Intelligent reinforced concrete vibrating method based on machine vision

Abstract

The invention discloses an intelligent reinforced concrete vibrating method based on machine vision. The method comprises the following steps: analyzing the change of the area of a concrete mixture region in adjacent acquired images, determining a quasi-static moment in a concrete paving process, analyzing the contour characteristics of the concrete mixture region at the quasi-static moment, and determining the pixel coordinates of an approximate center point of each region. Orderly screening is carried out according to the specification requirement of the distance between the adjacent vibration points and the energy efficiency optimization target, and the pixel coordinates of the actual vibration points are obtained. According to the mapping relation between the pixel space and the driver space, driving parameters of point-by-point vibration of the mechanical arm are obtained; analyzing the surface structure and distribution characteristics of the image of the outermost ring layer of the action area of the current vibration point location in the vibration process, and obtaining the moment when the vibration quality of the current action area is qualified; the automatic detection of the quasi-static vibrating moment in the concrete paving process, the generation of ordered vibrating point positions and the automatic quality control of weak areas in the vibrating action range related to each point in the vibrating process are realized.

Description

Technical Field

[0001] This invention relates to concrete vibration methods, and more particularly to a machine vision-based intelligent vibration method for reinforced concrete. Background Technology

[0002] Concrete vibration is a crucial step in construction, directly affecting the density and structural performance of concrete. However, traditional vibration methods rely primarily on manual operation, which is not only labor-intensive and involves harsh working environments but also suffers from inadequate quality control. Insufficient vibration time (under-vibration) leads to residual air bubbles within the concrete, resulting in insufficient density; excessive vibration time (over-vibration) can cause segregation, forming quality defects such as honeycomb and pitting. Especially in concrete areas under reinforcing mesh, standardized vibration is difficult to achieve due to the limitations of traditional equipment and the lack of experience among workers. Furthermore, the lag in vibration quality evaluation often leads to rework, wasting human and material resources. Currently, research on mechanized and automated concrete vibration methods focuses primarily on plain concrete or simple construction scenarios. However, in actual complex environments, especially for concrete vibration operations in areas with reinforcing mesh, effective intelligent vibration control methods are still lacking. Summary of the Invention

[0003] Purpose of the invention: The present invention aims to provide a machine vision-based intelligent vibration method for reinforced concrete to improve the vibration quality.

[0004] Technical solution: The intelligent vibration method for reinforced concrete based on machine vision described in this invention includes the following steps:

[0005] The machine vision camera acquires images of the work area before vibration, performs instance segmentation, and obtains a mask image of the concrete mixture; based on the change in the area of ​​the concrete mixture in adjacent frames of acquired images, the quasi-static moment in the concrete paving process is determined.

[0006] The image of the working area at the quasi-static moment is used as the original image at the vibratory moment. After instance segmentation, morphological processing is performed to extract the contour features of the concrete mixture area and calculate the geometric centroid of each contour to form a pixel coordinate group of the approximate center point.

[0007] Based on the specifications for the spacing between vibration points and the energy efficiency optimization target, a greedy algorithm is used to filter the pixel coordinate groups of the approximate center points to obtain the ordered pixel coordinate groups of the vibration points.

[0008] Based on the calibration parameters of the machine vision camera and the hand-eye transformation matrix, the pixel space coordinates are converted into the workspace coordinates of the robotic arm; the driving parameters are calculated based on the inverse kinematics model of the robotic arm, a point-by-point vibration path is generated, and vibration begins.

[0009] By acquiring real-time images of the area where the vibration point is located, the vibration quality is determined based on the surface texture characteristics of the area. If it is not qualified, vibration continues; otherwise, vibration stops and the machine moves to the next vibration point until all vibration points have been traversed, thus completing the vibration of the current work area.

[0010] Furthermore, the specific steps for determining the quasi-static moment during the concrete paving process are as follows:

[0011] Based on the concrete mixture mask image corresponding to the working area image of the adjacent frame, the area change value of the concrete mixture in the adjacent frame is obtained. If the area change value is less than or equal to the preset threshold, the current time is determined to be the quasi-static vibratory time.

[0012] Further morphological processing, including opening and erosion operations, is used to optimize the feature sharpness of the concrete mixture mask image.

