CN115562284A - Method for realizing automatic inspection by high-speed rail box girder inspection robot - Google Patents

Abstract

The invention discloses a method for realizing automatic inspection by a high-speed rail box girder inspection robot, comprising the following steps: collecting surrounding environment data and robot absolute position information, and generating an environment map; combining the robot absolute position information with the environment map, based on The path planning algorithm obtains the inspection path of the robot; controls the robot to move, and takes a global shot of the inside of the high-speed rail box girder in real time to obtain the image information of the high-speed rail box girder; builds the initial HED edge detection model, trains the initial HED edge detection model, and obtains Target HED edge detection model; input the real-time collected high-speed rail box girder image information into the target HED edge detection model for processing, obtain the target defect image of the high-speed rail box girder, and realize the defect detection of the high-speed rail box girder. The method provided by the invention for realizing the automatic patrol inspection of the high-speed rail box girder has fast detection speed and high efficiency, and can prevent workers from flaw detection in a high-risk working environment.

Description

A method for automatic inspection by a high-speed rail box girder inspection robot

technical field

The invention belongs to the field of defect inspection of high-speed rail box girders, and in particular relates to a method for automatic inspection of high-speed rail box girder inspection robots.

Background technique

In my country's economic development, high-speed rail occupies a relatively important position, and its quality determines the operation safety of high-speed rail and smooth roads. As the key component of the high-speed rail bridge, its box girder bears the train load transmitted by the high-speed rail track. Under the combined effect of various unfavorable factors, such as traffic load, gradual degradation of building materials, and the influence of different environments, etc., it will cause certain damage to the overall structure and internal structure of the high-speed rail box girder. In extreme cases, fatigue cracking will occur, seriously Threats to the safety of high-speed rail operations.

Domestic inspections of high-speed rail box girders are mostly manual inspections, which have low detection efficiency, poor detection accuracy, and high missed detection rate, and the staff have been working in high-risk, harsh, and cruel working environments for a long time, which is not conducive to the staff physical and mental health. However, the existing automatic inspection methods for box girders in colleges and universities cannot adapt to the complex environment in high-speed rail box girders, and the detection accuracy is low and the utilization rate is not high. Therefore, it is an urgent problem to be solved to realize the rapid and automatic detection of the high-speed rail box girder and provide support for the subsequent maintenance guarantee.

Contents of the invention

The purpose of the present invention is to provide a method for realizing automatic inspection by a high-speed rail box girder inspection robot, so as to solve the above-mentioned problems in the prior art.

In order to achieve the above object, the present invention provides a method for automatic inspection of a high-speed rail box girder inspection robot, comprising the following steps:

Collect surrounding environment data and robot absolute position information, and generate an environment map based on the surrounding environment data;

Combining the absolute position information of the robot with the environment map, and obtaining the inspection path of the robot based on a path planning algorithm;

Based on the inspection path of the robot, the robot is controlled to move, and the inside of the high-speed rail box girder is globally photographed in real time to obtain image information of the high-speed rail box girder;

Build an initial HED edge detection model, and train the initial HED edge detection model based on the high-speed rail box girder image information to obtain a target HED edge detection model;

The image information of the high-speed rail box girder collected in real time is input into the target HED edge detection model for processing, and the target defect image of the high-speed rail box girder is obtained to realize the defect detection of the high-speed rail box girder.

Optionally, the inspection robot is equipped with a NB-IOT sensor, a panoramic camera, a laser radar, and a synchronization controller; the panoramic camera includes a stereo image sensor; and the laser radar includes a laser scanner and a laser scanning sensor.

Optionally, the process of collecting the surrounding environment data and the absolute position information of the robot includes,

The NB-IOT sensor establishes a connection with the NB-IOT base station set up in the surrounding environment to obtain the position information of the inspection robot; the panoramic camera and the laser radar scan and measure the surroundings of the inspection robot, Obtain relative three-dimensional point cloud data of the surrounding environment; register the position information with the relative three-dimensional point cloud data to obtain absolute three-dimensional point cloud data, and then obtain absolute position information of the robot.

Optionally, before obtaining the inspection path of the robot based on the path planning algorithm, the relative attitude information of the robot in the surrounding environment is collected based on the attitude sensor, and the relative attitude information of the robot is combined with the absolute position information. The inspection path of the robot is planned in the map; wherein, the attitude sensor is mounted on the robot.

Optionally, the robot is equipped with an industrial detection camera to take a global picture of the high-speed rail box girder, and the industrial detection camera includes a top detection camera, a left detection camera, a right detection camera and a bottom detection camera.

