Pipeline burial depth detection method and system based on image recognition

Abstract

The invention belongs to the technical field of pipeline construction, and particularly discloses a pipeline burial depth detection method and system based on image recognition. The detection method comprises the following steps: acquiring contours of a to-be-detected pipeline and a detection plate, and adjusting the detection plate to enable center lines of the two to coincide; the background plate is arranged above the detection plate in parallel, and the levelness of the detection plate is adjusted based on the color recognition technology and a deep learning model, so that the detection plate is located on the horizontal section of the to-be-detected pipeline; adjusting the background plate according to the coordinates of the positioner; respectively acquiring a first actual burial depth and a second actual burial depth through a detection plate laser transmitter and a detection light beam, and averaging the first actual burial depth and the second actual burial depth as an actual measurement value; and finally, comparing an actual measurement value with the theoretical burial depth to judge whether the burial depth is qualified, and determining whether reworking is needed according to a difference value between the two. Based on an image recognition technology, the invention provides a brand-new pipeline burial depth detection method capable of effectively solving the precision problem, and ensures that the pipeline burial depth is accurately detected from the beginning of construction of pipeline laying.

Description

Technical Field

[0001] The invention relates to the technical field of gas pipeline construction, and in particular to a pipeline buried depth detection method based on image recognition. Background Art

[0002] In the process of modern infrastructure construction and maintenance, pipeline engineering occupies a pivotal position, widely involving many key areas such as urban water supply and drainage, gas transmission, petrochemical pipelines, etc. The precise buried depth of the pipeline is not only related to the quality of engineering construction, but also one of the core elements to ensure the subsequent long-term safety of the pipeline. The commonly used pipeline buried depth detection methods during the installation process include level measurement method, total station trigonometric height measurement method and simple rope measurement method.

[0003] The defects of the above detection methods are: when the level measurement method is actually operated, the cumbersome measurement and calculation process greatly increases the probability of human error. For example, the reading of the level rod is easily deviated due to the parallax of the observer, and it is quite difficult to find a stable reference point; although the total station trigonometric height measurement method is efficient and convenient, it can face the complex and changeable construction site environment. The atmospheric refraction coefficient causes the final result to deviate greatly from the true value, especially in long-distance and large height difference measurement tasks. The error can even reach several centimeters, which seriously deviates from the requirements of precise construction; as for the simple rope measurement method, because it relies too much on manual vertical placement operations, the verticality of the rope cannot be guaranteed in complex working conditions such as deep foundation pits and narrow trenches, and the scale reading error is significant. Therefore, there is an urgent need for a new pipeline depth detection method that can effectively overcome the problem of accuracy, to ensure that the pipeline laying is accurate from the beginning of construction, and to ensure the subsequent long-term operation and maintenance. Summary of the invention

[0004] In view of this, according to one aspect of the present application, a pipeline buried depth detection method based on image recognition is provided, and the detection method comprises the following steps: S1, placing the detection plate on the top of the pipeline to be tested, respectively obtaining the contours of the pipeline to be tested and the detection plate, and adjusting the detection plate according to the distance between the contour edges of the pipeline to be tested and the detection plate, so that the center line of the detection plate and the center line of the pipeline to be tested are on the same straight line; S2, place the background plate parallel to the detection plate, and adjust the horizontality of the detection plate based on color recognition technology and deep learning model so that the detection plate is located on the horizontal section of the pipeline to be tested; S3, after the detection plate is adjusted to be horizontal, the spatial position coordinates of the detection plate are obtained by a first locator at the center of the bottom of the detection plate, and the spatial position coordinates of the background plate are obtained by a second locator at the center of the background plate; the background plate is adjusted according to the difference between the spatial position coordinates of the two, so that the center line of the background plate and the center line of the detection plate are located on the same straight line; S4, using the laser transmitter at the top center of the detection board to emit a laser beam to the receiver of the background board to obtain a first actual buried depth of the pipeline to be measured, and using the detection board to emit a detection beam to the background board to obtain a second actual buried depth of the pipeline to be measured, and taking the average of the first actual buried depth and the second actual buried depth as the actual measured value; S5, judging whether the burial depth is qualified according to the difference between the actual measured value and the theoretical burial depth, specifically: If the difference between the actual measured value and the theoretical burial depth is within the threshold range, it is judged as qualified, otherwise it is judged as unqualified, and whether rework is required is determined based on the difference between the two.

[0005] Preferably, the adjustment process of the horizontality of the detection plate is: Step 1: Build a deep learning model, collect the actual projection and actual color data of the emission light beam of the detection board on the background board under different conditions, pre-process the actual projection and actual color data to obtain the original data set, divide the original data set into a test set and a training set, use the training set to train the deep learning model, and evaluate the model performance through the test set; Step 2: Adjust the shape and color of the light source of the detection board to emit light beams of different colors and shapes to the background board; Step 3: The background plate correspondingly collects the actual projection and actual color data of the light beam emitted by the detection plate each time; Step 4: Input the actual projection and actual color data collected by the background plate into the trained deep learning model; Step 5: According to the theoretical buried depth of the pipeline, calculate the theoretical projection and theoretical color data of each emission on the background board; and calculate the deviation values ​​of the actual projection and actual color data from the theoretical projection and theoretical color data respectively, and judge whether the horizontality of the detection board is qualified according to the deviation value: If the deviation value is within the threshold range, the horizontality of the detection plate is determined to be qualified. If the deviation value is greater than the threshold, the detection plate is adjusted accordingly until the horizontality of the detection plate is qualified.

[0006] Preferably, the actual projection and the theoretical projection respectively include the shape and area of ​​the projection; The actual color data and the theoretical color data include the saturation, chroma and brightness of the color respectively.