[0013] Furthermore, the specifications for the spacing between vibration points and the energy efficiency optimization targets for vibration point selection are as follows:

[0014] The distance between adjacent vibration points is greater than or equal to the radius of action of the vibrator and less than or equal to 1.5 times the radius of action of the vibrator;

[0015] To prevent missed vibrations, ensure no area is missed by covering unvibrated areas;

[0016] To prevent over-vibration, minimize the overlap of the effective working area between vibration points;

[0017] Optimize the vibration path and reduce the total number of vibration points to achieve energy efficiency optimization.

[0018] Furthermore, a greedy algorithm is used to filter the pixel coordinate groups of the approximate center point, as follows:

[0019] Select the point closest to the initial position of the robotic arm from the pixel coordinate group as the starting point;

[0020] In each round of screening, the remaining point set is traversed, the new coverage area of ​​the vibration coverage area corresponding to each point to the currently uncovered area is calculated, and the overlapping area of ​​the coverage area formed with the selected points is calculated.

[0021] Prioritize selecting the point with the largest newly added coverage area as the actual vibration point. If there are multiple candidate points, select the one with the smallest overlap area with the selected point and add it to the set of selected points.

[0022] Repeat the above steps until all concrete areas are completely covered, or there are no points that meet the screening criteria.

[0023] Furthermore, the specific steps for generating the point-to-point vibration path for the robotic arm are as follows:

[0024] The pixel space coordinates of the vibration point are converted into workspace coordinates using machine vision camera calibration parameters and hand-eye transformation matrix.

[0025] Joint space parameters are calculated using the inverse kinematics model of the robotic arm;

[0026] The joint space parameters are mapped to the actual control commands of the actuator to generate a point-by-point vibration path.

[0027] Furthermore, the surface texture features of the affected area, including contrast, entropy, uniformity, and energy, are used to determine the compactness and quality of concrete after vibration.

[0028] Furthermore, the process for determining the quality of vibration compaction is as follows:

[0029] Cut off the outermost layer of the vibratory compaction point's effective area, designating it as a weak zone;

[0030] The weak areas are converted to grayscale and filtered for noise reduction;

[0031] Calculate the gray-level co-occurrence matrix to obtain the texture feature index of the vibration compaction quality;

[0032] Determine whether the vibration quality is up to standard.

[0033] Furthermore, the outermost layer of the vibration point's effective area is an outer ring area with the current vibration point as the center and the radial distance of the vibrator as the radius.

[0034] Furthermore, the weak areas within the radial range of the vibrator are designated as key monitoring areas.

[0035] Beneficial Effects: Compared with existing technologies, the significant advantages of this invention are: 1. This invention utilizes machine vision technology and intelligent algorithms, taking into account the impact of concrete flowability on the accuracy of visual concrete segmentation during the paving process. Based on the change in the area of ​​the concrete mixture in adjacent frames of acquired images, it determines the quasi-static moment in the concrete paving process, realizing automatic detection of the quasi-static vibratory moment in the concrete paving process under the reinforcing mesh; 2. This invention considers the vibration specification requirements and the goal of reducing energy and improving efficiency during robotic arm operation, realizing the orderly selection and path optimization of actual vibration points; 3. This invention considers the decreasing trend of the energy of the vibrator as it propagates radially, realizing automated control of vibration quality in weak areas; 4. This invention is applicable to complex reinforcing mesh areas, reducing manual intervention and optimizing the vibration process through mechanical automation, improving vibration efficiency and construction quality, and has broad application prospects and promotional value. Attached Figure Description

[0036] Figure 1 A flowchart for determining the quasi-static vibratory time of concrete;

[0037] Figure 2 A flowchart for extracting the center points of multiple blocks in a concrete mixture;

[0038] Figure 3 A flowchart showing the actual vibration points;

[0039] Figure 4 A flowchart for determining the driving parameters of the robotic arm for point-by-point vibration compaction;

[0040] Figure 5 A flowchart illustrating the method for determining the quality of vibration compaction. Detailed Implementation

[0041] The invention will now be further described with reference to the accompanying drawings.

[0042] The intelligent vibration method for reinforced concrete based on machine vision described in this invention includes the following steps:

[0043] (1) The machine vision camera acquires images of the working area before vibration, performs instance segmentation, and obtains a mask image of the concrete mixture; based on the change value of the area of ​​the concrete mixture in the acquired images of adjacent frames, the quasi-static moment in the concrete paving process is determined.