Optionally, the process of training the initial HED edge detection model includes, delineating the boundary of each target object in the high-speed rail box girder image, and saving it as a vector data format; Converted into an edge image after conversion, the high-speed rail box girder image and the edge image are used to make edge training samples based on the preset production size, and the initial HED edge detection model is trained based on the edge training samples.

Optionally, the process of obtaining the internal defect image of the high-speed rail box girder includes: segmenting the image of the high-speed rail box girder based on a preset ratio to obtain several image blocks; inputting the several image blocks into the target HED edge detection model Perform detection and splicing to obtain a complete edge probability map; perform binarization, skeleton extraction, and vector-derived processing on the complete edge probability map to obtain a vector polygon; simplify the vector polygon to obtain the high-speed rail The image segmentation result of box girder, and then obtain the target defect image of described high-speed rail box girder;

Among them, the simplification process includes eliminating the aliasing of the boundary of the vector polygon and finely breaking the hole of the vector polygon.

Optionally, the process of realizing the defect detection of the high-speed rail box girder includes, based on the position information of the industrial inspection camera mounted on the robot, obtaining the relative position information of the target defect image of the high-speed rail box girder; Absolute position information: Based on the relative position information of the target defect image and the absolute position information of the robot, the position information of hidden danger points in the high-speed rail box girder is obtained, and the patrol inspection of the high-speed rail box girder is completed.

Technical effect of the present invention is:

The method for realizing automatic inspection of high-speed rail box girders provided by the present invention can not only realize automatic detection of defects in high-speed rail box girders, but also has fast detection speed and high efficiency, and can save staff from working in high-risk, harsh and cruel working environments. lower flaw detection;

The invention can obtain the position information of hidden danger points in the high-speed rail box girder according to the relative position information of the target defect image of the high-speed rail box girder and the absolute position information of the inspection robot, which is beneficial to the later maintenance and repair work and improves the work efficiency of repair and maintenance , can provide timely maintenance and strong support for the safe operation of high-speed rail;

The HED edge detection model adopted by the present invention does not need to set any parameters during use, which greatly simplifies the difficulty of use; and the segmentation effect of the model is better, the segmentation scale is finer, and the boundary is more accurate, which greatly improves the detection of defect points of high-speed rail box girders. the accuracy.

Description of drawings

The drawings constituting a part of the application are used to provide further understanding of the application, and the schematic embodiments and descriptions of the application are used to explain the application, and do not constitute an improper limitation to the application. In the attached picture:

FIG. 1 is a flow chart of an automatic inspection method in an embodiment of the present invention.

detailed description

It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and embodiments.

It should be noted that the steps shown in the flowcharts of the accompanying drawings may be performed in a computer system, such as a set of computer-executable instructions, and that although a logical order is shown in the flowcharts, in some cases, The steps shown or described may be performed in an order different than here.

Embodiment one

As shown in Figure 1, a method for automatic inspection of a high-speed rail box girder inspection robot is provided in this embodiment, including the following steps:

Collect the surrounding environment data and the absolute position information of the robot, and generate an environment map based on the surrounding environment data; combine the absolute position information of the robot with the environment map, and obtain the inspection path of the robot based on the path planning algorithm; control the robot based on the inspection path of the robot. Move, and take a global shot of the inside of the high-speed rail box girder in real time to obtain the image information of the high-speed rail box girder; build an initial HED edge detection model, and train the initial HED edge detection model based on the high-speed rail box girder image information to obtain the target HED edge detection model; The image information of the high-speed rail box girder collected in real time is input into the target HED edge detection model for processing, and the target defect image of the high-speed rail box girder is obtained to realize the defect detection of the high-speed rail box girder.

Practical, the inspection robot is equipped with NB-IOT sensors, panoramic cameras, lidars, and synchronization controllers; panoramic cameras include stereo image sensors; lidars include laser scanners and laser scanning sensors.

Practical, the process of collecting the surrounding environment data and the absolute position information of the robot includes: the NB-IOT sensor establishes a connection with the NB-IOT base station set up in the surrounding environment to obtain the position information of the inspection robot; The surrounding environment of the robot is scanned and measured to obtain the relative three-dimensional point cloud data of the surrounding environment; the position information and the relative three-dimensional point cloud data are registered to obtain the absolute three-dimensional point cloud data, and then the absolute position information of the robot is obtained.

The NB-IOT base station in this embodiment can receive GNSS signals and obtain the absolute spatial coordinates of the NB-IOT base station. The NB-IOT sensor takes advantage of the wide coverage area and strong penetrating ability of the narrowband IoT signal, and can establish data exchange with the NB-IOT base station. Both the panoramic camera and the laser radar in this embodiment are used to obtain the three-dimensional space point cloud data of the surrounding environment of the inspection robot.

