Method for detecting flatness of concrete surface based on machine vision

Abstract

The present application relates to the technical field of flatness detection, and specifically relates to a concrete surface flatness detection method based on machine vision, comprising fusion sensing collection: controlling the movement of a detection vehicle carrying an electromagnetic sensor and a three-dimensional camera, synchronously acquiring concrete surface point cloud data and steel mesh spatial coordinate data; steel deformation compensation: calculating the steel deflection deformation variable based on the acquired steel mesh coordinate data, reversely compensating the point cloud data, and generating steel mesh corrected surface point cloud; vibration feature analysis: identifying the vibration trace features on the steel mesh corrected surface point cloud, correlating the construction process parameters according to the trace geometric parameters, and outputting the real flatness evaluation and the vibration quality report. The present application realizes high-precision height correction by introducing a dynamic deformation field and an adjustable deformation transmission coefficient, improves the authenticity and comparability of the overall flatness evaluation, outputs the targeted parameter adjustment suggestions through the report module, and realizes the closed loop from surface error identification to construction process optimization.

Description

Technical Field

[0001] This invention relates to the field of flatness detection technology, and in particular to a method for detecting the flatness of concrete surfaces based on machine vision. Background Technology

[0002] In existing concrete construction quality inspection, surface flatness remains an important indicator for evaluating the quality of structural pouring and the compliance with construction standards. However, traditional flatness inspection mainly relies on manual stringing, mechanical level, or laser level, which has problems such as low operational efficiency, limitation by human experience, and difficulty in fully covering the construction area. In addition, these methods are difficult to obtain information about the underlying structure related to the reinforcement layout, and cannot effectively identify surface errors caused by reinforcement deformation. In particular, misjudgment and missed detection are more likely to occur in complex floor slabs and beam-column junction areas.

[0003] There is a lack of effective testing methods for vibration quality. Currently, on-site construction largely relies on manual experience to determine whether vibration is sufficient, lacking a data-driven quantitative indicator system. Some studies based on 3D point clouds have attempted to analyze surface defects, but they have failed to effectively integrate the physical relationship between rebar displacement and vibration marks, making it difficult to intelligently identify and locate insufficient or excessive vibration. In addition, the interpolation reconstruction process generally uses low-order models, ignoring the interference of local fluctuations on flatness statistics, which affects the accuracy and interpretability of subsequent evaluations. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides a method for detecting the flatness of concrete surfaces based on machine vision.

[0005] A machine vision-based method for detecting the flatness of concrete surfaces includes the following steps:

[0006] S1, Integrated Sensor Acquisition: Controls the movement of the inspection vehicle equipped with electromagnetic sensors and a 3D camera to simultaneously acquire point cloud data of the concrete surface and spatial coordinate data of the steel mesh;

[0007] S2, Rebar Deformation Compensation: Based on the acquired rebar mesh coordinate data, calculate the rebar deflection deformation, perform reverse compensation on the point cloud data, and generate a corrected surface point cloud of the rebar mesh;

[0008] S3, Vibration Feature Analysis: Identify vibration mark features on the point cloud of the corrected surface of the steel mesh, associate construction process parameters with the geometric parameters of the vibration marks, and output the true flatness evaluation and vibration quality report.

[0009] Furthermore, S1 includes:

[0010] S11, Path planning control: Based on the design spacing of the steel mesh, plan the movement path of the inspection vehicle and set the path spacing to the magnification factor of the design spacing;

[0011] S12, Pose Synchronous Acquisition: When the inspection vehicle moves at a set speed, the system synchronously acquires two types of data at each time point, including surface point cloud acquired by the 3D camera and magnetic field strength acquired by the electromagnetic sensor.

[0012] S13, Rebar Coordinate Calculation: Gaussian filtering and Laplace operator processing are applied to the magnetic field data to extract the center position of the rebar, and the spatial coordinates of the rebar are generated by combining the preset embedment depth;

[0013] S14, Spatiotemporal Alignment and Fusion: Unify the point cloud and rebar coordinates into the same coordinate system, construct a fused dataset, and distinguish surface points from rebar points by identifiers.

[0014] Furthermore, S12 includes:

[0015] S121, Surface point cloud acquisition: During the movement of the inspection vehicle along the preset path, the spatial point cloud data of the concrete surface is acquired in real time through a 3D camera, and the surface structure including height information is recorded.