[0044] Acquire adjacent original images at intervals of 'a' frames; segment the adjacent original images to obtain their respective concrete mixture regions; obtain adjacent concrete segmentation masks for their respective concrete mixture regions; obtain adjacent areas of the adjacent concrete segmentation masks based on their contour features; subtract the adjacent areas to obtain the area change value of the concrete mixture regions in the adjacent original images; determine the state of the concrete mixture at the time of acquisition of the next image based on the area change value; retain the next original image whose area change value is less than or equal to a preset area change threshold, determine that the state of the concrete mixture in the next original image is quasi-static, and determine the vibratory compaction time of the quasi-static concrete mixture. Retain the next original image whose area change value is greater than the preset area change threshold, determine that the state of the concrete mixture in the next original image is dynamic, re-acquire new original images, and process them using the specific scheme described in this paragraph until the area change value is less than or equal to the preset area change threshold.

[0045] (2) Take the image of the working area at the quasi-static moment as the original image at the vibratory moment, perform morphological processing after instance segmentation, extract the contour features of the concrete mixture area, and determine the pixel coordinate group of the approximate center point of all concrete mixture mask areas.

[0046] The original image at the vibratory compaction time is segmented to obtain an instance segmentation mask image of the concrete mixture; morphological processing is performed on the instance segmentation mask image to obtain a concrete mixture mask image with clearer feature information; the morphological processing includes opening operation and erosion; contour features in the concrete mixture mask image are obtained; based on the contour features, the approximate center point pixel coordinate group of all concrete mixture mask regions is obtained.

[0047] (3) Based on the standard requirements for the spacing between vibration points and the energy efficiency optimization target, a greedy algorithm is used to filter the pixel coordinate group of the approximate center point to obtain the ordered pixel coordinate group of the vibration point.

[0048] Based on the specifications for the spacing between adjacent vibration points, the prerequisite of qualified vibration quality, and the target of the robotic arm operation, the approximate center point pixel coordinate group is filtered to obtain an ordered actual vibration point pixel coordinate group; the specifications for the spacing between adjacent vibration points include:

[0049] The distance between adjacent vibration points shall not be greater than 1.5 times the effective radius of the vibrator and not less than the effective radius of the vibrator.

[0050] The prerequisites for passing the vibration compaction quality include:

[0051] To prevent under-vibration and missed vibration, the effective working area should be as large as possible, with each vibration point as the center and the radial distance of the vibrator as the radius.

[0052] To prevent over-vibration, the overlapping area of ​​the effective working area should be minimized, with each vibration point as the center and the radial distance of the vibrator as the radius.

[0053] The goals of reducing energy and improving efficiency when operating a robotic arm include:

[0054] Energy reduction and efficiency improvement. To achieve the goal of energy reduction and efficiency improvement, the number of actual vibration points should be minimized.

[0055] The selection process is performed using a greedy algorithm, the basic idea of ​​which includes:

[0056] Starting from the current point, select the nearest unvisited point each time; under the condition of being closest, add the standard requirements for the spacing between adjacent vibration points, the premise of qualified vibration quality, and the target when the robotic arm is operating as filtering conditions, until all points have been traversed; the "all points" are the approximate center point pixel coordinate group of all concrete mixture mask areas.

[0057] The specific steps of the greedy algorithm to implement the filtering include:

[0058] (a) Initial point selection

[0059] The starting point is selected from the pixel coordinates of the approximate center point that is closest to the initial pose of the robotic arm.

[0060] (b) Sorting and Filtering

[0061] The coordinate points in the approximate center point pixel coordinate group are sorted and filtered; the sorting rule is the point closest to the previous selected point; the filtering requirement is the standard requirement of the spacing between adjacent vibration points, and candidate points are obtained starting from the point closest to the selected point.

[0062] (c) Prioritize coverage of uncovered areas

[0063] Prioritize selecting points that can cover more uncovered areas of the concrete mixture, ensuring coverage of as many areas not yet covered by other points as possible. This is achieved by calculating the coverage area of ​​the candidate points.

[0064] (d) Reduce overlapping areas

[0065] Each time a new point is selected, it is essential to ensure that the overlap area between the newly selected point and the previous point forming the effective working area is as small as possible. This is indirectly achieved by optimizing the relative positions between the candidate points.

[0066] (e) Repeat steps (c) and (d).

[0067] Each time, from the remaining points in the candidate pool, repeat steps (c) and (d) to select points that minimize overlap while covering more area. Continue until the concrete mixture area is covered.

[0068] (f) Termination conditions

[0069] The algorithm terminates when the concrete mixture area is covered or when there are no more candidate points that meet the criteria.

