A robot nondestructive testing method and system for high-altitude pipelines

Abstract

The present application relates to a robot nondestructive testing method and system for high-altitude pipelines, including obtaining several standard images of historical standard pipelines and several defective images of defective pipelines, and storing all the standard images and defective images as an initial data set; standardizing the initial data set to obtain a standard data set; establishing an initial pipeline detection model, and training and optimizing the initial pipeline detection model according to the standard data set to obtain a pipeline detection model; sequentially obtaining several images of the pipeline to be detected collected by the radar equipment, and uniformly standardizing the acquired images to obtain several images to be detected; sequentially inputting all the images to be detected into the pipeline detection model to obtain actual detection results. The present application can combine ultrasonic technology and image analysis and processing to greatly improve the detection efficiency of high-altitude pipelines and the accuracy of the obtained detection results.

Description

Technical Field

[0001] The present application relates to the field of robot detection technology, and in particular to a robot non-destructive detection method and system applied to high-altitude pipelines. Background Art

[0002] High-altitude pipelines shuttle between high-rise buildings, elevated roads and mountains. During use, these pipelines are exposed to the outdoor environment and are affected by rain, snowfall, atmospheric conditions and other conditions for a long time. The pipelines are easily polluted, corroded or damaged, which in turn affects the transportation of the corresponding materials, causing huge economic losses and may even threaten human health and safety. Therefore, it is of great significance to regularly inspect the health status of pipelines.

[0003] At present, experts and scholars have developed a variety of pipeline robots that can enter complex and changeable unstructured pipeline environments and complete pipeline inspection, cleaning and maintenance tasks through the operating devices they carry, greatly reducing the danger of manual operations. However, most of the pipeline robots currently in use use a supporting wheel-foot mechanism and a crawler mechanism to enter the interior of the pipeline for inspection. When encountering a large amount of accumulated oil and dust in the pipeline, this inspection method can easily cause the robot to slip or be unable to move forward due to these oil or dust during the movement process, resulting in the inability to successfully complete the inspection of the entire pipeline, making the inspection efficiency of high-altitude pipelines low. Summary of the invention

[0004] In order to improve the problem that the inspection robot may not be able to successfully complete the inspection of the entire pipeline when there are impurities inside the pipeline, resulting in low inspection efficiency of high-altitude pipelines, the present application provides a robot non-destructive inspection method and system for high-altitude pipelines.

[0005] In the first aspect, the present application provides a robot nondestructive testing method for high-altitude pipelines, which adopts the following technical solution:

[0006] A robot nondestructive testing method for high-altitude pipelines is based on a robot nondestructive testing system, which includes a radar device for transmitting and receiving ultrasonic waves, and is characterized in that the robot nondestructive testing method includes:

[0007] Acquire several standard images of historical standard pipelines and several defective images of defective pipelines, and store all the standard images and defective images as an initial data set;

[0008] Standardizing the initial data set to obtain a standard data set;

[0009] Establishing an initial pipeline detection model, and training and optimizing the initial pipeline detection model according to the standard data set to obtain a pipeline detection model;

[0010] sequentially acquiring a plurality of images of the pipeline to be inspected collected by the radar device (200), and uniformly standardizing the acquired images to obtain a plurality of images to be inspected;

[0011] Inputting all the images to be detected into the pipeline detection model in sequence to obtain actual detection results;

[0012] The training and optimization of the initial pipeline detection model according to the standard data set to obtain the pipeline detection model includes: dividing the standard data set into a training set and a test set according to a preset ratio, wherein both the training set and the test set include the standard image and the defect image that have been uniformly standardized;

[0013] Training the pipeline detection initial model using the training set;

[0014] Testing and optimizing the trained initial pipeline detection model using the test set to obtain a pipeline detection model;

[0015] The training of the pipeline detection initial model by using the training set includes:

[0016] Input all images in the training set into the pipeline detection initial model, divide each image into a number of detection areas, and extract the local feature matrix of all the detection areas corresponding to each image;

[0017] Classify all images in the pipeline inspection initial model to obtain standard images and defect images;

[0018] The average value of the local feature matrices corresponding to all the standard images is calculated, and the average value is used as the standard feature matrix. The calculation formula of the standard feature matrix is ​​as follows:

[0019]

[0020] Among them, X 训，标准 (n,i,j) represents the local feature matrix of the nth standard image in the detection area (i,j) in the training set; represents the standard feature matrix; a n(e,f) , e=1 or 2, f=1 or 2, represents the element in the e-th row and f-th column of the local feature matrix corresponding to the n-th standard image; N represents the total number of standard images; e=1 or 2, f=1 or 2, indicating the average value of the elements in the eth row and fth column in the local feature matrix corresponding to all standard images;

[0021] Obtaining a defect feature matrix based on the standard feature matrix and the local feature matrix of the defect image to complete the training of the pipeline detection initial model, wherein the pipeline detection initial model includes a plurality of detection sub-models for detecting different defect types;

[0022] The step of obtaining a defect feature matrix based on the standard feature matrix and the local feature matrix of the defect image to complete the training of the initial pipeline detection model includes:

[0023] The local feature matrix of each detection area of ​​each defect image is respectively compared with the standard feature matrix of the corresponding detection area to obtain a first phase difference. The calculation formula of the first phase difference is:

[0024]

[0025] Among them, X 缺陷 (a,i,j) represents the local feature matrix of the a-th defect image in the detection area (i,j) in the training set; represents a standard feature matrix corresponding to all detection areas (i, j); d represents X 缺陷 (a,i,j) and The Euclidean distance between e,f , e=1 or 2, f=1 or 2, represents the feature matrix X 缺陷 The element in row e and column f in (a,i,j); b e,f ,e=1 or 2,f=1 or 2, represents the feature matrix The element in the e-th row and the f-th column; D1 represents the first phase difference, and its value range is [0,1];

[0026] Obtaining a feature matrix to be merged corresponding to some of the detection areas in each of the defect images according to the first phase difference, and updating the standard feature matrix according to all the feature matrices to be merged;

