Concrete vibration monitoring method and system based on image recognition and three-dimensional point cloud

Abstract

The invention relates to a concrete vibration monitoring method and system based on image recognition and three-dimensional point cloud, and the method comprises the following steps: obtaining an image of a concrete surface in a vibration process and three-dimensional point cloud data of the concrete surface, marking coarse aggregate and bubbles on the concrete surface, and obtaining an image sample data set; using the image sample data set to train a target detection algorithm, and identifying aggregate and bubbles in the image; counting the number of coarse aggregates and the number of bubbles in the image; extracting the z coordinate value of each point on the vibration surface from the point cloud data to obtain the elevation difference of the whole vibration surface; establishing respective correlation models of the aggregate quantity, the bubble quantity, the elevation difference and the strength, the porosity and the segregation index by utilizing a regression model; and evaluating whether the vibration effect is qualified or not according to a prediction result of the correlation model. Real-time monitoring can be achieved in the vibrating process, and comprehensive and accurate vibrating quality evaluation is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of visual monitoring of concrete vibration, and specifically relates to a method and system for monitoring concrete vibration based on image recognition and three-dimensional point cloud. Background Art

[0002] Vibration is a crucial step in the process of preparing concrete. Through the vibration of the vibrating rod, the internal particles of the concrete are rearranged, and the voids and bubbles existing inside the concrete are discharged, so as to achieve the purpose of improving the compactness and quality of the concrete. During the vibration construction process, when there is no obvious collapse on the concrete surface, no more bubbles rise, the surface floats, and the surface is slurry-covered and flat, it reaches compaction. At present, the quality control of vibration at the construction site mainly relies on the experience of workers, and the vibration quality is greatly affected by human factors.

[0003] The development process of the concrete vibration quality monitoring method has evolved from the initial intuitive judgment relying on the experience of workers to the precise monitoring of vibration depth, time, and trajectory using advanced GPS, UWB positioning technology, sensors and other devices. Although these technologies have improved the objectivity of monitoring, they have not visually represented the vibration quality of concrete. With the progress of computer technology, especially in the field of deep learning, image recognition technology has begun to be applied to concrete quality monitoring, but mainly focuses on the recognition of worker behavior and the positioning of vibrating rods and vibration areas. The existing vibration detection methods use parameters such as vibration frequency and amplitude as inputs and surface quality as outputs for monitoring, which has hysteresis, cannot perform dynamic monitoring during the vibration process, and has poor monitoring accuracy and poor adjustability with vibration parameters as variables. Most of the existing methods are limited to the two-dimensional level and cannot comprehensively reflect the three-dimensional dynamic changes of concrete quality during the vibration process. Therefore, a method combining two-dimensional images of concrete, three-dimensional surface undulations of concrete, and related concrete hardening properties is proposed to solve the above problems. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the technical problem to be solved by the present invention is: to provide a method and system for monitoring concrete vibration based on image recognition and three-dimensional point cloud.

[0005] The technical solution adopted by the present invention to solve the above technical problems is:

[0006] In the first aspect, the present invention provides a method for monitoring concrete vibration based on image recognition and three-dimensional point cloud, and the method includes the following contents:

[0007] Obtain images of the concrete surface and three-dimensional point cloud data of the concrete surface during the vibration process, accurately mark the collected image samples, and the marking objects are the coarse aggregates and bubbles on the concrete surface to obtain an image sample data set;

[0008] Train the object detection algorithm using the image sample dataset to identify the aggregates and air bubbles in the image;

[0009] Count the respective quantities of coarse aggregates and air bubbles in the image;

[0010] Extract the z - coordinate value of each point on the vibrating surface from the point cloud data to obtain the elevation difference of the entire vibrating surface;

[0011] Establish a correlation model between the visual monitoring results and the hardened concrete performance: Obtain the strength, porosity, segregation index of the concrete specimens prepared under different vibration times with different concrete formulations, as well as the aggregate quantity, air bubble quantity, and elevation difference at the corresponding vibration times;

[0012] Use the regression model to establish the respective correlation models between the aggregate quantity, air bubble quantity, elevation difference and strength, porosity, segregation index;

[0013] Evaluate whether the vibration effect is qualified based on the prediction results of this correlation model.