[0070] (4) Based on the calibration parameters of the machine vision camera and the hand-eye transformation matrix, the pixel space coordinates are converted into the workspace coordinates of the robotic arm; the driving parameters are calculated based on the inverse kinematics model of the robotic arm, the point-by-point vibration path is generated, and the vibration begins.

[0071] An ordered set of actual vibration point pixel coordinates resides in pixel space; the pixel space is the image captured by the camera; the pixel space is transformed into a workspace using camera calibration parameters and a hand-eye transformation matrix; the workspace is the specific pose based on the robot arm's base coordinates; the workspace is transformed into a joint space using an inverse kinematics model; the joint space is an abstract model of the vibration robot arm system, described by the rotation angle or translation distance of each joint; the joint space is transformed into a actuator space using calibration parameters obtained through linear interpolation mapping to obtain the driving parameters for the robot arm's point-to-point vibration; the transformation to the actuator space also requires consideration of the actuator's placement and joint structure. Based on the driving parameters for the robot arm's point-to-point vibration, combined with the ordered set of actual vibration point pixel coordinates, an ordered set of driving parameters is obtained.

[0072] The driver drives the robotic arm to move to the first actual vibration point to perform vibration work according to the first set of driving parameters in the ordered driving parameter group; until the ordered driving parameter group has been traversed, that is, each actual vibration point has been vibrated to a qualified state, it indicates that the vibration work in the working area is over.

[0073] (5) By collecting images of the area where the vibration point is located in real time, the vibration quality is determined based on the surface texture characteristics of the area. If it is not qualified, vibration continues; otherwise, vibration is stopped and the next vibration point is moved to vibrate until all vibration points are traversed and the vibration of the current work area is completed.

[0074] During vibration at each actual vibration point, an image of the current vibration point during the vibration process is acquired. The image is cropped, retaining the outermost ring within the effective area of ​​the current vibration point. The outermost ring is a circle centered on the current vibration point, with a radius greater than or equal to 0.6 times the radial distance of the vibrator and less than or equal to the radial distance of the vibrator. Because the energy of the vibrator decreases radially, although the ring area is within the vibration effective range, it requires a longer vibration time than the inner ring area and is considered a weak area that must be given special attention. The original image of the weak area is used to obtain the concrete mixture area within the weak area; the concrete mixture area is then converted to grayscale and filtered for noise reduction to obtain a target area image with more obvious features; the grayscale co-occurrence matrix of the target area image is calculated to obtain texture feature indicators; the texture feature indicators include: contrast, entropy value, energy, uniformity, and correlation; based on the texture feature indicators, the vibration quality at the current moment is determined; based on the vibration quality, the qualified vibration time is determined; based on the qualified vibration time, the vibration duration at the current vibration point is obtained.

[0075] By automatically detecting the quasi-static vibratory moment during the concrete paving process, automatically generating orderly vibration points, and automatically determining the vibration quality of weak areas within the current vibration range in real time, intelligent vibration operation is achieved, reducing energy consumption and improving efficiency.

[0076] In summary, this invention identifies and acquires concrete mix areas by classifying and predicting the image features of the work area before vibration. By analyzing the changes in the area of ​​the concrete mix region in adjacent acquired images, the quasi-static moment in the concrete paving process is determined, and the quasi-static image of the concrete mix and its regional information are extracted. Further analysis of the contour features of the concrete mix region is conducted, and the approximate center point pixel coordinates of each region are calculated and obtained. Based on the specifications for the spacing between adjacent vibration points and energy efficiency optimization targets, orderly screening is performed to obtain the pixel coordinates of the actual vibration points. Based on the mapping relationship between pixel space and driver space, the driving parameters for point-by-point vibration by the robotic arm are obtained. By analyzing the surface structure and distribution characteristics of the outermost layer image of the current vibration point's action area during vibration, the moment when the vibration quality of the current action area is qualified is obtained. This invention achieves automatic detection of the quasi-static vibrable moment in the concrete paving process, generation of orderly vibration points, and automatic quality control of weak areas within the vibration action range of each point during the vibration process.