[0027] Obtaining a suspected defect feature matrix corresponding to a certain detection area in each of the defect images according to the first phase difference; testing each of the suspected defect feature matrices, and obtaining a defect feature matrix according to all the suspected defect feature matrices that pass the test;

[0028] The step of obtaining a feature matrix to be merged corresponding to certain detection areas in each defect image according to the first phase difference, and updating the standard feature matrix according to all the feature matrices to be merged includes:

[0029] When the first phase difference is less than the phase difference threshold, the local feature matrix corresponding to the first phase difference is used as the feature matrix to be merged, and the standard feature matrix is ​​updated according to all the feature matrices to be merged. The calculation formula for the update is as follows:

[0030]

[0031] Among them, X 待合并 (n,i,j) represents the feature matrix to be merged; represents the standard feature matrix before updating; represents the updated standard feature matrix; b e,f ,e=1 or 2,f=1 or 2, represents the feature matrix The element in row e and column f in c n(e,f) , e=1 or 2, f=1 or 2, representing the element in the e-th row and f-th column of the n-th feature matrix to be merged; M represents the total number of feature matrices to be merged.

[0032] In a specific implementation scheme, each of the suspected defect feature matrices is tested, and a defect feature matrix is ​​obtained based on all the suspected defect feature matrices that have passed the test, including:

[0033] After obtaining each suspected defect feature matrix, the corresponding detection area is used as the inspection area, and each inspection area is divided twice to obtain several inspection sub-areas;

[0034] Extract the test sub-feature matrix of each test sub-region, and take the average value of all the test sub-feature matrices as the test feature matrix;

[0035] Calculate the phase difference parameters between each inspection feature matrix and the suspected defect feature matrix corresponding to the corresponding inspection area;

[0036] If the phase difference parameter is less than a preset phase difference threshold, a defect feature matrix is ​​obtained according to all inspection sub-feature matrices corresponding to the phase difference parameter.

[0037] In a specific embodiment, the calculation formula of the phase difference parameter is as follows:

[0038]

[0039] X2=X 疑似缺陷 (i,j);

[0040]

[0041] Among them, γ(X1,X2) represents the phase difference parameter between X1 and X2; That is, X1 represents the inspection feature matrix corresponding to the inspection area (i, j); X 疑似缺陷 (i, j) represents the suspected defect feature matrix corresponding to the inspection area (i, j); X d ', d = 1 or 2, represents the transposed matrix of the feature matrix X1; X d ' T ,d=1 or 2, represents the feature matrix X d 'The transposed matrix; w represents the feature matrix X d ,d=1 or 2, the number of rows; X 检验子 (u,v) represents the inspection sub-region feature matrix corresponding to the inspection sub-region (u,v); U represents the number of inspection sub-regions included in the rows of the inspection region (i,j); V represents the number of inspection sub-regions included in the columns of the inspection region (i,j).

[0042] In a specific implementation scheme, the calculation formula of the defect feature matrix is ​​as follows:

[0043]

[0044] Among them, X 缺陷 (i, j) represents the defect feature matrix corresponding to the inspection area (i, j); X 检验子 (u,v),u≤m,v≤m, represents the feature matrix corresponding to the inspection sub-region located in the uth row and vth column in the inspection region corresponding to the suspected defect feature matrix.

[0045] In a second aspect, the present application provides a smart terminal, which adopts the following technical solution:

[0046] An intelligent terminal comprises a memory and a processor, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the robot nondestructive testing method as described in the first aspect.

[0047] In a third aspect, the present application provides a robot nondestructive testing system, which adopts the following technical solution:

[0048] A robot nondestructive testing system including the intelligent terminal as described in the second aspect, characterized in that it includes:

[0049] The robot has a magnetic suction cup installed at the bottom, and is used to crawl on the outer wall of the pipeline to be inspected along the length direction of the pipeline to be inspected;

[0050] A radar device is arranged on the robot, and transmits and receives ultrasonic waves to the pipeline to be inspected, so as to obtain a plurality of images of the pipeline to be inspected;

[0051] The intelligent terminal is connected to the robot and the radar device for communication, and is used to obtain a plurality of images of the pipeline to be inspected collected by the radar device, and obtain actual inspection results of the pipeline to be inspected based on the obtained images.

[0052] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution:

[0053] A computer-readable storage medium, wherein at least one instruction, at least one program, code set or instruction set is stored in the readable storage medium, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the robot non-destructive testing method as described in the first aspect.

[0054] In summary, the present application includes at least one of the following beneficial technical effects:

[0055] 1. Ultrasonic testing can be used to directly detect the collected images through intelligent terminals to find tiny defects in the pipeline, such as cracks, corrosion, etc. The high sensitivity of the processor enables these defects that may be difficult to detect with the naked eye or other detection methods to be detected in time;

[0056] 2. Compared with traditional destructive testing methods, ultrasonic testing will not cause any damage to the pipeline, thus ensuring the integrity and safety of the pipeline after testing;

[0057] 3. Ultrasonic technology can penetrate thicker pipe walls and conduct comprehensive inspections inside the pipe. It is not affected by the surface conditions of the pipe and has wide adaptability.

[0058] 4. The intelligent terminal can provide instant detection results in the process of acquiring images collected by radar equipment, which is convenient for timely discovery and handling of potential problems, reducing downtime and maintenance costs; it can record data during the detection process, providing strong support for subsequent data analysis and pipeline status assessment;

[0059] 5. When the detection method provided in the present application encounters a large amount of accumulated oil and dust in the pipeline, the robot can carry the radar equipment and move normally on the surface of the pipeline, avoiding slipping or being unable to move forward due to oil or dust during the movement, thereby successfully completing the detection of the entire pipeline and greatly improving the detection efficiency of high-altitude pipelines. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a structural schematic diagram of a robot nondestructive testing system according to an embodiment of the present application.

[0061] Figure 2It is a flow chart of a robot nondestructive testing method applied to high-altitude pipelines according to another embodiment of the present application.

[0062] Figure 3 It is a structural diagram used to reflect the relationship between the standard / defect image and the detection area in another embodiment of the present application.