[0014] Furthermore, the process of obtaining the elevation difference is as follows:

[0015] Collect the point cloud data of the concrete surface during the concrete vibration process, and remove the noise and invalid data in the collected point cloud data;

[0016] After that, perform Poisson three - dimensional reconstruction on the point cloud data to restore the concrete surface, and calculate the protrusion height H of each aggregate from the reconstructed three - dimensional surface according to formula (1) i , and obtain the elevation difference ΔΗ of the entire vibrating surface according to formula (2) through the protrusion heights of all aggregates;

[0017] H i =Z(P i ) - Z avg (P i ) (1)

[0018] ΔΗ=|max(H i ) - min(H i )| (2)

[0019] Among them, Z(P i ) is the height of aggregate point i, and P i is the point cloud coordinate of the i - th aggregate; Z avg (P i ) is the average height of the surrounding neighborhood of aggregate point i.

[0020] In the second aspect, the present invention provides a concrete vibration monitoring system based on image recognition and three - dimensional point cloud, adopting the method described above. The system includes:

[0021] The vision module includes an industrial camera for collecting two-dimensional image data of the concrete surface and a binocular point cloud camera for collecting three-dimensional point cloud data of the concrete surface;

[0022] The power supply module is used to provide stable voltage and current for the system, ensure the normal operation of each module, has a power management function, can monitor the power consumption, and remind the user to charge or replace the battery when the power is insufficient;

[0023] The switch module includes a manual switch and an automatic start-stop mode;

[0024] The data storage module includes a local storage unit, the cloud, and a storage management unit;

[0025] The data processing module is used to identify and count the aggregates and air bubbles in the image data, perform three-dimensional reconstruction on the collected point cloud data and calculate the surface elevation difference, and predict the concrete hardening performance;

[0026] The display module includes an operation interface and an alarm light, and is used to interact with the operator, provide functions such as system status monitoring, data display, and error prompt, so that the operator can quickly understand the system working status and the quality of the vibration process; when uneven or incomplete vibration is detected, a red light is displayed, and when the vibration is completed, a green light is displayed;

[0027] The control module is used to coordinate the work of each module and communicate with the above-mentioned vision module, power supply module, switch module, data storage module, display module and data processing module.

[0028] Furthermore, the system further includes a user terminal, and the user terminal communicates with the control module.

[0029] Compared with the prior art, the beneficial effects of the present invention are:

[0030] The system of the present invention obtains the concrete surface image and three-dimensional surface data through an industrial camera and a depth camera, and the monitoring parameters include coarse aggregates, the number of air bubbles and the surface elevation difference. Combining the object detection algorithm and three-dimensional point cloud reconstruction, the surface quality information is obtained in real time, and the concrete hardening performance (such as strength, porosity, segregation index) is correlated through a regression model to achieve a comprehensive and accurate vibration quality assessment.

[0031] The method of the present invention can monitor in real time during the vibration process, avoid the lag of monitoring after a single vibration is completed in the prior art, directly monitor the state of the concrete itself, improve the accuracy of prediction, and can perform real-time dynamic assessment and quality prediction, having great application prospects. Description of the Drawings

[0032] Figure 1Schematic flowchart of a vision monitoring method for concrete vibration based on image recognition and 3D point cloud according to an embodiment of the present invention.

[0033] Figure 2(a) is a schematic diagram of the installation position of the vision module under manual vibration equipment.

[0034] Figure 2(b) is a schematic diagram of the installation position of the vision module under automatic vibration equipment.

[0035] Figure 3(a) is an image sample collected during the vibration experiment.

[0036] Figure 3(b) is the image data after manually marking the aggregates and bubbles. Detailed implementation manners

[0037] The present invention will be further explained below in conjunction with embodiments and the accompanying drawings, but this is not used as a limitation to the protection scope of the present application.