Claims

1. A machine vision-based intelligent vibration method for reinforced concrete, characterized in that, Includes the following steps: The machine vision camera acquires images of the working area before vibration, performs instance segmentation, and obtains a mask image of the concrete mixture; based on the change in the area of ​​the concrete mixture in adjacent frames of acquired images, the quasi-static moment in the concrete paving process is determined. The image of the working area at the quasi-static moment is used as the original image at the vibratory moment. After instance segmentation, morphological processing is performed to extract the contour features of the concrete mixture area. The centroid of the contour of each area is calculated as the approximate center point to form a pixel coordinate group of the approximate center point. Based on the specifications for the spacing between vibration points and the energy efficiency optimization target, a greedy algorithm is used to filter the pixel coordinate groups of the approximate center points to obtain the ordered pixel coordinate groups of the vibration points. Based on the calibration parameters of the machine vision camera and the hand-eye transformation matrix, the pixel space coordinates are converted into the workspace coordinates of the robotic arm; the driving parameters are calculated based on the inverse kinematics model of the robotic arm, a point-by-point vibration path is generated, and vibration begins. By acquiring real-time images of the area where the vibration point is located, the vibration quality is determined based on the surface texture characteristics of the area being vibrated. If it is not qualified, vibration continues. Conversely, if the vibration fails, stop vibrating and move to the next vibration point, continuing until all vibration points have been visited to complete the vibration of the current work area.

2. The intelligent vibration method for reinforced concrete based on machine vision according to claim 1, characterized in that, The specific steps for determining the quasi-static moment during concrete paving are as follows: Based on the concrete mixture mask image corresponding to the working area image of the adjacent frame, the area change value of the concrete mixture in the adjacent frame is obtained. If the area change value is less than or equal to the preset threshold, the current time is determined to be the quasi-static vibratory time.

3. The intelligent vibration method for reinforced concrete based on machine vision according to claim 1, characterized in that, Morphological processing includes opening and erosion operations.

4. The intelligent vibration method for reinforced concrete based on machine vision according to claim 1, characterized in that, The specifications for the spacing between vibration points and the energy efficiency optimization targets for selecting vibration points are as follows: The distance between adjacent vibration points is greater than or equal to the radius of action of the vibrator and less than or equal to 1.5 times the radius of action of the vibrator; To prevent missed vibrations, ensure no area is missed by covering unvibrated areas; To prevent over-vibration, minimize the overlap of the effective working area between vibration points; Optimize the vibration path and reduce the total number of vibration points to achieve energy efficiency optimization.

5. The intelligent vibration method for reinforced concrete based on machine vision according to claim 4, characterized in that, A greedy algorithm is used to filter the approximate center point pixel coordinate group, specifically including: Select the point closest to the initial position of the robotic arm from the set of pixel coordinates of the approximate center point as the starting point; In each round of screening, the remaining point set is traversed, the new coverage area of ​​the vibration coverage area corresponding to each point to the currently uncovered area is calculated, and the overlapping area of ​​the coverage area formed with the selected points is calculated. Prioritize selecting the point with the largest newly added coverage area as the actual vibration point. If there are multiple candidate points, select the one with the smallest overlap area with the selected point and add it to the set of selected points. Repeat the above steps until all concrete areas are completely covered, or there are no points that meet the screening criteria.

6. The intelligent vibration method for reinforced concrete based on machine vision according to claim 1, characterized in that, The specific steps for generating the point-to-point vibration path for the robotic arm are as follows: The pixel space coordinates of the vibration point are converted into workspace coordinates using machine vision camera calibration parameters and hand-eye transformation matrix. Joint space parameters are calculated using the inverse kinematics model of the robotic arm; The joint space parameters are mapped to the actual control commands of the actuator to generate a point-by-point vibration path.

7. The intelligent vibration method for reinforced concrete based on machine vision according to claim 1, characterized in that, The surface texture features of the affected area include contrast, entropy, uniformity, and energy.

8. The intelligent vibration method for reinforced concrete based on machine vision according to claim 7, characterized in that, The process for determining the quality of vibration compaction is as follows: Cut off the outermost layer of the vibratory compaction point's effective area, designating it as a weak zone; The weak areas are converted to grayscale and filtered for noise reduction; Calculate the gray-level co-occurrence matrix to obtain the texture feature index of the vibration compaction quality; Determine whether the vibration quality is up to standard.

9. The intelligent vibration method for reinforced concrete based on machine vision according to claim 8, characterized in that, The outermost layer of the vibratory point's effective area is an outer ring area with the current vibratory point as the center and the radial distance of the vibrator as the radius.

10. The intelligent vibration method for reinforced concrete based on machine vision according to claim 9, characterized in that, Weak areas within the radial range of the vibrator should be the focus of monitoring.

Images