[0007] According to another aspect of the present application, a pipeline buried depth detection system based on image recognition is provided. The detection system adopts a pipeline buried depth detection method based on image recognition. The detection system includes: A detection plate is installed on a horizontal section of the top of the pipeline to be tested; The background plate is used to assist the detection plate in completing the detection process and is located directly above the detection plate; A level adjustment module is installed at the bottom of the detection plate, the level adjustment module is buckled on the top of the pipeline to be tested, and a contact sensor is provided on the side of the level adjustment module facing the pipeline to be tested; The installation module is located on the ground of the pipeline to be tested and is used to adjust the positions of the background plate and the detection plate; The acquisition module includes: an image acquisition unit for acquiring the contours of the pipeline to be tested and the detection plate, a spatial position acquisition unit for acquiring the spatial position coordinates of the detection plate and the background plate, a distance detection unit for measuring the first actual buried depth of the pipeline to be tested, and an image recognition unit for realizing color recognition and a second actual buried depth; The image acquisition unit is located directly above the detection board; The spatial position acquisition unit includes a first locator installed at the center of the bottom of the detection plate, and a second locator installed at the center of the background plate; The distance detection unit and the image recognition unit are respectively installed on the detection plate and the background plate, and the image recognition unit is arranged on the peripheral side of the distance detection unit; The processing control module is respectively connected to the detection board, the contact sensor, the installation module, the level adjustment module, the image acquisition unit, the first locator, the second locator, the distance detection unit and the image recognition unit.

[0008] Preferably, the detection board is an intelligent light-emitting board; The center line of the detection plate coincides with the center line of the level adjustment module; The distance detection unit includes a laser transmitter installed at the center of the top of the detection plate and a receiver installed at the center of the bottom of the background plate; The image recognition unit includes a light source of the detection plate and a collector arranged on the background plate; The laser emitter is located at the center of the light source; The collector is located on the periphery of the receiver, and the collector is arranged corresponding to the light source of the detection board; The laser transmitter, the receiver, the light source and the collector are respectively connected to the processing control module; The contact sensor is mounted at the center of the bottom of the level adjustment module; The second locator is installed at the center of the background plate.

[0009] Preferably, the mounting module is a mechanical clamp.

[0010] Preferably, the level adjustment module comprises a level adjustment portion, a clamping portion and a mounting plate which are sequentially connected from bottom to top; The top and bottom of the mounting plate are respectively provided with a first concave mounting opening and a second concave mounting opening; The first locator is located in the first concave mounting opening; The contact sensor is installed in the second concave installation opening; The detection plate is installed on the top of the mounting plate; The clamping parts are respectively located on the outer sides of the contact sensors, and at least two clamping parts are provided correspondingly, and the two clamping parts are clamped on both sides of the pipeline to be tested by moving closer to or farther away from each other; The level adjustment part is an automatic telescopic device, and one telescopic end faces underground.

[0011] Preferably, an adjustable telescopic rod is respectively provided between the clamping portion and the mounting plate; The adjustable telescopic rod is arranged along the radial direction of the pipeline to be tested, the telescopic ends of the adjustable telescopic rod are arranged facing each other and are respectively connected to the clamping parts, the opposite ends of the adjustable telescopic rod are respectively connected to the mounting plates, and the adjustable telescopic rod is an automatic telescopic rod.

[0012] Preferably, the clamping portion includes a first clamping portion and a second clamping portion with openings facing each other; The first clamping portion and the second clamping portion respectively include a connecting plate and a clamping plate; The connecting plates are respectively vertically arranged on the mounting plates, one end of the connecting plates is respectively connected to the telescopic end of the adjusting telescopic rod, and are driven by the adjusting telescopic rod to move along the radial direction of the pipeline to be tested on the mounting plate; The clamping plates are respectively installed on opposite sides of the connecting plate, and the opposite sides of the clamping plates are respectively arc-shaped.

[0013] Compared with the prior art, the beneficial effects of this application are: During the measurement process, the present application uses an image acquisition unit to accurately obtain the contours of the pipeline to be measured and the detection plate, and accurately adjusts the position of the detection plate according to the distance between the edges of the contours of the two, ensuring that the center line of the detection plate completely coincides with the center line of the pipeline to be measured. This effectively avoids the depth error caused by the deviation of the measuring point. For example, the distance measured on the side of the pipeline will be smaller than the distance measured at the vertex, so that the distance measured by the laser transmitter emitting laser from the top center of the detection plate can accurately reflect the actual depth of the pipeline to be measured, greatly improving the measurement accuracy. At the same time, the horizontality of the detection plate is finely adjusted using color recognition technology and deep learning models so that it is accurately located on the horizontal section of the pipeline to be measured, further ensuring the accuracy of the measurement benchmark and reducing the source of error.

[0014] This application obtains the spatial position coordinates of the two through the first locator at the bottom of the detection plate and the second locator at the center of the background plate, and accurately adjusts the background plate according to the coordinate difference to achieve that the background plate is parallel to the detection plate and the center lines coincide. This precise positioning method significantly improves the accuracy of collecting projection and color data, provides a more reliable data foundation for deep learning models, thereby enhancing the accuracy of the model, ensuring that the buried depth measurement is based on an accurate relative position relationship, and effectively guarantees the reliability of the measurement results.

[0015] In this application, color recognition technology and laser measurement are combined to obtain multiple actual burial depth detection data (i.e., the first actual burial depth and the second actual burial depth) and the average value is taken as the actual measurement value, which effectively improves the accuracy of pipeline burial depth detection. During the entire detection process, from the contact adjustment between the detection plate and the pipeline to the calibration of the horizontality and centerline position, each link is closely connected and works together to ensure that each step of the operation provides strong support for the accuracy of the final measurement result, effectively reducing the cumulative error and making the measurement result closer to the true value.