[0063] Figure 4 Another embodiment of the present application Figure 3 A in the middle is an enlarged view showing the relationship between the detection area and the detection sub-area.

[0064] Explanation of reference numerals: 100, robot; 200, radar equipment; 300, intelligent terminal; 310, processor; 320, memory. DETAILED DESCRIPTION

[0065] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0066] The following is a further detailed description of an embodiment of a robot nondestructive testing method and system for high-altitude pipelines of the present application in conjunction with the drawings in the specification.

[0067] An embodiment of the present application discloses a robot nondestructive testing system.

[0068] Reference Figure 1 A robot nondestructive testing system includes a robot 100, a radar device 200 and an intelligent terminal 300. Among them, a magnetic suction cup is installed at the bottom of the robot 100. The robot 100 is used to adsorb on the outer wall of the steel pipeline (pipeline to be inspected) through the magnetic suction cup set at the bottom and crawl along the length direction of the pipeline to be inspected. The magnetic suction cup can adjust the magnetic force so that the robot 100 can move stably on parts of different sizes and slopes on the pipeline. The robot 100 in this example takes a magnetic adsorption wall climbing robot as an example, which is a prior art and will not be repeated here. The radar device 200 can be detachably installed on the robot 100 body, and is used to transmit and receive ultrasonic waves to the pipeline to be inspected to obtain several images of the pipeline to be inspected.

[0069] The robot 100 and the radar device 200 are both connected to the intelligent terminal 300 for communication. The intelligent terminal 300 of this embodiment includes a processor 310 and a memory 320, which are used to obtain a number of images of the pipeline to be inspected currently collected by the radar device 200 for processing by the processor 310, and obtain the actual inspection result of the pipeline to be inspected currently. The memory 320 is used to store at least one instruction, at least one program, code set or instruction set; when the processor 310 runs at least one instruction, at least one program, code set or instruction set, it executes the following steps of a robot nondestructive inspection method applied to high-altitude pipelines.

[0070] It should be noted that, in the process of obtaining the detection results, the intelligent terminal 300 of this embodiment will control the robot 100 to move in a specified direction along the outer wall of the steel pipe based on a preset pathfinding program or an operator's remote control program. This is prior art and will not be described in detail in this application.

[0071] Another embodiment of the present application discloses a robot nondestructive testing method applied to high-altitude pipelines.

[0072] Reference Figure 2 A robot nondestructive testing method for high-altitude pipelines is provided. Based on the above-mentioned robot nondestructive testing system, the robot nondestructive testing method includes:

[0073] S100, obtaining a number of standard images of historical standard pipelines and a number of defective images of defective pipelines, and storing all the standard images and defective images as an initial data set;

[0074] Among them, the standard image and the defect image both refer to the cross-sectional image of a certain part of the pipeline collected by the radar device 200. These cross-sectional images are reconstructed by the signal transmitted by the radar device 200 and propagated inside the pipeline and reflected back. The radar device 200 receives the reflected signal and analyzes its time delay, amplitude, phase and other parameters to infer the structure and characteristics of the inside of the pipeline, and displays it in the form of an image. The image used to reflect the internal situation of the pipeline obtained by ultrasound is a prior art and will not be repeated here. The standard image and defect image of this embodiment can come from an industrial defect detection data set, or it can be a data set collected in previous daily inspections. These images are all labeled with two labels, namely, the pipeline type (straight pipe / bend pipe) label and the defect condition label (standard / defect, where the label is defective and the specific defect type will be indicated at the same time); for the convenience of calculation, the initial data set of this embodiment includes several standard images and one defect image representing different defect types, as shown in Table 1.1 below:

[0075] Table 1.1

[0076] Initial Dataset Label - Pipeline Type Label-defect situation Standard image 1 Straight pipe / Standard image 2 Straight pipe / Defective image 3 Straight pipe Defects-Cracks Defective image 4 Straight pipe Defects - Corrosion Defective image 5 Straight pipe Defect - Wall thickness reduction ...... ...... ...... Standard Image 6 Pipe bending / Standard Image 7 Pipe bending / Defective image 8 Pipe bending Radius Variation Defective image 9 Pipe bending crack Defect image 10 Pipe bending Flattened cross section Defective image 11 Pipe bending Wall thickness reduction ...... ...... ......

[0077] Generally speaking, the labels of the images in the initial data set are set by manual detection, and the labels of the newly added images are first automatically detected by the intelligent terminal 300, and then manually verified and confirmed, and then the images and labels are added to the initial data set. The initial data set includes standard images labeled as straight pipes / bent pipes (i.e., pipes without defects) and defective images labeled as xx defect + straight pipes / bent pipes (i.e., pipes with only a certain defect).

[0078] After S100, the following steps are also included:

[0079] S110, standardizing the initial data set to obtain a standard data set;

[0080] Specifically, the unified standardization in S110 includes preprocessing steps such as denoising, enhancement, edge detection, and edge cropping for all images (standard / defective images) in the initial data set to improve image clarity and facilitate subsequent image detection; among them, the standard data set includes standard images and defective images with a unified format and after image enhancement.

[0081] S200, establishing an initial pipeline detection model, and training and optimizing the initial pipeline detection model according to a standard data set to obtain a pipeline detection model;

[0082] Among them, the pipeline detection model includes a straight pipe detection model for detecting straight pipe defects and a bent pipe detection model for detecting bent pipe defects; the straight pipe detection model and the bent pipe detection model each include a number of detection sub-models, which are used to detect different defect types, and different detection sub-models in the straight pipe detection model and the bent pipe detection model will have different priorities according to the occurrence rates of different defect types in different pipelines (straight pipes / bent pipes); specifically, the relationship between the pipeline detection model / pipeline detection initial model and the detection sub-model is shown in Table 2.1 below:

[0083] Table 2.1

[0084]

[0085] As shown in Table 2.1 above, when gas or liquid passes through a straight pipe for a long time, corrosion is the most likely defect to occur, considering the chemical substances or environmental factors (such as humidity, temperature, etc.) on the inner and outer walls of the pipe, so it has the highest priority; cracks are caused by factors such as pipe materials, stress concentration or poor welding, and are usually formed by a small crack that slowly expands and extends. The occurrence rate is high and it is ranked second priority... Due to its special curved shape, the elbow pipe may be deformed during the manufacturing or installation of the elbow pipe if it is improperly operated or affected by external forces. During use, it will gradually manifest itself through radius changes, cross-section flattening, etc. Therefore, in the defect detection of the elbow pipe, the detection sub-models corresponding to the above two defect types have a higher priority...