In this embodiment, the NB-IOT base station includes a movable NB-IOT sensing device and GNSS receivers connected thereto, and the number is at least three, which are distributed around the periphery of the environment to be measured.

The purpose of using the synchronous controller in this embodiment is to enable the above-mentioned measured position information and collected 3D point cloud data to have a unified time reference, thereby improving the accuracy of the finally acquired point cloud data. When the location information of the inspection robot and all point cloud data are collected, they can be uploaded to the computer, and then the location information of the inspection robot and the relative three-dimensional point cloud data are registered by using the point cloud data registration method, and further Eliminate errors, thereby improving the accuracy of the final absolute 3D point cloud data.

It can be implemented, before obtaining the inspection path of the robot based on the path planning algorithm, the relative attitude information of the robot in the surrounding environment is collected based on the attitude sensor, and the relative attitude information of the robot is combined with the absolute position information. The inspection path of the robot is planned on the map; among them, the attitude sensor is mounted on the robot.

It can be implemented. When the inspection robot inspects according to the planned inspection path, it relies on the industrial inspection camera on the robot to take a global picture of the high-speed rail box girder. The industrial inspection camera includes a top inspection camera, a left inspection camera, and a right inspection camera. Surface detection camera and bottom surface detection camera.

It can be implemented that the robot is also equipped with an infrared sensor and/or an ultrasonic sensor to detect obstacles; when an obstacle is detected on the final walking path, the robot stops walking and judges whether the obstacle existence time reaches the preset threshold; if the obstacle exists If the time does not reach the preset threshold, it is determined that the obstacle is a temporary obstacle, and the robot continues to walk according to the final walking path after the obstacle disappears; if the existence time of the obstacle reaches the preset threshold, it is determined that the obstacle is a fixed obstacle After the robot bypasses the obstacle, it returns to the final walking path and continues to walk according to the final walking path. Wherein, the determination of the preset threshold can be selected according to actual needs.

It can be implemented, after the global image of the high-speed rail box girder is obtained based on the industrial inspection camera, the image of the high-speed rail box girder is analyzed and processed based on the HED edge detection model, and then the target defect image is obtained.

Specifically, the initial HED edge detection model is constructed, and the initial HED edge detection model is trained. The training process includes, outlining the boundary of each target object in the high-speed rail box girder image, and saving it as a vector data format; After rasterization, it is converted into an edge image, and the high-speed rail box girder image and the edge image are used to make edge training samples based on the preset production size. Based on the edge training samples, the initial HED edge detection model is trained to obtain the target HED edge detection model.

Segment the image of the high-speed rail box girder based on the preset ratio to obtain several image blocks; input several image blocks into the target HED edge detection model for detection and splicing to obtain a complete edge probability map; for the complete edge probability map Perform binarization, skeleton extraction and vector export processing to obtain vector polygons; simplify vector polygons to obtain image segmentation results of high-speed rail box girders, and then obtain target defect images of high-speed rail box girders;

Among them, the simplification process includes eliminating the aliasing of the vector polygon boundary and finely breaking the holes of the vector polygon to make it have a better visual effect. First use the area threshold method to eliminate finely divided polygons, set a threshold, and delete polygons whose area is smaller than the threshold. To eliminate holes, you should first read the geometric points of each polygon of the vector, keep only the outermost geometric points, and discard the inner geometric points.

Further, based on the position information of the industrial detection camera mounted on the robot, the relative position information of the target defect image of the high-speed rail box girder is obtained; the absolute position information of the robot is obtained in real time based on the NB-IOT sensor; based on the relative position information of the target defect image and The absolute position information of the robot can obtain the position information of hidden danger points in the box girder of the high-speed rail, and complete the inspection of the box girder of the high-speed rail.

After the HED edge detection model adopted in this embodiment is trained, no parameters need to be set during use, and the individual parameters required in the process of processing the model calculation result data can also be automatically calculated through image features. When using it, you only need to specify the image to be segmented and the output segmentation vector path, which greatly simplifies the difficulty of use. Compared with the object-oriented and depth segmentation models, the segmentation effect of this embodiment is better, the segmentation scale is finer, and the boundary is more accurate. And the trained model can run under the GPU, and its parallel computing efficiency is significantly higher than that of the CPU. Compared with the object-oriented image segmentation algorithm that only uses the CPU, the image segmentation efficiency can be greatly improved.