[0016] Various operations such as image acquisition, detection data acquisition, and adjustment of the horizontality of the detection board are automated with the help of deep learning models. The horizontality of the detection board is automatically judged based on the deviation between the actual projection and actual color data and the theoretical data, and adjustments are made. The horizontality adjustment part of the horizontality adjustment module and the adjustment telescopic rod of the clamping part are both automatic devices, which greatly reduces manual operation errors and significantly improves detection efficiency. In addition, the clamping part of the detection board can flexibly change the spacing by adjusting the telescopic rod, and can firmly clamp the pipelines to be tested of different diameters, so that it can be widely adapted to the buried depth detection needs of pipelines of various specifications, and can play an efficient and accurate detection role in different pipeline engineering scenarios, effectively reducing engineering costs and time costs, and improving the overall quality and efficiency of engineering construction and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of the positions of the detection plate, background plate and pipeline to be tested in this application; Figure 2 It is a vertical cross-sectional schematic diagram of the detection board in this application; Figure 3 A bottom view schematic diagram of the detection board in this application; In the figure: 1. pipeline to be tested; 2. level adjustment part; 3. connecting plate; 4. clamping plate; 5. contact sensor; 6. mounting plate; 7. first positioner; 8. detection plate; 9. laser transmitter; 10. background plate; 11. receiver; 12. second positioner; 13. slider; 14. adjustment telescopic rod; 15. mounting slot. DETAILED DESCRIPTION

[0018] As attached Figure 1A pipeline buried depth detection method based on image recognition is shown, and the detection method comprises the following steps: S1, placing the detection plate on the top of the pipeline to be tested, respectively obtaining the contours of the pipeline to be tested and the detection plate, and adjusting the detection plate according to the distance between the contour edges of the pipeline to be tested and the detection plate, so that the center line of the detection plate and the center line of the pipeline to be tested are on the same straight line; S2, place the background plate parallel to the detection plate, and adjust the horizontality of the detection plate based on color recognition technology and deep learning model so that the detection plate is located on the horizontal section of the pipeline to be tested; S3, after the detection plate is adjusted to be horizontal, the spatial position coordinates of the detection plate are obtained by a first locator at the center of the bottom of the detection plate, and the spatial position coordinates of the background plate are obtained by a second locator at the center of the background plate; the background plate is adjusted according to the difference between the spatial position coordinates of the two, so that the center line of the background plate and the center line of the detection plate are located on the same straight line; S4, using the laser transmitter at the top center of the detection board to emit a laser beam to the receiver of the background board to obtain a first actual buried depth of the pipeline to be measured, and using the detection board to emit a detection beam to the background board to obtain a second actual buried depth of the pipeline to be measured, and taking the average of the first actual buried depth and the second actual buried depth as the actual measured value; S5, judging whether the burial depth is qualified according to the difference between the actual measured value and the theoretical burial depth, specifically: If the difference between the actual measured value and the theoretical burial depth is within the threshold range, it is judged as qualified, otherwise it is judged as unqualified, and whether rework is required is determined based on the difference between the two.

[0019] It should be noted that: In the present application, the contour of the pipeline to be tested refers to the contour of a certain section of the pipeline to be tested, rather than the contour of the entire pipeline to be tested. A fixed boundary is circled on the ground, and then the contour of the pipeline to be tested under the boundary is obtained by an image acquisition device; generally, four points of a certain section of the pipeline to be tested, two points form a group and are located on opposite sides of the pipeline to be tested, and the contour is composed of four points; In this application, the center line of the detection plate refers to the vertical perpendicular line of the geometric center of the detection plate; the center line of the pipeline to be tested refers to the vertical perpendicular line of the geometric center of the contour projection area of ​​a certain section of the pipeline to be tested; the center line of the background plate refers to the vertical perpendicular line of the geometric center of the background plate.

[0020] Furthermore, the adjustment process of the horizontality of the detection plate is as follows: Step 1: Build a deep learning model, collect the actual projection and actual color data of the emission light beam of the detection board on the background board under different conditions, pre-process the actual projection and actual color data to obtain the original data set, divide the original data set into a test set and a training set according to a reasonable ratio (such as 7:3 or 8:2), use the training set to train the deep learning model, and evaluate the model performance through the test set; Step 2: Adjust the shape and color of the light source of the detection board to emit light beams of different colors and shapes to the background board; Step 3: The background plate collects the actual projection and actual color data of the light beam emitted by the detection plate each time; Step 4: Input the actual projection and actual color data collected by the background plate into the trained deep learning model; Step 5: According to the theoretical buried depth of the pipeline, calculate the theoretical projection and theoretical color data of each emission on the background board; and calculate the deviation values ​​of the actual projection and actual color data from the theoretical projection and theoretical color data respectively, and judge whether the horizontality of the detection board is qualified according to the deviation value: If the deviation value is within the threshold range, the horizontality of the detection plate is determined to be qualified. If the deviation value is greater than the threshold, the detection plate is adjusted accordingly until the horizontality of the detection plate is qualified.

[0021] It should be noted that: The qualified threshold of buried depth in this application is determined according to the requirements of pipeline construction specifications and the actual allowable error range of the project. Since the buried depth error allowed by relevant standards in urban gas pipeline construction is within ±0.5 cm, the threshold of the difference between the actual measured value and the theoretical buried depth is set to ±0.5 cm. For the threshold of the horizontality of the detection plate, through a large number of experiments and simulation analysis, considering the requirements for pipeline installation accuracy and the propagation of measurement errors, it is determined that when the shape deviation value of the actual projection and the theoretical projection is within 5%, the area deviation value is within 3%, and the saturation, chromaticity and brightness deviation values ​​of the color data are within 0.05, the horizontality of the detection plate is qualified.

[0022] The process of building a deep learning model is: 1) Data collection: Set up the collection environment: ensure that the detection plate and the background plate are stable and under appropriate lighting conditions to reduce interference from external factors and make the collected data more accurate and consistent.