[0086] It should be noted that the training and optimization process of the initial pipeline detection model in S200 mainly includes two parts: obtaining a standard feature matrix and obtaining a defect feature matrix corresponding to each defect type; wherein the standard feature matrix is ​​a detection condition for determining whether there are defects in the image to be detected of the pipeline in the subsequent S400, and the defect feature matrix is ​​a detection condition for determining the specific defect type existing in the image to be detected; the intelligent terminal 300 obtains the standard feature matrix corresponding to each straight pipe / bent pipe detection model through the following steps S221-S222, and stores the standard feature matrix in the corresponding detection model, and obtains several defect feature matrices corresponding to each straight pipe / bent pipe detection model through the following step S223, and stores each defect feature matrix for different defect types in the corresponding detection sub-model in the corresponding detection model.

[0087] Specifically, since the principles of the training optimization steps of the straight pipe detection model and the bent pipe detection model in S200 are consistent, this embodiment takes the straight pipe detection model in the pipeline detection initial model as an example to illustrate the training optimization process of the pipeline detection initial model in S200. S200 includes:

[0088] S210, dividing the standard data set into a training set and a test set according to a preset ratio;

[0089] In this embodiment, 70% of the standard images and defect images in the initial data set are used as a training set, and 30% are used as a test set.

[0090] S220, training a straight pipe detection model by using all images in a training set whose pipe type is a straight pipe;

[0091] Specifically, S220 includes:

[0092] S221, inputting all images of the straight pipe type in the training set into the straight pipe detection model, dividing each image into a plurality of detection areas, and extracting the local feature matrix of all the detection areas corresponding to each image;

[0093] Among them, refer to Figure 3 In this embodiment, each image in the training set is divided into n*n grids (n in this embodiment is 10 as an example), starting with the first detection area in the upper left corner of each image, and naming each divided detection area as the position (i, j) of the detection area, that is, (i, j) indicates that the detection area is located in the i-th row and j-th column of the corresponding image, i increases from left to right, j increases from top to bottom, and 1≤i≤n, 1≤j≤n; the local feature matrix of this embodiment is X m,l(a,i,j) represents, where m is one of the training set and the test set, l is one of the standard and the defect, and the feature matrix represents the feature matrix of the a-th image l in m in the detection area (i,j).

[0094] It should be noted that all feature matrices in this embodiment (except the defect feature matrix in S2234) are based on two-dimensional arrays as examples. These feature matrices are used to represent the digital image data of an image, where the rows of the feature matrix correspond to the height of the image, the columns correspond to the width of the image, the elements in the feature matrix correspond to the pixels of the image, and the values ​​of the elements are the grayscale values ​​of the pixels. Since the grayscale value in the grayscale image ranges from 0 to 255, the closer to white, the larger the value, and the closer to black, the smaller the value; the local feature matrix in this embodiment is based on For example, the pipeline detection model of this embodiment can perform defect detection on the pipeline according to the specific manifestations of different defect types on the pipeline (the arrangement characteristics of grayscale values ​​in a certain detection area), such as: (combined with the following S200-S400) when the defect type of a certain detection area in a certain image to be detected on the pipeline to be detected is wall thinning, since the area representing the pipe wall on the image is reduced (that is, the gray-white area increases), the feature matrix to be matched corresponding to the detection area will show the characteristic of lower outer circle element values.

[0095] S222, calculating the average value of the local feature matrices corresponding to all standard images, and using the average value as the standard feature matrix;

[0096] Among them, the standard feature matrix of this embodiment is Indicates that the feature matrix represents the average value of the local feature matrix of all standard images in the training set in the detection area (i, j), which is used to reflect the gray value distribution characteristics of the standard image. The calculation formula is as follows:

[0097]

[0098] Among them, X 训，标准 (n,i,j) represents the local feature matrix of the nth standard image in the detection area (i,j) in the training set; represents the standard feature matrix; a n(e,f) , e=1 or 2, f=1 or 2, represents the element in the e-th row and f-th column of the local feature matrix corresponding to the n-th standard image; N represents the total number of standard images; e=1 or 2, f=1 or 2, which represents the average value of the elements in the e-th row and f-th column in the local feature matrix corresponding to all standard images.

[0099] S223, obtaining a defect feature matrix corresponding to each defect type based on the standard feature matrix and the local feature matrices of all defect images to complete the training of the straight tube detection model.

[0100] Among them, the defect feature matrix is ​​used to reflect the gray value distribution characteristics of the defect image; the defect feature matrices of different defect types will be stored in the corresponding detection sub-model; specifically, S223 includes:

[0101] S2231, performing phase difference comparison on the local feature matrix of each detection area of ​​each defect image and the standard feature matrix of the corresponding detection area to obtain a first phase difference;

[0102] Among them, the local feature matrix corresponding to the defect image is X 缺陷 (a,i,j) indicates that the feature matrix represents the feature matrix of the a-th defect image in the detection area (i,j) in the training set; the first phase difference of this embodiment is mainly obtained by the Euclidean distance calculation formula, and the formula is:

[0103]

[0104]

[0105] Among them, X 缺陷 (a,i,j) represents the local feature matrix of the a-th defect image in the detection area (i,j) in the training set; represents a standard feature matrix corresponding to all detection areas (i, j); d represents X 缺陷 (a,i,j) and The Euclidean distance between e,f , e=1 or 2, f=1 or 2, represents the feature matrix X 缺陷 The element in row e and column f in (a,i,j); b e,f ,e=1 or 2,f=1 or 2, represents the feature matrix The element in the e-th row and f-th column; D1 represents the first phase difference, and its value range is [0,1].