[0038] Embodiment 1

[0039] In this embodiment, a vision monitoring system for concrete vibration based on image recognition and 3D point cloud is provided, including:

[0040] A vision module, including an industrial camera for collecting two-dimensional image data of the concrete surface, and a binocular point cloud camera for collecting three-dimensional point cloud data of the concrete surface;

[0041] A power module for providing stable voltage and current for the system to ensure the normal operation of each module, having a power management function, capable of monitoring the power consumption situation, and reminding the user to charge or replace the battery when the power is insufficient;

[0042] A switch module, including a manual switch and an automatic start-stop mode. The switch function allows the user to manually control the start and stop of the system. During the operation, the user can start and stop the monitoring system at any time to adjust the operation progress or process the data; the automatic start-stop mode automatically enables the monitoring function according to the working state of the vibration equipment, starts data collection automatically when the vibrator starts, and stops data recording automatically after vibration is completed;

[0043] A data storage module, including a local storage unit, a cloud, and a storage management unit; the local storage unit is used to save the real-time collected images, point cloud data, and processed data; the cloud is used for backing up and archiving the data to ensure the long-term preservation and security of the data. Cloud storage enables multiple workstations and operators to share the data, facilitating remote monitoring and data analysis; the storage management unit is used for classifying and storing data, supporting classification storage of data by time, project, or equipment number, facilitating quick retrieval and use in the later stage;

[0044] A data processing module, which is used to identify and count aggregates and bubbles in image data, perform 3D reconstruction on the collected point cloud data, calculate the surface elevation difference, and judge the concrete hardening performance;

[0045] A control module, which is used to coordinate the work of each module, and realizes the control of hardware devices (such as cameras, storage, data processing modules, etc.) through an embedded controller or an industrial computer; according to external inputs (such as the working status of the vibrator, monitoring tasks, etc.), automatically adjust parameters such as the shooting angle, resolution, and exposure of the camera. At the same time, according to the amount of collected data, control the use of storage space to avoid overload or data loss;

[0046] A display module, which is used to interact with the operator, provide functions such as system status monitoring, data display, and error prompt, so that the operator can quickly understand the system working status and the quality of the vibration process; when uneven vibration or incomplete vibration is detected, display a red light, and when the vibration is completed, display a green light.

[0047] Embodiment 2

[0048] As Figure 1 shown, the visual monitoring method for concrete vibration quality based on image recognition and 3D point cloud in this embodiment includes the following steps:

[0049] 1) Visual module installation

[0050] Taking a building floor slab as an example, during the concrete pouring and vibration construction process, taking the operation area range of the vibration equipment as the monitoring target, install a visual module above it. This module integrates an industrial camera and a structured light depth point cloud camera, and install an illumination device independently in the operation site to ensure the normal data collection work of the two types of visual sensors.

[0051] The installation of the visual module is mainly divided into two situations:

[0052] (1) When the vibration equipment is of the manually held type, then this visual module is fixed on an independent movable bracket, as shown in Figure 2(a), where 1 refers to the visual module.

[0053] (2) When the vibration equipment is of the automated type, such as a crawler vibrator truck, then this visual module is fixed at the end of the vibrator holder, as shown in Figure 2(b), where 1 refers to the visual module.

[0054] 2) Collection of visual monitoring data for concrete vibration (i.e., surface images and 3D point cloud data)

[0055] During the concrete pouring construction process, the vibrating equipment is inserted into the concrete to perform vibration compaction work (Note: the purpose of vibration is to make the aggregate evenly distributed and expel bubbles, so as to achieve a more compact concrete effect). The visual module is turned on synchronously, where the industrial camera is used to capture images of the concrete surface, and the depth point cloud camera collects three-dimensional point cloud data of the concrete surface to ensure that the surface of the vibration area can be widely covered. In this embodiment, the two-dimensional image and three-dimensional point cloud data of the concrete surface are collected at a frame rate of 1 frame per 15 frames, or the image data and the three-dimensional point cloud data are subsequently aligned using timestamp synchronization.