[0023] Multiple acquisitions: According to the operation in step 2, the detection board emits light beams (such as light beams of different shapes and colors) to the background board multiple times. The background board uses corresponding image acquisition equipment (such as high-definition cameras, etc.) to completely acquire the projection formed by each light beam emitted by the detection board and the corresponding color information, and records them as image data. These image data are accompanied by corresponding color feature descriptions, etc., to form an original data set.

[0024] 2) Data preprocessing: Image normalization: For the collected image data, uniformly adjust the image size, pixel range, etc. For example, normalize all images to a specific width and height pixel value (such as 500×500 pixels) to ensure that the data collected in different batches are consistent in size, which is convenient for subsequent processing.

[0025] Color correction: Since color deviation may occur due to factors such as lighting during the acquisition process, the image color is calibrated to a standard color space (such as the RGB standard color gamut) through a color correction algorithm to ensure the accuracy of the color data.

[0026] Data enhancement: To expand the data set and improve the generalization ability of the model, operations such as rotation, flipping, and adding noise are used to appropriately transform the image to generate more image data with different forms but the same essential features. For example, the original image can be rotated 10 degrees clockwise, flipped horizontally, etc., and its corresponding color data records can be updated accordingly.

[0027] Divide the data set: Divide the original data set after the above processing into a test set and a training set according to a certain ratio (commonly used ratios such as 8:2 or 7:3, etc.). The test set is used to evaluate the performance of the model later, and the training set is used in the model training process.

[0028] 3) Feature extraction: Choose a suitable feature extraction algorithm: For example, you can use the classic architecture of convolutional neural network (CNN) (such as AlexNet, VGG, etc.) to automatically extract features from images. The convolution layer in these network structures can capture local features of different scales in the image, such as edges and textures, through convolution kernels of different sizes. The pooling layer further compresses and refines the features, reducing the data dimension while retaining key features.

[0029] Extract color-related features: For color data, you can use methods such as statistical color histograms to obtain features such as the distribution ratio of each color channel (R, G, B) in the image. You can also use more complex methods such as color moments to describe color features. These features, together with visual features such as image shape, can be used as input for subsequent model learning.

[0030] 4) Model training: Use a common deep learning model for image classification or regression, such as ResNet's deep residual network. Through residual connections, the model can better learn the mapping relationship between image features and target output (i.e., the feature representation corresponding to the theoretical projection and theoretical color data).

[0031] Define the loss function: Depending on the specific task requirements, if the goal is to predict the exact position of the projection and the color value, the mean square error (MSE) can be used as the loss function to measure the error between the theoretical projection and theoretical color data predicted by the model and the actual projection and actual color data.

[0032] Set the optimization algorithm: commonly used ones such as stochastic gradient descent (SGD) can dynamically adjust the model's weight update strategy according to the set parameters such as the learning rate, so that the model can continuously optimize and converge in the direction of reducing the loss function.

[0033] Training process: Input the training set data into the selected model, and iterate continuously in multiple training rounds (epochs) according to the set batch size (such as inputting 32 images and their corresponding color data each time). Use the back propagation algorithm to calculate the gradient and update the model's weight parameters, so that the model gradually learns the mapping rules from input images and color data to output theoretical projections and theoretical color data.

[0034] 5) Model evaluation and verification (can be performed during training): Test on the test set: After a certain number of training rounds, input the test set data into the trained model to obtain the model's prediction results (i.e. the predicted theoretical projection and theoretical color data).

[0035] Evaluation index calculation: Calculate relevant evaluation indexes. For example, for regression tasks, if the prediction is continuous values ​​such as projection position coordinates, the root mean square error (RMSE) is calculated to measure the degree of deviation between the predicted value and the true value. The prediction of color data is similar. These indicators are used to judge the performance of the model. If the performance does not meet expectations, the model parameters can be adjusted, the training frequency can be increased, or the data preprocessing can be optimized to retrain the model.

[0036] Among them, in one embodiment of the present application, the ResNet-50 architecture is finally selected because it has good performance in processing image feature extraction and regression tasks. After multiple verifications, it can effectively learn the mapping relationship between image features and buried depth data in this pipeline buried depth detection scenario. During training, the learning rate is set to 0.001, the training rounds (epochs) are 100, and the batch size is 64. During the model verification process, if the root mean square error (RMSE) is greater than the set threshold (such as 0.5 cm), the model is optimized by increasing the amount of training data, adjusting the number of network layers, or optimizing the convolution kernel parameters, and retraining and verification until a satisfactory accuracy index is achieved.

[0037] Further, the actual projection and theoretical projection include projection shape and area respectively; The actual color data and the theoretical color data include the saturation, chroma and brightness of the color respectively.

[0038] A pipeline buried depth detection system based on image recognition, the detection system comprising: The detection plate 8 is installed on the horizontal section of the top of the pipeline 1 to be detected; The background plate 10 is used to assist the detection plate 8 in completing the detection process and is located directly above the detection plate 8; A level adjustment module is installed at the bottom of the detection plate 8, and the level adjustment module is buckled on the top of the pipeline to be tested 1. A contact sensor 5 is provided on the side of the level adjustment module facing the pipeline to be tested; An installation module is located on the ground of the pipeline to be tested and is used to adjust the positions of the background plate 10 and the detection plate 8; The acquisition module includes: an image acquisition unit for acquiring the contours of the pipeline to be tested and the detection plate, a spatial position acquisition unit for acquiring the spatial position coordinates of the detection plate and the background plate, a distance detection unit for measuring the first actual buried depth of the pipeline to be tested, and an image recognition unit for realizing color recognition and a second actual buried depth; The image acquisition unit is located directly above the detection board; The spatial position acquisition unit includes a first locator 7 installed at the center of the bottom of the detection plate 8, and a second locator 12 installed at the center of the top of the background plate 10; The distance detection unit and the image recognition unit are respectively mounted on the detection plate 8 and the background plate 10, and the image recognition unit is arranged on the peripheral side of the distance detection unit; The processing control module is respectively connected to the detection board 8, the contact sensor 5, the installation module, the level adjustment module, the image acquisition unit, the first locator 7, the second locator 12, the distance detection unit and the image recognition unit.