[0106] S2232, obtaining a feature matrix to be merged corresponding to certain detection areas in each defect image according to the first phase difference, and updating the standard feature matrix according to all the feature matrices to be merged;

[0107] Specifically, S2232 includes:

[0108] S22321, when the first phase difference is less than the phase difference threshold, taking the local feature matrix corresponding to the first phase difference as the feature matrix to be merged, and updating the standard feature matrix according to all the feature matrices to be merged;

[0109] Among them, the feature matrix to be merged refers to the local feature matrix of all detection areas without defects in the defective image (the detection area corresponding to the local feature matrix with a high similarity to the standard feature matrix can be regarded as a normal area without defects); therefore, the standard feature matrix in S223 can be updated according to the feature matrix to be merged to continuously improve the accuracy of the standard feature matrix representing the intact standard image, and the calculation formula for completing the update is as follows:

[0110]

[0111] Among them, X 待合并 (n,i,j) represents the feature matrix to be merged (the local feature matrix of the nth defect image in the detection area (i,j) in the training set); represents the standard feature matrix before updating; represents the updated standard feature matrix; b e,f ,e=1 or 2,f=1 or 2, represents the feature matrix The element in row e and column f in c n(e,f) , e=1 or 2, f=1 or 2, representing the element in the e-th row and f-th column of the n-th feature matrix to be merged; M represents the total number of feature matrices to be merged.

[0112] S2233, obtaining a suspected defect feature matrix corresponding to a certain detection area in all defect images of each defect type according to the first phase difference;

[0113] Specifically, S2233 includes:

[0114] S22331, when the first phase difference is greater than or equal to a preset phase difference threshold, taking the local feature matrix corresponding to the first phase difference as a suspected defect feature matrix;

[0115] Among them, the suspected defect feature matrix refers to the local feature matrix of a certain detection area where the defect is specifically located in the defect image (the detection area corresponding to the local feature matrix with a higher degree of difference from the standard feature matrix can be regarded as a defect area where defects exist); the difference threshold of this embodiment takes 0.2 as an example.

[0116] S2234, each suspected defect feature matrix is ​​tested, and a defect feature matrix is ​​obtained based on all suspected defect feature matrices that pass the test;

[0117] Specifically, S2234 includes:

[0118] S22341, after obtaining each suspected defect feature matrix, the corresponding detection area is used as the inspection area, and each inspection area is divided twice to obtain a plurality of inspection sub-areas;

[0119] Among them, refer to Figure 4 , in this embodiment, each inspection area is sub-divided twice using an m*m (m < n) grid. For the convenience of subsequent calculations, m in this embodiment is taken as 2.

[0120] S22342. Extract the inspection sub-feature matrix of each inspection sub-area, and take the average value of all inspection sub-feature matrices as the inspection feature matrix;

[0121] Among them, the inspection sub-feature matrix is represented by X 检验子 (u, v) in this embodiment, which refers to the feature matrix corresponding to the inspection sub-area located in the u-th row and v-th column in a certain inspection area; the calculation method of the inspection feature matrix is the same as that of the standard feature matrix in S223, which will not be elaborated here.

[0122] S22343. Calculate the difference parameter between each inspection feature matrix and the suspected defect feature matrix corresponding to the corresponding inspection area;

[0123] Among them, the difference parameter is used to reflect the degree of difference between the inspection feature matrix and the corresponding suspected defect feature matrix. The larger the value, the greater the difference between the two; the calculation formula of the difference parameter in this embodiment is as follows:

[0124]

[0125] X2 = X 疑似缺陷 (i, j);

[0126]

[0127] Among them, γ(X1, X2) represents the difference parameter between X1 and X2; (i.e., X1) represents the inspection feature matrix corresponding to the inspection area (i, j); X 疑似缺陷 (i, j) represents the suspected defect feature matrix corresponding to the inspection area (i, j); X d '(d = 1 or 2) represents the transpose matrix of the feature matrix X1; X d ' T (d = 1 or 2) represents the transpose matrix of the feature matrix X d '; w represents the number of rows of the feature matrix X d (d = 1 or 2), which is taken as 2 in this embodiment (a two-dimensional matrix has 2 rows of elements); X 检验子 (u, v) represents the inspection sub-feature matrix corresponding to the inspection sub-area (u, v); U represents the number of inspection sub-areas included in the row of the inspection area (i, j), which is taken as 2 in this embodiment (i.e., m in S22341); V represents the number of inspection sub-areas included in the column of the inspection area (i, j), which is taken as 2 in this embodiment (i.e., m in S22341).

[0128] S22344, if the phase difference parameter is less than a preset phase difference threshold, a defect feature matrix is ​​obtained according to all inspection sub-feature matrices corresponding to the phase difference parameter;

[0129] In this embodiment, the phase difference threshold is taken as 0.3 as an example; the calculation formula of the defect feature matrix is ​​as follows:

[0130]

[0131] Among them, X 缺陷 (i, j) represents the defect feature matrix corresponding to the inspection area (i, j). In this implementation, the feature matrix is ​​composed of the inspection sub-region feature vectors (two-dimensional) corresponding to the four inspection sub-regions. Therefore, the defect feature matrix takes a four-dimensional array as an example (the specific number of rows and columns is determined by the grid size (m*m) of the secondary division in S22341); X 检验子 (u,v),u≤m,v≤m, represents the feature matrix corresponding to the inspection sub-region located in the uth row and vth column in the inspection region corresponding to the suspected defect feature matrix.

[0132] S22345, if the phase difference parameter is greater than or equal to the phase difference threshold, the suspected defect feature matrix corresponding to the phase difference parameter is regarded as a detection error. At this time, the phase difference threshold in S2231 is adjusted and steps S2231-S2234 are repeated until the defect feature matrix is ​​obtained.

[0133] Specifically, since the same defective straight pipe / curved pipe image may include several normal areas without defects, but only includes an abnormal area where a defect is located, there must be only one defective straight pipe / curved pipe feature matrix finally obtained in S223 (when the defect position covers multiple detection areas, this embodiment can continuously adjust the phase difference threshold in S22331 so that in S22344, only the defective straight pipe / curved pipe feature matrix of the detection area with the most obvious defects is retained), and each defect feature matrix is ​​stored in the detection sub-model of the corresponding defect type in the corresponding straight pipe / curved pipe detection model.