The method for realizing automatic inspection of high-speed rail box girders provided by this embodiment can not only realize automatic detection of defects in high-speed rail box girders, but also has fast detection speed and high efficiency, which can save staff from working in high-risk, harsh and cruel work. Flaw detection in the environment, while avoiding the subjectivity of inspection results, can provide timely maintenance and strong support for the safe operation of high-speed rail.

The above is only a preferred embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any skilled person familiar with the technical field can easily think of changes or Replacement should be covered within the protection scope of this application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims (8)

1. A method for realizing automatic inspection by a high-speed rail box girder inspection robot is characterized by comprising the following steps:

collecting surrounding environment data and absolute position information of a robot, and generating an environment map based on the surrounding environment data;

combining the absolute position information of the robot with an environment map, and obtaining a patrol route of the robot based on a route planning algorithm;

controlling the robot to move based on the routing inspection path of the robot, and carrying out global shooting on the interior of the high-speed railway box girder in real time to obtain image information of the high-speed railway box girder;

constructing an initial HED edge detection model, and training the initial HED edge detection model based on the image information of the high-speed railway box girder to obtain a target HED edge detection model;

and inputting the image information of the high-speed rail box girder acquired in real time into the target HED edge detection model for processing to obtain a target defect image of the high-speed rail box girder, thereby realizing the defect detection of the high-speed rail box girder.

2. The method for realizing automatic inspection by the inspection robot for the high-speed railway box girder according to claim 1,

the inspection robot is provided with an NB-IOT sensor, a panoramic camera, a laser radar and a synchronous controller; the panoramic camera includes a stereoscopic image sensor; the lidar includes a laser scanner and a laser scanning sensor.

3. The method for realizing automatic inspection by the inspection robot for the high-speed railway box girder according to claim 2, is characterized in that,

the process of collecting the ambient data and the absolute position information of the robot includes,

the NB-IOT sensor is connected with an NB-IOT base station set in the surrounding environment to acquire the position information of the inspection robot; the panoramic camera and the laser radar scan and measure the periphery of the inspection robot to obtain relative three-dimensional point cloud data of the surrounding environment; and registering the position information and the relative three-dimensional point cloud data to obtain absolute three-dimensional point cloud data, and further obtain absolute position information of the robot.

4. The method for realizing automatic inspection by the inspection robot for the high-speed railway box girder according to claim 1, is characterized in that,

before obtaining the inspection path of the robot based on a path planning algorithm, acquiring relative attitude information of the robot in the surrounding environment based on an attitude sensor, combining the relative attitude information of the robot with absolute position information, and planning the inspection path of the robot in an environment map based on the path planning algorithm; wherein the attitude sensor is mounted on the robot.

5. The method for realizing automatic inspection by the inspection robot for the high-speed railway box girder according to claim 1,

carry on industry detection camera on the robot and carry out the global shooting to the high-speed railway case roof beam, industry detection camera includes top surface detection camera, left surface detection camera, right flank detection camera and bottom surface detection camera.

6. The method for realizing automatic inspection by the inspection robot for the high-speed railway box girder according to claim 1,

the training process of the initial HED edge detection model comprises the steps of drawing the boundary of each target object in the high-speed railway box girder image and storing the boundary in a vector data format; rasterizing the vector data, converting the rasterized vector data into an edge image, manufacturing an edge training sample of the high-speed railway box girder image and the edge image based on a preset manufacturing size, and training the initial HED edge detection model based on the edge training sample.

7. The method for realizing automatic inspection by the inspection robot for the high-speed railway box girder according to claim 1,

the process of obtaining the internal defect image of the high-speed rail box girder comprises the steps of dividing the image of the high-speed rail box girder based on a preset proportion to obtain a plurality of image blocks; inputting the image blocks into a target HED edge detection model for detection and splicing to obtain a complete edge probability graph; carrying out binarization, skeleton extraction and vector derivation processing on the complete edge probability map to obtain a vector polygon; simplifying the vector polygon to obtain an image segmentation result of the high-speed railway box girder, and further obtain a target defect image of the high-speed railway box girder;

the simplification processing comprises eliminating sawtooth on the boundary of the vector polygon and finely crushing the vector polygon holes.

8. The method for realizing automatic inspection by the inspection robot for the high-speed railway box girder according to claim 1, is characterized in that,

the method comprises the steps that the relative position information of a target defect image of the high-speed railway box girder is obtained based on the position information of an industrial detection camera carried on a robot; acquiring absolute position information of the robot in real time based on the NB-IOT sensor; and acquiring the position information of the hidden danger points in the high-speed rail box girder based on the relative position information of the target defect image and the absolute position information of the robot, and finishing the inspection of the high-speed rail box girder.

Images