[0039] It should be noted that: In the present application, the detection plate 8 is placed on the horizontal section of the top of the pipeline 1 to be detected to provide a detection reference; The background plate 10 is placed directly above the detection plate 8 for auxiliary detection.

[0040] The level adjustment module is buckled on the top of the pipeline to be tested 1 and is used to carry the detection plate 8. The level adjustment module is used to adjust the levelness of the detection plate 8, monitor the fitting state with the pipeline to be tested 1 through the contact sensor 5 arranged at the bottom, and accurately obtain the spatial position coordinates of the detection plate 8 through the first locator 7 arranged on the inner side of the top.

[0041] The installation module is located on the ground of the pipeline to be tested, and is used to install the detection plate 8 and the level adjustment module on the top of the pipeline to be tested 1, and to adjust the position of the background plate so that it is placed parallel to and directly above the detection plate 8; The image acquisition unit in the acquisition module is located directly above the pipeline 1 to be tested and is used to acquire the contour image of the pipeline 1 to be tested and the detection plate 8; in the present application, the image acquisition unit is any one of an industrial 3D scanner or a 3D line laser profile measuring instrument; The spatial position acquisition unit in the acquisition module includes a first locator 7 located at the center of the bottom of the detection plate 8 and a second locator 12 located at the center of the background plate 10, which are used to obtain the spatial coordinates of the detection plate 8 and the background plate 10 respectively; The distance detection unit in the acquisition module is arranged at the top center of the detection plate 8, and is used to measure the first actual buried depth of the pipeline 1 to be tested; the image recognition unit is used to recognize the color and calculate the second actual buried depth; The processing control module is the "intelligent center" of the entire system. In this application, the processing control module is a programmable controller, which is connected to the image acquisition unit to receive the contour image information of the pipeline to be tested and the detection plate collected by the image acquisition unit. Based on these raw materials, the controller uses the image analysis algorithm in S1 to measure the distance between the contour edges of the pipeline to be tested 1 and the detection plate 8, and then accurately adjusts the position of the detection plate 8 so that the center line of the detection plate 8 coincides with the center line of the pipeline to be tested 1; The processing control module is connected to the image recognition unit to obtain the color recognition data collected by the collector on the background plate 10. In S2, the processing control module uses the color change information fed back by the image recognition unit and combines it with the deep learning model to issue dynamic adjustment instructions for the shape and color of the detection plate 8 to calibrate the horizontal state of the detection plate 8 so that it is located on the horizontal section of the pipeline to be tested.

[0042] The processing control module is connected to the spatial position acquisition unit to obtain the spatial position coordinates of the detection board 8 and the background board 10 fed back by the first locator 7 (located at the center of the side of the horizontal adjustment module facing the detection board / the center of the bottom of the detection board) and the second locator 12 (located at the center of the background board). In S3, according to the difference between the coordinates of the two, an instruction is sent to the installation module to adjust the posture of the background board so that the background board 10 is parallel to the detection board 8 and the center lines coincide.

[0043] The processing control module is in communication with the distance detection unit, receives the first actual buried depth data of the pipeline to be tested measured by the distance detection unit, and combines it with the second actual buried depth data inferred from the image recognition unit according to the mapping relationship between color and depth, and completes the average value calculation in S4 to obtain the actual measurement value of the pipeline 1 to be tested.

[0044] The processing control module is connected to the contact sensor 5 of the level adjustment module to sense the degree of fit between the level adjustment module and the pipeline 1 to be tested in real time, so as to ensure the accuracy of the detection benchmark during the level adjustment of the detection plate 8, and assist in completing the relevant requirements for the level adjustment of the detection plate in the detection method. By working closely with these units, the processing control module realizes the orderly advancement and precise execution of each step of the pipeline buried depth detection method based on image recognition, ensuring the efficient and stable operation of the entire detection system.

[0045] Furthermore, the detection board 8 is an intelligent light-emitting board; The center line of the detection plate 8 coincides with the center line of the level adjustment module; The distance detection unit includes a laser transmitter 9 installed at the center of the top of the detection plate and a receiver 11 installed at the center of the bottom of the background plate; The image recognition unit includes a light source of the detection plate 8 and a collector installed at the bottom of the background plate 10; The laser emitter 9 is located at the center of the light source; The collector is located on the peripheral side of the receiver 11; The laser transmitter 9, the receiver 11, the light source and the collector are respectively connected to the processing control module; The contact sensor 5 is installed at the center of the bottom of the level adjustment module; The second locator 12 is installed at the center of the background plate 10 .

[0046] It should be noted that: In one embodiment of the present application, the intelligent lighting of the detection board 10 (i.e., the intelligent lighting board) is mainly realized based on light emitting diode (LED) technology and intelligent control, wherein the LED is the core component. The color is adjusted by a program, for example, various colors can be mixed by changing the ratio of the three colors in the RGB (red, green, blue) color mode.

[0047] In terms of shape adjustment, the smart light-emitting board has multiple small light-emitting units (i.e. light-emitting diodes (LEDs)) built into the board body. These light-emitting units are combined into different shapes through program control, such as being arranged into circles, squares or other custom shapes.