[0134] It should be noted that, in this embodiment, even if the defect type labels corresponding to the two defect feature matrices from the straight pipe detection model and the bent pipe detection model are the same, their respective defect feature matrices will be different due to the differences in the materials of the straight pipe and the bent pipe, the force direction when transmitting substances, etc. Therefore, separately training the initial pipeline detection models applied to straight pipes and bent pipes will help to further improve the accuracy of nondestructive testing.

[0135] S230, testing and optimizing the trained straight pipe detection model by testing all images in the test set whose pipe type is a straight pipe, to obtain a straight pipe detection model that passes the test;

[0136] Specifically, S230 includes:

[0137] S231, input all images in the test set into the corresponding trained straight tube detection model;

[0138] Specifically, in step S231, each image in the test set is sequentially input into each detection sub-model in the straight pipe / bent pipe detection model according to its label for detection (the standard / defective image of the pipe type is a straight pipe is input into the straight pipe detection model, and the standard / defective image of the pipe type is a bent pipe is input into the bent pipe detection model). This embodiment takes a detection sub-model for detecting crack defects in the straight pipe detection model as an example to illustrate the steps of detecting and verifying a certain image to be tested (a defective image in the test set) with a crack defect label, and the steps are as follows:

[0139] S232, dividing the image to be tested into a plurality of detection areas, and extracting local feature matrices of all detection areas corresponding to the image to be tested;

[0140] The region segmentation method in step S232 is the same as that in step S221.

[0141] S233, performing a phase difference comparison between the local feature matrix extracted from each detection area and the standard feature matrix corresponding to the corresponding detection area, to obtain a plurality of second phase differences corresponding to each detection area one by one;

[0142] Among them, the second phase difference is used to reflect the degree of difference between the characteristic matrix of a certain detection area of ​​the image to be tested and the characteristic matrix of the corresponding detection area in the standard image. When this value is large, it can be preliminarily judged that there is some defect in the detection area corresponding to this value in the image to be tested.

[0143] S234, obtaining a detection result based on the second phase difference;

[0144] The detection result includes a first judgment result of whether the pipeline has a defect and a second judgment result reflecting the defect type, and each detection result corresponds to an image to be tested; specifically, S234 includes:

[0145] S2341, when there is a second phase difference greater than or equal to the phase difference threshold in S22331, the first judgment result is that there is a defect in the pipeline, and the defect type judgment mechanism is triggered at this time to obtain a second judgment result, and the first judgment result and the second judgment result are used as the detection result;

[0146] Specifically, the defect type judgment mechanism obtains the second judgment result including:

[0147] S23411, taking the detection area corresponding to the second phase difference as a defect area, and sequentially inputting the local feature matrix corresponding to the defect area into corresponding detection sub-models until a second judgment result is obtained;

[0148] Among them, sequential input means completing the input of the local feature matrix according to the priority of the detection sub-model (the probability of occurrence of each defect type); specifically, the step of obtaining the defect type according to the local feature matrix corresponding to the defect area for each detection sub-model includes:

[0149] S234111, dividing the defect region into defect sub-regions for a second time, and obtaining a feature matrix to be matched according to a local feature matrix of each defect sub-region;

[0150] The secondary segmentation of the defect area in this embodiment is the same as the secondary segmentation of the inspection area in S22341, and the elements of the feature matrix to be matched correspond to the local feature matrix of each defect sub-region, that is:

[0151]

[0152] Among them, X 待匹配 (i, j) represents the feature matrix to be matched (i.e., the local feature matrix corresponding to the defect area (i, j)). In this implementation, the feature matrix to be matched takes a four-dimensional array as an example (each local feature matrix is ​​a two-dimensional matrix), and the specific number of rows and columns is determined by the grid size (m*m) of the secondary division in S22341; X 缺陷子 (u,v),u≤m,v≤m, represents the local feature matrix corresponding to the defect sub-region in the u-th row and v-th column in the defect region.

[0153] S234112, performing a phase difference comparison between the feature matrix to be matched and the defect feature matrix in the current detection sub-model to obtain a third phase difference corresponding to the defect area;

[0154] Among them, the third phase difference is used to reflect the degree of difference between the image to be tested and the defect image. When this value is small, it can be further judged that the defect type existing in the defect area in the image to be tested is the same as the defect type corresponding to the current detection sub-model.

[0155] S234113, when the third phase difference is less than the phase difference threshold in S22344, the second judgment result is that the defect type is consistent with the defect type corresponding to the current detection sub-model;

[0156] The second judgment result of this embodiment takes the defect type as a crack as an example.

[0157] S234114, when the third phase difference is greater than or equal to the phase difference threshold, the local feature matrix corresponding to the defect area is input into the detection sub-model of the next priority, and steps SS234112-S234113 are repeated until the second judgment result is obtained.

[0158] S2342, when all the second phase differences are less than the phase difference threshold, the first judgment result is that the pipeline is normal, and the first judgment result is directly used as the detection result.

[0159] S235, obtaining a test result corresponding to each image to be tested according to the first judgment result and the second judgment result; wherein the test result is successful when any one of the following conditions is met:

[0160] ① The first judgment result is that the pipeline is normal and the image to be tested is a standard image;

[0161] ② The first judgment result is that there is a defect in the pipeline and the image to be tested is a defective image. At the same time, the second judgment result is consistent with the defect type of the defective image.

[0162] S236, if the proportion of all images to be tested whose corresponding test results are successful in the test set reaches the preset accuracy threshold, the test is considered to have passed; otherwise, the phase difference threshold in S22321 and the phase difference threshold in S22344 are adjusted and the process returns to S220 to retrain the straight tube detection model until a straight tube detection model that passes the test is obtained;

[0163] The accuracy threshold of this embodiment is 90% as an example; the straight pipe detection model includes several detection sub-models that have passed the test.