[0056] 3) Real-time evaluation of concrete vibration quality

[0057] Based on the data collected by the two types of visual sensors, the concrete vibration quality evaluation results are generated in real time, specifically identifying the number of coarse aggregates and bubbles in the two-dimensional image of the concrete surface, and counting the elevation difference of the three-dimensional point cloud data of the concrete surface, so as to judge whether the performance of the concrete after hardening meets the requirements. The specific steps are as follows:

[0058] 3.1 Using YOLO target detection algorithm for identification and statistics of coarse aggregate and bubbles on concrete surface

[0059] (1) Construction of image sample dataset

[0060] The collection of concrete vibration image samples is mainly based on experiments, and vibration experiments are carried out on multiple types of concrete grade materials. During the experiment, an industrial camera is used to capture images. A total of 100 valid images are captured for each group of concrete samples. The collected images are shown in Figure 3(a). In this embodiment, the LabelImg software is used to accurately mark the collected image samples. The marked objects are the coarse aggregate and bubbles on the surface of the concrete (the marked bubbles are the bubbles that overflow from the inside during the vibration process, and the bubbles on the mold wall are not marked). A bmp image file containing a marking box and a txt text file describing the marking information are generated, as shown in Figure 3(b), forming the final image sample data set. The image sample data set is divided into a training set, a validation set, and a test set.

[0061] (2) Target recognition training

[0062] In this embodiment, the YOLO target detection algorithm environment is built on the Pycharm platform using the Pytorch framework to perform image recognition, and the image sample data set is used to train it. After the training is completed, the target recognition model performance is evaluated through the validation set to ensure the recognition accuracy and stability of the target recognition model.

[0063] (3) Writing a detection target counting script

[0064] In this embodiment, a statistical module is written in the Python language, and the statistical module is used to count and classify aggregates and bubbles through a counting script.

[0065] The above target recognition model is used for identifying coarse aggregates and bubbles on the surface of concrete during vibration. It is integrated with the statistical module in an industrial camera. The industrial camera captures images, real-time identifies the coarse aggregates and bubbles in the images, and counts their respective quantities.

[0066] 3.2 Calculation of elevation difference on the surface of concrete during vibration

[0067] During the concrete vibration process, a depth point cloud camera captures the point cloud data of the concrete surface, and removes the noise and invalid data in the captured point cloud data. Poisson three-dimensional reconstruction is performed on the point cloud data to restore the concrete surface. The protrusion height of each aggregate is calculated from the reconstructed three-dimensional surface, and the elevation difference of the entire surface is calculated through the protrusion heights of all aggregates. The specific calculation process is as follows:

[0068] (1) Calculate the protrusion height H of each aggregate i

[0069] H i =Z(P i )-Z avg (P i )

[0070] Z(P i ) is the height of aggregate point i;

[0071] P i is the point cloud coordinate of the i-th aggregate;

[0072] Z avg (P i ) is the average height of the surrounding neighborhood of aggregate point i;

[0073] (2) Calculate the elevation difference ΔΗ of the entire surface

[0074] ΔΗ=|max(H i )-min(H i )|.

[0075] The depth point cloud camera captures the point cloud and obtains the elevation difference of the concrete surface in real time.

[0076] 3.3 Establish a correlation model between visual monitoring results and the properties of hardened concrete

[0077] In order to effectively evaluate the concrete vibration effect, it aims to obtain the performance (compressive strength, porosity, segregation index) of the hardened concrete based on the number of coarse aggregates, the number of air bubbles, and the elevation difference on the concrete surface during the vibration process obtained above, as the judgment basis for vibration control. To achieve this goal, in this embodiment, a certain C30 concrete material is taken as an example to illustrate the establishment process of the correlation model based on experimental data:

[0078] (1) Specimen preparation

[0079] Mix ratio: Determine the mix ratio of C30 concrete, including the dosage of materials such as cement, water, sand, and stone. Among them, cement is 210 kg / m 3 , sand is 810 kg / m 3 , crushed stone is 1010 kg / m 3 , water is 165 kg / m 3 , fly ash is 70 kg / m 3 , mineral powder S95 is 110 kg / m 3 , and the size of the specimen is a cube with a standard side length of 150 mm.

[0080] Vibration frequency: 50 HZ.

[0081] Vibration time: During the experiment, the vibration time is set to 0 s, 5 s, 10 s, 15 s, 20 s, 25 s, and 30 s respectively. Record the vibration state of the concrete at each time point, and stop the vibration for subsequent processing.