[0048] In terms of intelligent control, there is a control circuit inside the intelligent light-emitting board. The control circuit is connected to the processing control module, and the control circuit can receive signals from the processing control module. After receiving the instruction to adjust the color or shape, the control circuit can control the light and dark of these light-emitting units to spell out different shapes; and correspondingly control the LED chip to make electrons and holes recombine, release energy in the form of photons, and thus emit light. LED chips of different materials can emit light of different colors. By adjusting the brightness of LED lamp beads of different colors, a variety of colors can be mixed.

[0049] Furthermore, the installation module is a mechanical gripper, and the controllers of the mechanical grippers are respectively connected to the processing control modules.

[0050] It should be noted that the synchronous movement of the detection plate and the level adjustment module, as well as the movement of the background plate, are achieved by mechanical grippers, which are commercially available.

[0051] Further, the level adjustment module comprises a level adjustment portion, a clamping portion and a mounting plate 6 which are sequentially connected from bottom to top; The top and bottom of the mounting plate 6 are respectively provided with a first concave mounting opening and a second concave mounting opening; The first locator 7 is located in the first concave mounting opening; The contact sensor 5 is installed in the second concave installation opening; The depth of the first concave installation opening matches the overall height of the first locator 7 to prevent affecting the installation of the detection board 8; The depth of the second concave installation opening matches the overall height of the contact sensor 5, that is, after the contact sensor 5 is installed, it is parallel to the bottom surface of the installation plate 6; The laser transmitter 9 is installed on the top of the detection plate 8 in the exact center; The receiver 11 is installed at the exact center of the bottom of the background plate 10 , and the second locator 12 is installed at the exact center of the top of the background plate 10 .

[0052] The detection plate 8 is installed on the top of the mounting plate 6, and when the detection plate 8 is installed on the top of the mounting plate 6, the first locator 7 is inserted into the first concave mounting opening; The clamping parts are respectively located on the outside of the contact sensor 5, and at least two clamping parts are correspondingly provided. The two clamping parts are clamped on both sides of the pipeline to be tested 1 by moving closer to or farther away from each other, so as to fix the detection plate 8 on the horizontal section of the top of the pipeline to be tested 1 through the horizontal adjustment module; The level adjustment part 2 is an automatic telescopic device, and one end of the telescopic device faces underground. In one embodiment of the present application, the level adjustment part 2 is an electric telescopic rod, and its controller is electrically connected to the processing control module; Furthermore, an adjustable telescopic rod 14 is provided between the clamping portion and the mounting plate 6; The adjustable telescopic rod 14 is arranged along the radial direction of the pipeline 1 to be tested, the telescopic ends of the adjustable telescopic rod 14 are arranged facing each other and are respectively connected to the clamping parts, and the opposite ends of the adjustable telescopic rod 14 are respectively connected to the mounting plate 6, the adjustable telescopic rod 14 is an automatic telescopic rod, and the controller of the adjustable telescopic rod 14 is electrically connected to the processing control module; Further, the clamping portion includes a first clamping portion and a second clamping portion with openings facing each other; The first clamping part and the second clamping part respectively include a connecting plate 3 and a clamping plate 4; The connecting plates 3 are respectively vertically arranged on the mounting plates 6, and the connecting plates 3 are respectively connected to the telescopic ends of the adjusting telescopic rods 14, and are driven by the adjusting telescopic rods 14 to move along the radial direction of the pipeline 1 to be tested on the mounting plates 6; The clamping plates are respectively installed on opposite sides of the connecting plate, and the opposite sides of the clamping plates are respectively arc-shaped.

[0053] It should be noted that: In the present application, the size of the mounting plate 6 is larger than the size of the detection plate 8; the outer surface of the mounting plate 6 is subjected to anti-corrosion treatment; The plate body of the mounting plate 6 is a metal plate, and the bottom of the detection plate 8 is an electromagnetic plate body. The control switch of the electromagnetic plate body is electrically connected to the processing control module. The processing control module controls whether the detection plate 8 has magnetism by controlling the power-on status of the detection plate 8, so as to achieve magnetic connection with the mounting plate 6; In the present application, in order to realize that the connecting plate 3 can move on the mounting plate 6 and ensure its stability, as shown in the attached Figure 2 and 3As shown, in one embodiment of the present application, two groups of clamping parts (i.e., 4 in total) are arranged at intervals along the extension direction of the mounting plate 6 (i.e., the axial direction of the pipeline 1 to be tested), and two mounting grooves 15 are correspondingly arranged at the bottom of the mounting plate 6, and the mounting grooves 15 are arranged along the longitudinal direction of the mounting plate 6 (i.e., the radial direction of the pipeline 1 to be tested), and two adjustable telescopic rods 14 are arranged in one of the mounting grooves 15, and the two adjustable telescopic rods 14 are respectively arranged along the extension direction of the mounting groove 15, and the opposite ends of the two adjustable telescopic rods 14 are respectively fixed to the ends of the mounting groove 15, and the opposite ends of the two adjustable telescopic rods 14 are telescopic ends, which are respectively fixedly connected with the connecting plate 3, and the top of the connecting plate 3 is vertically connected to the adjustable telescopic rod 14; The two side walls of the installation groove 15 are correspondingly provided with sliding grooves, and the two sides of the connecting plate 3 at one end of the installation groove are correspondingly provided with sliders 13, the sliders 13 are respectively connected to the opposite sides of the same connecting plate 3, and the sliders 13 are slidably connected to the sliding grooves; The bottom end of the connecting plate 3 is fixedly connected to the top end of the levelness adjusting part 2, wherein the center lines of the levelness adjusting part 2 and the connecting plate 3 are located on the same straight line; the opposite sides of the two connecting plates 3 located in the same mounting groove 15 are respectively connected to the clamping plates 4, and the opposite sides of the two clamping plates 4 are arc-shaped surfaces.