[0164] After S236, the intelligent terminal 300 will train the straight pipe detection model by using all images of curved pipes in the training set according to the same principle (i.e., S220), and test and optimize the straight pipe detection model by using all images of curved pipes in the test set (i.e., S230) to obtain a curved pipe detection model that passes the test, and then use the straight pipe detection model and the curved pipe detection model that pass the test as the pipeline detection model (the pipeline detection model includes the straight pipe detection model that passes the test and the curved pipe detection model that passes the test).

[0165] It should be noted that, when all images of the test set in S230 have been used for testing, but a pipeline detection model that passes the test has not yet been obtained, the number of samples in the test set can be increased by techniques such as data augmentation. This is an existing technology and will not be described in detail here.

[0166] S300, sequentially acquiring a plurality of images of the current pipeline to be inspected collected by the radar device 200, and uniformly standardizing the acquired images to obtain a plurality of images to be inspected;

[0167] Among them, since the pipeline is relatively large and the same pipeline is connected by several straight pipes and curved pipes, this embodiment takes the example of obtaining an image from one end of the current pipeline to be detected to the other end and every 10 meters; the unified standardization in S300 is the same as the unified standardization in S110.

[0168] S400, inputting all the images to be detected into the pipeline detection model in sequence to obtain an actual detection result corresponding to all the images to be detected;

[0169] The actual detection result here can be regarded as the detection result in S234, including the first judgment result of whether the pipeline has defects and the second judgment result reflecting the defect type. The difference is that the actual detection result in S400 corresponds to all the images to be detected of the pipeline to be detected (that is, only when the first judgment results corresponding to all the images to be detected are that the pipeline is normal, the actual detection result is that the pipeline is normal; when the first judgment result corresponding to any image to be detected is that the pipeline has defects, the actual detection result is that the pipeline has defects; the actual detection result of this embodiment is as follows: the pipeline is normal or the pipeline has corrosion and crack defects in the detection area (i1, j1) of the straight pipe / bend at the Xth position, and there is a corrosion defect in the detection area (i2, j2)... (where the Xth position can be automatically determined by the intelligent terminal 300 according to the detection order of the images to be detected); specifically, the detection process of the image to be detected in the pipeline detection model is consistent with the detection process principle of the image to be tested in the above steps S231-S234, and the only difference is that:

[0170] In the detection test process of the test image in S234, since all the straight pipe / bent pipe defect images in the standard data set of S110 have clear defect type labels (that is, each straight pipe / bent pipe defect image includes only one defect), in this process, as long as one detection sub-model obtains the second judgment result in S234, the detection is stopped;

[0171] In the actual inspection process of the image to be inspected in S400, considering that there may be multiple defects in one position (the same inspection area) in the actual pipeline, even if a defect has been detected in a certain inspection sub-model during the process, the inspection will continue until all corresponding inspection sub-models are traversed (that is, all defect types are obtained); at the same time, before determining the existence and type of defects, the intelligent terminal 300 will first determine whether the image to be inspected currently obtained corresponds to a straight pipe or a curved pipe in the pipeline according to the movement trajectory of the robot 100, and thereby input each image to be inspected into the corresponding straight pipe / curved pipe detection model.

[0172] It should be noted that in the actual detection process of S400, when the first judgment result of the image to be detected is that there is a defect in the pipeline, but the second judgment result cannot be obtained by traversing all detection sub-models, the intelligent terminal 300 will highlight all defective areas and directly use the first judgment result as the actual detection result to facilitate the detection personnel to manually detect all defective areas to determine the defect type; considering that the above situation usually occurs due to the appearance of unknown defect types in the image to be detected, the image to be detected can be used as the corresponding defective straight pipe / bent pipe image in the initial data set in S100 after manual processing (excluding defect features of other known defect types), and added to the training of the initial pipeline detection model in S200 to help the model adapt to more defect type detection in continuous detection.

[0173] Based on the same inventive concept mentioned above, an embodiment of the present application also discloses an intelligent terminal, which includes a memory and a processor, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set can be loaded and executed by the processor to implement a robot non-destructive inspection method for high-altitude pipelines provided in the above method embodiment.

[0174] Based on the same inventive concept mentioned above, an embodiment of the present application also discloses a computer-readable storage medium, which stores at least one instruction, at least one program, code set or instruction set. The at least one instruction, at least one program, code set or instruction set can be loaded and executed by a processor to implement a robot non-destructive inspection method for high-altitude pipelines provided in the above method embodiment.