[0082] Four specimens are prepared for each group of concrete, three for strength testing and one for porosity testing and segregation index testing, with a total of 28 specimens prepared.

[0083] (2) Monitoring data of two types of visual sensors, industrial cameras and depth point cloud cameras

[0084] Surface topography data acquisition: Use an industrial camera to capture the images of the concrete material surface during the entire vibration process in real time, and identify them using the target recognition model in step 3.1. Use a depth point cloud camera to capture the three-dimensional point cloud data of the concrete material surface during the entire vibration process in real time, and calculate the height difference within the monitoring range based on this data to reflect the flatness of the concrete surface.

[0085] (3) Performance measurement

[0086] Stop the vibration and cure at 0 s, 5 s, 10 s, 15 s, 20 s, 25 s, and 30 s of concrete vibration respectively. After 28 days of curing, test the strength, porosity, and segregation index of the concrete at the corresponding time.

[0087] (4) Model establishment

[0088] Draw the curve graphs of the number of aggregates and air bubbles on the concrete surface, the elevation difference curve graph of the concrete surface, and the strength change graph, porosity change graph, and segregation index change graph during the vibration of each group of C30 concrete.

[0089] Take the number of aggregates on the concrete surface as independent variable 1, the number of air bubbles on the concrete surface as independent variable 2, the elevation difference of the concrete surface as independent variable 3, and the concrete strength, porosity, and segregation index as target variables (dependent variables) respectively.

[0090] Preprocess the data, that is, remove the abnormal data in independent variable 1, independent variable 2, and independent variable 3, and divide the data set into a 70% training data set and a 30% test data set.

[0091] Select a regression model: In this embodiment, the regression model can be a non-linear regression model, a support vector machine regression model, or a neural network regression model. Because the number of aggregates and air bubbles is not linearly related to the strength, a support vector machine regression model or a neural network regression model can be selected for prediction in the experiment, and the model with better regression fitting between the two is selected.

[0092] Model optimization: Use methods such as cross-validation to optimize the hyperparameters of the model.

[0093] Respectively obtain the respective correlation models of concrete strength, porosity, and segregation index.

[0094] 4) Evaluate the data collected on site

[0095] Embed the above correlation model, and based on the image recognition data and three-dimensional point cloud data, judge whether the vibration effect is qualified. When the predicted data of the three performance parameters of strength, porosity, and segregation index hardly change before and after, and these performance parameters tend to be stable, it indicates that the concrete has reached the expected density and quality, and the vibration process is completed. And there are red and green indicator lights. When the vibration quality monitoring result does not meet the requirements, the red light is displayed, otherwise the green light is displayed. At this time, it is prompted that the vibration equipment should stop working.

[0096] Embodiment 3

[0097] The concrete vibration visual monitoring system based on image recognition and three-dimensional point cloud in this embodiment includes the following functional modules:

[0098] (1) Visual module

[0099] Industrial Camera: An industrial camera with high resolution is mainly used for real-time acquisition of two-dimensional image data on the concrete surface. The camera is equipped with a high-speed shutter and a high-sensitivity sensor, capable of obtaining clear images in complex industrial environments. Especially during the vibration process, the fast-moving surface can still maintain high image quality. These images provide basic data for subsequent image processing and vibration quality assessment.

[0100] Binocular Point Cloud Camera: Using binocular stereo vision technology, two cameras are used to capture the same scene from different angles. Combining the parallax principle, the three-dimensional point cloud data of the concrete surface is calculated. This point cloud data can be used to generate a three-dimensional model of the concrete surface, accurately reflecting surface undulations, particle distributions, etc. during the vibration process, evaluating the vibration quality, and analyzing whether there are phenomena of missed vibration or over-vibration.

[0101] (2) Power Supply Module

[0102] The power supply module provides stable voltage and current for the system to ensure the normal operation of each module. According to the system power consumption, the power supply module may include components such as batteries, transformers, AC / DC converters, etc. to adapt to different voltage requirements. Batteries and UPS (uninterruptible power supply) systems can ensure that the monitoring system can still continue to work during power failures, avoiding data loss. It has a power management function, capable of monitoring the power consumption situation and reminding users to charge or replace the battery when the power is insufficient.