[0054] This application is also applicable to the later buried depth monitoring after the construction of the pipeline to be tested is completed, which is realized by an electromagnetic wave transmitter and a matching electromagnetic wave receiver. The specific implementation method is: During the construction process, after the difference between the actual buried depth and the theoretical buried depth of the pipeline to be tested is found to be within the qualified buried depth threshold range, the processing module controls the magnetic plate at the bottom of the detection plate to be powered off, and then the detection plate and the first positioner are taken out together by the mechanical gripper, and the electromagnetic wave transmitter is placed in the first concave installation opening, and then the entire level adjustment module and the electromagnetic wave transmitter are buried together with the pipeline to be tested, and the electromagnetic wave receiver is installed on the ground near the pipeline to be tested, and attention is paid to the waterproof and anti-corrosion treatment of the electromagnetic wave receiver and the electromagnetic wave; the electromagnetic wave is sent to the corresponding electromagnetic wave receiver through the battery transmitter, and the electromagnetic wave is recorded, and the electromagnetic wave is used as a reference; In the later monitoring process, when extreme weather occurs or major environmental changes occur in the area where the pipeline to be tested is located, the processing control module sends electromagnetic waves to the corresponding electromagnetic wave receiver through the electromagnetic wave transmitter. When the electromagnetic wave data emitted by the electromagnetic wave transmitter to the electromagnetic wave receiver deviates greatly from the reference electromagnetic wave and exceeds the safety range, the processing control module issues an early warning prompt to promptly discover whether there is a drastic change in the buried depth of the pipeline to be tested.

[0055] The specific testing process of this application is: 1. Make the center line of the pipeline 1 to be tested and the center line of the detection plate on the same straight line. The specific process is as follows: The contour of the pipeline to be tested is obtained through the image acquisition unit, and its center line in the vertical direction is determined through the algorithm processing in the processing control module; the detection plate 8 is powered on, so that the detection plate 8 is installed on the top of the level adjustment module, and the detection plate 8 and the level adjustment module are lowered. After the level adjustment module contacts the pipeline 1 to be tested, since the contact sensor 5 is connected to the processing control module, a signal is sent to the processing control module. At this time, the processing control module controls the image acquisition unit to collect the contours of the detection plate 8 and the contours of the pipeline 1 to be tested. According to the distance between the contour of the detection plate 8 and each side of the contour of the pipeline 1 to be tested, the corresponding control installation module adjusts the detection plate 8 and the level adjustment module so that the center line of the detection plate 8 and the level adjustment module is on the same straight line with the center line of the pipeline 1 to be tested; ensure that the detection reference is accurate, and this process is achieved using a high-precision image analysis algorithm; 2. Adjust the horizontality of the detection plate so that the detection plate 8 is located on the horizontal section of the pipeline 1 to be tested, specifically: The processing control module emits light beams of different shapes and colors to the background plate 10 by changing the light-emitting shape and light-emitting color of the detection plate 8 for multiple times. The collector on the background plate 10 responds by collecting the data of the light-emitting shape and projection area each time, as well as the synchronized color data. The collected projection and color data are pre-processed and divided into a training set, a test set and a training learning model. Then, the new collected data is input to obtain the actual projection and actual color. At the same time, the theoretical projection and theoretical color data are calculated according to the theoretical burial depth. The detection plate 8 is adjusted according to the deviation value until the deviation value is within the threshold range to determine that the horizontality of the detection plate 8 is qualified. At this time, the detection plate 8 is located on the horizontal section of the pipeline 1 to be tested. 3. Through the coordinate deviation of the spatial position coordinates of the first locator 7 and the spatial position coordinates of the second locator 12, the processing control module controls the installation module to adjust the background plate 10 accordingly, so that the center line of the background plate 10, the center line of the detection plate 8 and the center line of the pipeline 1 to be tested are located on the same straight line; 4. The laser transmitter 9 installed at the top center of the detection board emits a laser beam to the receiver 11 to obtain the first actual burial depth, and the detection board emits a detection beam to obtain the second actual burial depth. The average of the two is taken as the actual measurement value to improve data reliability.

[0056] 5. Determination of measurement results: Compare the actual measurement value with the theoretical burial depth. If the difference is within the threshold, it is considered qualified. Otherwise, it is unqualified and rework is determined based on the difference, so as to achieve efficient evaluation and quality control of the test results. 6. If the buried depth of the pipeline to be tested is determined to be qualified, the detection board 8 is powered off through the processing control module, and the detection board 8 is taken out through the electromagnetic clamp, and the level adjustment module is buried synchronously with the pipeline to be tested.

Claims

1. A pipeline buried depth detection method based on image recognition, characterized in that: The detection method comprises the following steps: S1, placing the detection plate on the top of the pipeline to be tested, respectively obtaining the contours of the pipeline to be tested and the detection plate, and adjusting the detection plate according to the distance between the contour edges of the pipeline to be tested and the detection plate, so that the center line of the detection plate and the center line of the pipeline to be tested are on the same straight line; S2, place the background plate parallel to the detection plate, and adjust the horizontality of the detection plate based on color recognition technology and deep learning model so that the detection plate is located on the horizontal section of the pipeline to be tested; S3, after the detection plate is adjusted to be horizontal, the spatial position coordinates of the detection plate are obtained by a first locator at the center of the bottom of the detection plate, and the spatial position coordinates of the background plate are obtained by a second locator at the center of the background plate; the background plate is adjusted according to the difference between the spatial position coordinates of the two, so that the center line of the background plate and the center line of the detection plate are located on the same straight line; S4, using the laser transmitter at the top center of the detection board to emit a laser beam to the receiver of the background board to obtain a first actual buried depth of the pipeline to be measured, and using the detection board to emit a detection beam to the background board to obtain a second actual buried depth of the pipeline to be measured, and taking the average of the first actual buried depth and the second actual buried depth as the actual measured value; S5, judging whether the burial depth is qualified according to the difference between the actual measured value and the theoretical burial depth, specifically: If the difference between the actual measured value and the theoretical burial depth is within the threshold range, it is judged as qualified, otherwise it is judged as unqualified, and whether rework is required is determined based on the difference between the two.