[0175] It should be understood that the "plurality" mentioned in this article refers to two or more. "And / or" describes the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0176] A person skilled in the art will appreciate that all or part of the above embodiments may be implemented through hardware or through a program to instruct related hardware, and the program may be stored in a computer-readable storage medium. The storage medium mentioned above may include, for example, various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0177] The above description is only an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A robot nondestructive testing method for high-altitude pipelines, based on a robot nondestructive testing system, the system comprising a radar device (200) for transmitting and receiving ultrasonic waves, characterized in that: The robot nondestructive testing method comprises: Acquire a number of standard images of historical standard pipelines and a number of defective images of defective pipelines, and store all of the standard images and the defective images as an initial data set; Standardizing the initial data set to obtain a standard data set; Establishing an initial pipeline detection model, and training and optimizing the initial pipeline detection model according to the standard data set to obtain a pipeline detection model; sequentially acquiring a plurality of images of the pipeline to be inspected collected by the radar device (200), and uniformly standardizing the acquired images to obtain a plurality of images to be inspected; Inputting all the images to be detected into the pipeline detection model in sequence to obtain actual detection results; The training and optimization of the initial pipeline detection model according to the standard data set to obtain the pipeline detection model includes: dividing the standard data set into a training set and a test set according to a preset ratio, wherein both the training set and the test set include the standard image and the defect image that have been uniformly standardized; Training the pipeline detection initial model using the training set; Testing and optimizing the trained initial pipeline detection model using the test set to obtain a pipeline detection model; The training of the pipeline detection initial model by using the training set includes: Input all images in the training set into the pipeline detection initial model, divide each image into a number of detection areas, and extract the local feature matrix of all the detection areas corresponding to each image; Classify all images in the pipeline inspection initial model to obtain standard images and defect images; The average value of the local feature matrices corresponding to all the standard images is calculated, and the average value is used as the standard feature matrix. The calculation formula of the standard feature matrix is ​​as follows: Among them, X 训，标准 (n,i,j) represents the local feature matrix of the nth standard image in the detection area (i,j) in the training set; represents the standard feature matrix; a n(e,f) , e=1 or 2, f=1 or 2, represents the element in the e-th row and f-th column of the local feature matrix corresponding to the n-th standard image; N represents the total number of standard images; or 2, f = 1 or 2, which means the average value of the elements in the e-th row and the f-th column in the local feature matrix corresponding to all standard images; Obtaining a defect feature matrix based on the standard feature matrix and the local feature matrix of the defect image to complete the training of the pipeline detection initial model, wherein the pipeline detection initial model includes a plurality of detection sub-models for detecting different defect types; The step of obtaining a defect feature matrix based on the standard feature matrix and the local feature matrix of the defect image to complete the training of the initial pipeline detection model includes: The local feature matrix of each detection area of ​​each defect image is respectively compared with the standard feature matrix of the corresponding detection area to obtain a first phase difference. The calculation formula of the first phase difference is: Among them, X 缺陷 (a,i,j) represents the local feature matrix of the a-th defect image in the detection area (i,j) in the training set; represents a standard feature matrix corresponding to all detection areas (i, j); d represents X 缺陷 (a,i,j) and The Euclidean distance between e,f , e=1 or 2, f=1 or 2, represents the feature matrix X 缺陷 The element in row e and column f in (a,i,j); b e,f ,e=1 or 2,f=1 or 2, represents the feature matrix The element in the e-th row and the f-th column; D1 represents the first phase difference, and its value range is [0,1]; Obtaining a feature matrix to be merged corresponding to some of the detection areas in each of the defect images according to the first phase difference, and updating the standard feature matrix according to all the feature matrices to be merged; Obtaining a suspected defect feature matrix corresponding to a certain detection area in each of the defect images according to the first phase difference; testing each of the suspected defect feature matrices, and obtaining a defect feature matrix according to all the suspected defect feature matrices that pass the test; The step of obtaining a feature matrix to be merged corresponding to certain detection areas in each defect image according to the first phase difference, and updating the standard feature matrix according to all the feature matrices to be merged includes: When the first phase difference is less than the phase difference threshold, the local feature matrix corresponding to the first phase difference is used as the feature matrix to be merged, and the standard feature matrix is ​​updated according to all the feature matrices to be merged. The calculation formula for the update is as follows: Among them, X 待合并 (n,i,j) represents the feature matrix to be merged; represents the standard feature matrix before updating; represents the updated standard feature matrix; b e,f ,e=1 or 2,f=1 or 2, represents the feature matrix The element in row e and column f in c n(e,f) , e=1 or 2, f=1 or 2, representing the element in the e-th row and f-th column of the n-th feature matrix to be merged; M represents the total number of feature matrices to be merged.

2. The robot nondestructive testing method according to claim 1, characterized in that: The step of testing each of the suspected defect feature matrices and obtaining a defect feature matrix according to all the suspected defect feature matrices that have passed the test includes: after obtaining each suspected defect feature matrix, taking the detection area corresponding to the matrix as a test area, and performing secondary segmentation on each test area to obtain a plurality of test sub-areas; Extract the test sub-feature matrix of each test sub-region, and take the average value of all the test sub-feature matrices as the test feature matrix; Calculate the phase difference parameters between each inspection feature matrix and the suspected defect feature matrix corresponding to the corresponding inspection area; If the phase difference parameter is less than a preset phase difference threshold, a defect feature matrix is ​​obtained according to all inspection sub-feature matrices corresponding to the phase difference parameter.

3. The robot nondestructive testing method according to claim 2, characterized in that: The calculation formula of the phase difference parameter is as follows: X2=X 疑似缺陷 (i,j); Among them, γ(X1,X2) represents the phase difference parameter between X1 and X2; That is, X1 represents the inspection feature matrix corresponding to the inspection area (i, j); X 疑似缺陷 (i, j) represents the suspected defect feature matrix corresponding to the inspection area (i, j); X d ', d = 1 or 2, represents the transposed matrix of the feature matrix X1; X d ' T ,d=1 or 2, represents the feature matrix X d 'The transposed matrix; w represents the feature matrix X d ,d=1 or 2, the number of rows; X 检验子 (u,v) represents the inspection sub-region feature matrix corresponding to the inspection sub-region (u,v); U represents the number of inspection sub-regions included in the rows of the inspection region (i,j); V represents the number of inspection sub-regions included in the columns of the inspection region (i,j).

4. The robot nondestructive testing method according to claim 3, characterized in that: The calculation formula of the defect feature matrix is ​​as follows: Among them, X 缺陷 (i, j) represents the defect feature matrix corresponding to the inspection area (i, j); X 检验子 (u,v),u≤m,v≤m, represents the feature matrix corresponding to the inspection sub-region located in the uth row and vth column in the inspection region corresponding to the suspected defect feature matrix.

5. An intelligent terminal, characterized in that: It comprises a memory and a processor, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the robot nondestructive testing method according to any one of claims 1 to 4.

6. A robot nondestructive testing system comprising the intelligent terminal according to claim 5, characterized in that: include: The robot (100) has a magnetic suction cup installed at the bottom thereof and is used for crawling on the outer wall of the pipeline to be inspected along the length direction of the pipeline to be inspected; A radar device (200) is arranged on the robot (100) and transmits and receives ultrasonic waves to the pipeline to be inspected, thereby obtaining a plurality of images of the pipeline to be inspected; The intelligent terminal (300) is connected in communication with the robot (100) and the radar device (200), and is used to acquire a plurality of images of the pipeline to be inspected collected by the radar device (200), and obtain actual inspection results of the pipeline to be inspected based on the acquired plurality of images.

7. A computer-readable storage medium, characterized in that: The readable storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by a processor to implement the robot nondestructive testing method as described in any one of claims 1 to 4.

Images