[0103] (3) Switch Module

[0104] It includes a manual switch and an automatic start-stop mode. The switch function allows users to manually control the start and stop of the system. During the operation, users can start and stop the monitoring system at any time to adjust the operation progress or process the data; the automatic start-stop mode automatically enables the monitoring function according to the working state of the vibration equipment. When the vibrator starts, data acquisition automatically begins, and when the vibration is completed, data recording automatically stops.

[0105] (4) Data Storage Module

[0106] Local Storage Unit: It includes local storage media such as hard disks and SSDs, used to save real-time acquired images, point cloud data, and processed data. For large-scale projects, a high-capacity storage unit can be provided to ensure that the system runs for a long time without losing data.

[0107] Cloud: Used for backing up and archiving data to ensure the long-term preservation and security of data. Cloud storage enables multiple workstations and operators to share data, facilitating remote monitoring and data analysis;

[0108] Storage Management Unit: Used for classified storage of data, supporting classified storage of data by time, project, or equipment number, facilitating fast retrieval and use in the later stage.

[0109] (5) Data processing module

[0110] The data processing module mainly integrates a target recognition model, a statistics module, an elevation difference calculation module, and a correlation model between visual monitoring results and hardened concrete performance to identify and count aggregates and bubbles in image data, perform three-dimensional reconstruction on the collected point cloud data and calculate the surface elevation difference, and predict the hardened performance of concrete.

[0111] Among them, the target recognition model is used to identify aggregates and bubbles;

[0112] The statistics module is used to count the number of aggregates and the number of bubbles in the image;

[0113] The elevation difference calculation module is used to calculate the elevation difference of the concrete vibration surface based on the point cloud data of the concrete vibration surface;

[0114] The correlation model is used to obtain the corresponding hardened concrete performance based on the number of aggregates, the number of bubbles, and the elevation difference;

[0115] Real-time processing and batch processing: For real-time monitoring during the vibration process, the data processing module can perform real-time data processing and provide instant feedback; while for the analysis of historical data, a batch processing mode is adopted to generate reports, statistical data, etc.

[0116] (6) Control module

[0117] Central control unit: The central control unit serves as the brain of the system and coordinates the work of each module. It controls hardware devices (such as cameras, storage, processing units, etc.) through an embedded controller or an industrial computer. The central control unit can communicate with modules such as sensors, cameras, and actuators in the system in real time to ensure the synchronization of data acquisition and processing.

[0118] Automation control and adjustment: The central control unit automatically adjusts parameters such as the shooting angle, resolution, and exposure of the camera according to external inputs (such as the working state of the vibrator, monitoring tasks, etc.). At the same time, it controls the use of storage space according to the amount of collected data to avoid overload or data loss.

[0119] (7) Display module

[0120] It includes an operation interface and an alarm light,

[0121] Operation interface: The central control unit interacts with the operator through the user interface, providing functions such as system status monitoring, data display, and error prompt, facilitating the operator to quickly understand the system working state and the quality of the vibration process.

[0122] Result feedback module: It is implemented by an alarm light. When the system detects uneven vibration or incomplete vibration, etc., the central control unit will control the result feedback module to display a red light. When the vibration is completed, a green light is displayed.

[0123] (8) User terminal

[0124] The user terminal communicates with the control module and can remotely obtain the vibration monitoring results.

[0125] Embodiment 4

[0126] This embodiment is a visual monitoring method for concrete vibration based on image recognition and three-dimensional point cloud. The method includes the following:

[0127] Obtain the image of the concrete surface and the three-dimensional point cloud data of the concrete surface during the vibration process, accurately mark the collected image samples. The marking objects are the coarse aggregates and air bubbles on the concrete surface, and obtain the image sample data set.

[0128] Use the image sample data set to train the target detection algorithm to identify the aggregates and air bubbles in the image.

[0129] Count the respective quantities of the coarse aggregates and air bubbles in the image.

[0130] Extract the z coordinate value of each point on the vibration surface from the point cloud data to obtain the elevation difference of the entire vibration surface.