2. The pipeline buried depth detection method based on image recognition according to claim 1 is characterized in that: The process of adjusting the levelness of the test plate is as follows: Step 1: Build a deep learning model, collect the actual projection and actual color data of the emission light beam of the detection board on the background board under different conditions, pre-process the actual projection and actual color data to obtain the original data set, divide the original data set into a test set and a training set, use the training set to train the deep learning model, and evaluate the model performance through the test set; Step 2: Adjust the shape and color of the light source of the detection board to emit light beams of different colors and shapes to the background board; Step 3: The background plate correspondingly collects the actual projection and actual color data of the light beam emitted by the detection plate each time; Step 4: Input the actual projection and actual color data collected by the background plate into the trained deep learning model; Step 5: According to the theoretical buried depth of the pipeline, calculate the theoretical projection and theoretical color data of each emission on the background board; and calculate the deviation values ​​of the actual projection and actual color data from the theoretical projection and theoretical color data respectively, and judge whether the horizontality of the detection board is qualified according to the deviation value: If the deviation value is within the threshold range, the horizontality of the detection plate is determined to be qualified. If the deviation value is greater than the threshold, the detection plate is adjusted accordingly until the horizontality of the detection plate is qualified.

3. The pipeline buried depth detection method based on image recognition according to claim 2 is characterized in that: The actual projection and the theoretical projection respectively include the shape and area of ​​the projection; The actual color data and the theoretical color data include the saturation, chroma and brightness of the color respectively.

4. A pipeline buried depth detection system based on image recognition, characterized in that: The detection system adopts any one of the pipeline buried depth detection methods based on image recognition as described in claims 1-3, and the detection system comprises: A detection plate is installed on a horizontal section of the top of the pipeline to be tested; The background plate is used to assist the detection plate in completing the detection process and is located directly above the detection plate; A level adjustment module is installed at the bottom of the detection plate, the level adjustment module is buckled on the top of the pipeline to be tested, and a contact sensor is provided on the side of the level adjustment module facing the pipeline to be tested; The installation module is located on the ground of the pipeline to be tested and is used to adjust the positions of the background plate and the detection plate; The acquisition module includes: an image acquisition unit for acquiring the contours of the pipeline to be tested and the detection plate, a spatial position acquisition unit for acquiring the spatial position coordinates of the detection plate and the background plate, a distance detection unit for measuring the first actual buried depth of the pipeline to be tested, and an image recognition unit for realizing color recognition and a second actual buried depth; The image acquisition unit is located directly above the detection board; The spatial position acquisition unit includes a first locator installed at the center of the bottom of the detection plate, and a second locator installed at the center of the background plate; The distance detection unit and the image recognition unit are respectively installed on the detection plate and the background plate, and the image recognition unit is arranged on the peripheral side of the distance detection unit; The processing control module is respectively connected to the detection board, the contact sensor, the installation module, the level adjustment module, the image acquisition unit, the first locator, the second locator, the distance detection unit and the image recognition unit.

5. The pipeline buried depth detection system based on image recognition according to claim 4 is characterized in that: The detection board is an intelligent light-emitting board; The center line of the detection plate coincides with the center line of the level adjustment module; The distance detection unit includes a laser transmitter installed at the center of the top of the detection plate and a receiver installed at the center of the bottom of the background plate; The image recognition unit includes a light source of the detection plate and a collector arranged on the background plate; The laser emitter is located at the center of the light source; The collector is located on the periphery of the receiver, and the collector is arranged corresponding to the light source of the detection board; The laser transmitter, the receiver, the light source and the collector are respectively connected to the processing control module; The contact sensor is mounted at the center of the bottom of the level adjustment module; The second locator is installed at the center of the background plate.

6. The pipeline buried depth detection system based on image recognition according to claim 4 is characterized in that: The mounting module is a mechanical gripper.

7. The pipeline buried depth detection system based on image recognition according to claim 4 is characterized in that: The level adjustment module comprises a level adjustment part, a clamping part and a mounting plate connected in sequence from bottom to top; The top and bottom of the mounting plate are respectively provided with a first concave mounting opening and a second concave mounting opening; The first locator is located in the first concave mounting opening; The contact sensor is installed in the second concave installation opening; The detection plate is installed on the top of the mounting plate; The clamping parts are respectively located on the outer sides of the contact sensors, and at least two clamping parts are provided correspondingly, and the two clamping parts are clamped on both sides of the pipeline to be tested by moving closer to or farther away from each other; The level adjustment part is an automatic telescopic device, and one telescopic end faces underground.

8. The pipeline buried depth detection system based on image recognition according to claim 7 is characterized in that: Adjustable telescopic rods are respectively arranged between the clamping portion and the mounting plate; The adjustable telescopic rod is arranged along the radial direction of the pipeline to be tested, the telescopic ends of the adjustable telescopic rod are arranged facing each other and are respectively connected to the clamping parts, the opposite ends of the adjustable telescopic rod are respectively connected to the mounting plates, and the adjustable telescopic rod is an automatic telescopic rod.

9. The pipeline buried depth detection system based on image recognition according to claim 7 is characterized in that: The clamping portion includes a first clamping portion and a second clamping portion with openings facing each other; The first clamping portion and the second clamping portion respectively include a connecting plate and a clamping plate; The connecting plates are respectively vertically arranged on the mounting plates, one end of the connecting plates is respectively connected to the telescopic end of the adjusting telescopic rod, and are driven by the adjusting telescopic rod to move along the radial direction of the pipeline to be tested on the mounting plate; The clamping plates are respectively installed on opposite sides of the connecting plate, and the opposite sides of the clamping plates are respectively arc-shaped.

Images