data is spliced and fused with the previous point cloud data, the above operation is repeated, the complete and correct point cloud data is obtained, and the final result with color information and coordinate value is transmitted to the terminal server.

[0112] Finally, the data processing system obtains the boundary line of the long axis direction of the corrugated pipe through the edge extraction algorithm, and obtains the center line of the corrugated pipe by combining the center line fitting method, so as to obtain the height value of the corrugated pipe at any point in the long axis direction. The data processing system uploads the data to the terminal server for storage and analysis, and is convenient for users to view.

[0113] In addition, the present application also provides an electronic device, comprising:

[0114] A memory is configured to store a software program.

[0115] A processor is connected with the memory and is configured to call and execute the software program to implement the corrugated pipe positioning detection method provided above.

[0116] When the computer program in the memory is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk, and various media that can store program codes.

[0117] The various embodiments described in this specification are presented for the purpose of illustrating the principles of the present application and its best mode of operation. Each of the embodiments described in this specification has been provided for the purpose of illustration only and the various embodiments are not intended to limit the present application in any way unless otherwise specifically indicated. The same parts and / or features of the various embodiments described in this specification can be referenced using the same reference numerals.

[0118] The principles and implementations of the present application have been described above with the specific examples. The above description of the embodiments is only for the purpose of helping to understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, the specific implementation and application range of the present application can be changed according to the idea of the present application. In summary, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method of detecting the positioning of a bellows, characterized by, The method comprises the following steps: establishing a training data set; training a neural network model based on the training data set to obtain a prediction model; obtaining point cloud data of each measuring station, and preprocessing the point cloud data; the point cloud data comprises color information and coordinate information; inputting the preprocessed point cloud data into the prediction model to obtain segmented point cloud data; performing visual processing on the segmented point cloud data to obtain a visual result; correcting the segmented point cloud data based on the color of the corrugated pipe in the visual result; performing algorithm fitting on the corrected segmented point cloud data to obtain height data of the corrugated pipe; wherein the step of correcting the segmented point cloud data based on the color of the corrugated pipe in the visual result comprises: statistically counting the RGB value of each point cloud in the segmented point cloud data according to the point cloud sequence number, and generating a scatter plot based on the RGB value; obtaining a color average threshold of the corrugated pipe, and obtaining a threshold range based on the color average threshold; when the RGB value is outside the threshold range, determining that the point cloud corresponding to the RGB value is an abnormal segmented point; locating the three-dimensional coordinate position of the abnormal segmented point based on the point cloud sequence number of the abnormal segmented point; generating a point cloud density histogram with the long axis coordinate of the corrugated pipe as the horizontal axis based on the three-dimensional coordinate position; based on the point cloud density histogram, regarding the area with a point cloud density lower than a preset value as an occlusion area; using a three-dimensional rectangular frame to frame the occlusion area, reading the three-dimensional coordinates of the three-dimensional rectangular frame, and determining the center point of the three-dimensional rectangular frame; constructing a spherical region with the center point as the center and a preset distance as the radius; re-scanning the point cloud data in the spherical region, and returning to execute "statistically counting the RGB value of each point cloud in the segmented point cloud data according to the point cloud sequence number, and generating a scatter plot based on the RGB value" until the RGB value of all point clouds is within the threshold range, thereby completing the correction of the segmented point cloud data.

2. The bellows position detection method according to claim 1, characterized by, The method for establishing a training data set comprises the following steps: establishing a whole model of a steel corrugated pipe; converting the whole model of the steel corrugated pipe into point cloud data to obtain a point cloud file; performing data enhancement processing on the point cloud data in the point cloud file to obtain a virtual data set; adding different RGB colors to the point cloud data in the virtual data set to obtain a point cloud virtual data set; obtaining real point cloud data of the corrugated pipe, and preprocessing the real point cloud data to obtain a real point cloud set; adding a corrugated pipe label to the real point cloud set and the point cloud virtual data set respectively; performing a hybrid splicing operation on the real point cloud set and the point cloud virtual data set to which the corrugated pipe label is added, thereby obtaining the training data set.

3. The bellows position detection method according to claim 1, characterized by, The neural network model is constructed based on a RandLA-Net network.

4. The bellows position detection method according to claim 1, characterized by, The preset distance is half of the distance between the center point and a specific measuring station; the specific measuring station is the nearest measuring station to the center point among all measuring stations.

5. The bellows position detection method according to claim 1, wherein The method for performing algorithm fitting on the corrected segmented point cloud data to obtain height data of the corrugated pipe comprises the following steps: Color information and coordinate information of the corrected segmented point cloud data are acquired, and a boundary line in a long axis direction of the corrugated pipe is obtained through an edge extraction algorithm, and a center line of the corrugated pipe is obtained by a center line fitting method; A height value of the corrugated pipe is determined according to the boundary line and the center line, and the height data is obtained.

6. A bellow positioning detection device characterized by, It comprises: An unmanned vehicle, a fixed module, a three-dimensional laser scanner, a data processing system and a communication module; The three-dimensional laser scanner is fixed on the unmanned vehicle through the fixed module; the three-dimensional laser scanner and the communication module are electrically connected with the data processing system; The three-dimensional laser scanner is used to acquire real point cloud data of the corrugated pipe; the data processing system is implanted with a software program; the software program is used to implement the corrugated pipe positioning detection method according to any one of claims 1-5 based on the real point cloud data.

7. The bellows position sensing device of claim 6, wherein, The data processing system comprises: A data set construction module for establishing a training data set; A model training module for training a neural network model to obtain a prediction model using the training data set; A preprocessing module for acquiring point cloud data of each measuring station and preprocessing the point cloud data; the point cloud data comprises color information and coordinate information; A data segmentation module for inputting the preprocessed point cloud data into the prediction model to obtain segmented point cloud data; A visualization processing module for visualizing the segmented point cloud data to obtain a visualization result; A data correction module for correcting the segmented point cloud data based on the corrugated pipe color in the visualization result; A height data determination module for algorithm fitting of the corrected segmented point cloud data to obtain height data of the corrugated pipe.

8. The bellows position sensing device of claim 6, wherein, It also comprises: A GPS positioning system, a terminal server, an unmanned vehicle remote control device, a three-dimensional laser scanner control device; The GPS positioning system and the terminal server are connected with the data processing system through the communication module; The unmanned vehicle remote control device is wirelessly connected with the unmanned vehicle; the three-dimensional laser scanner control device is wired or wirelessly connected with the three-dimensional laser scanner.

9. An electronic device, comprising: It comprises: A memory for storing a software program; A processor connected with the memory for calling and executing the software program to implement the corrugated pipe positioning detection method according to any one of claims 1-5.

Images