[0131] Establish a correlation model between the visual monitoring results and the properties of hardened concrete: Obtain the strength, porosity, segregation index of the concrete specimens prepared under different vibration times with different concrete formulations, as well as the aggregate quantity, air bubble quantity, and elevation difference under the corresponding vibration time.

[0132] Use the regression model to establish the respective correlation models between the aggregate quantity, air bubble quantity, elevation difference and strength, porosity, segregation index.

[0133] Evaluate whether the vibration effect is qualified based on the prediction results of this correlation model. During the vibration process, when these performance parameters (strength, porosity, and segregation index) all tend to be stable, it indicates that the concrete has reached the expected density and quality, and the vibration process is completed accordingly. This is used as the basis for evaluating whether the vibration effect is qualified.

[0134] In this embodiment, the target detection algorithm can be implemented by intelligent algorithms such as the YOLO series or the R-CNN series.

[0135] The parts not described in this invention are applicable to the prior art.

Claims

1. A concrete vibration monitoring method based on image recognition and three-dimensional point cloud, characterized in that: The method includes the following: Acquire images of the concrete surface during the vibration process and the three-dimensional point cloud data of the concrete surface, accurately mark the collected image samples, and mark the coarse aggregate and bubbles on the concrete surface to obtain an image sample data set; Use the image sample dataset to train the object detection algorithm to identify aggregates and bubbles in the image; Count the respective numbers of coarse aggregate and air bubbles in the image; Extract the z coordinate value of each point on the vibrated surface from the point cloud data to obtain the elevation difference of the entire vibrated surface; Establish a correlation model between visual monitoring results and hardened concrete performance: obtain the strength, porosity, segregation index of concrete specimens prepared at different vibration times under different concrete formulas, as well as the number of aggregates, number of bubbles, and elevation difference at the corresponding vibration time; The regression model is used to establish the respective correlation models between the number of aggregates, the number of bubbles, the elevation difference and the strength, porosity and segregation index; The prediction results of this correlation model are used to evaluate whether the vibration effect is qualified.

2. The method according to claim 1, characterized in that The process of obtaining the elevation difference is: During the concrete vibration process, point cloud data of the concrete surface is collected, and noise and invalid data in the collected point cloud data are removed; Then, the point cloud data is reconstructed using Poisson 3D reconstruction to restore the concrete surface. The protrusion height H of each aggregate is calculated from the reconstructed 3D surface according to formula (1): i , through the raised height of all aggregates, the elevation difference ΔH of the entire vibrated surface is obtained according to formula (2); H i =Z(P i )-Z avg (P i ) (1) ΔΗ=|max(H i )-min(H i )| (2) Among them, Z(P i ) is the height of aggregate point i, P i is the point cloud coordinate of the i-th aggregate; Z avg (P i ) is the average height of the neighborhood around aggregate point i.

3. A concrete vibration monitoring system based on image recognition and three-dimensional point cloud, characterized in that: The method according to claim 1 or 2, wherein the system comprises: A vision module, including an industrial camera for collecting two-dimensional image data of the concrete surface, and a binocular point cloud camera for collecting three-dimensional point cloud data of the concrete surface; The power module is used to provide the system with stable voltage and current to ensure the normal operation of each module. It has power management function, can monitor power consumption, and remind users to charge or replace batteries when the power is low; Switch module, including manual switch and automatic start-stop mode; Data storage module, including local storage unit, cloud, and storage management unit; Data processing module, used to identify and count aggregates and bubbles in image data, perform three-dimensional reconstruction of collected point cloud data, calculate surface elevation difference, and predict concrete hardening performance; The display module includes an operation interface and an alarm light, which is used to interact with the operator and provide system status monitoring, data display, and error prompt functions, so that the operator can quickly understand the system working status and the quality of the vibration process; when uneven or incomplete vibration is detected, a red light is displayed, and when the vibration is completed, a green light is displayed; The control module is used to coordinate the work of each module and communicate with the above-mentioned visual module, power module, switch module, data storage module, display module and data processing module.

4. The system according to claim 3, characterized in that The system also includes a user terminal in communication with the control module.

Images