[0016] S122, Magnetic field data acquisition: Obtain the magnetic field intensity distribution map of the detection area through an electromagnetic sensor array.

[0017] Furthermore, S13 includes:

[0018] S131, Gaussian smoothing: Gaussian kernel convolution is performed on the magnetic field intensity distribution collected by the electromagnetic sensor to smooth noise and enhance the response characteristics of the steel bar signal area;

[0019] S132, Center point identification and coordinate generation: The Laplacian operator is used to detect the minimum point in the convolution result, identify the center position of the rebar, and output the rebar coordinate point set in combination with the design burial depth information.

[0020] Furthermore, S2 includes:

[0021] S21, Establish an elastic foundation beam model: Based on the steel bar coordinate data, construct an elastic foundation beam model according to a single steel bar, and estimate its deflection change under concrete constraint and vibration pressure.

[0022] S22, Deformation field construction: Extend the steel bar deflection to the entire point cloud region, calculate the deformation of each surface point through interpolation, and generate a continuous deformation field;

[0023] S23, Reverse compensation calculation: The height of each point is compensated and corrected, and the surface depression caused by the settlement of the steel bars is repaired based on its deformation and transmission coefficient.

[0024] S24, Generate Corrected Point Cloud: Combine the corrected height values ​​with the original planar coordinates to output the compensated point cloud data.

[0025] Furthermore, S21 includes:

[0026] S211, Rebar Deformation Modeling: Based on the acquired rebar coordinate data, each rebar is treated as an independent calculation unit to establish an elastic foundation beam model;

[0027] S212, Deflection Calculation Expression: Using the elastic modulus of concrete, the stiffness of steel bars, and the construction vibration pressure parameters, calculate the deflection deformation of steel bars in the length direction.

[0028] Furthermore, S22 includes:

[0029] S221, Projection Matching and Neighborhood Search: For each point in the point cloud of the concrete surface, find its vertical projection position in the direction of the reinforcing bars and determine its four neighboring reinforcing bar nodes.

[0030] S222, value generation of continuous deformation field: based on the deflection value of the four neighboring steel reinforcement nodes, the deformation of each point cloud point is calculated by bilinear interpolation method to form a continuous deformation field covering the entire surface.

[0031] Furthermore, S23 includes:

[0032] S231, Compensation coefficient setting: Set the deformation transfer coefficient based on material properties and engineering experience;

[0033] S232, Surface Elevation Correction: Using the constructed deformation field, the height value of each point in the point cloud is compensated to repair the local surface depression area caused by the settlement of the reinforcing bars.

[0034] Furthermore, S3 includes:

[0035] S31, Vibration mark identification: In the corrected point cloud, suspected vibration mark areas are identified by analyzing height abrupt changes and surface tilt;

[0036] S32, Geometric parameter calculation: Calculate the average depth, equivalent width, and depth-to-width ratio for each identified vibration mark area;

[0037] S33, Process parameter correlation analysis: Based on the trace morphology and distribution density, a scoring model is constructed to determine whether there are problems with vibration.

[0038] S34, Flatness Correction: Remove areas with vibration marks and use interpolation algorithms to fill in missing data to generate a more realistic surface point cloud;

[0039] S35, Report Generation: Outputs vibration defect diagrams, flatness indicators, and construction parameter adjustment suggestions to form a complete quality assessment report.

[0040] Furthermore, S32 includes:

[0041] S321, Average Depth Calculation: Calculate the elevation difference between the internal points of each vibration mark area and the surrounding reference area, and calculate the overall average depression depth of the area.

[0042] S322, Equivalent width calculation: The projected area of ​​the vibration mark area is equivalent to a circle and converted into an equivalent diameter;

[0043] S323, Aspect Ratio Calculation: The aspect ratio is obtained by calculating the ratio of the average depth to the equivalent width.

[0044] The beneficial effects of this invention are:

[0045] This invention, based on machine vision fusion of a 3D camera and electromagnetic sensors, achieves synchronous acquisition of point cloud and steel mesh spatial information of concrete surface. It also constructs a refined compensation mechanism for construction error sources by physically modeling the deformation of steel bars using an elastic foundation beam model, effectively eliminating the interference of steel bar settlement on surface flatness. Compared with traditional methods based on global fitting or static calibration, this invention achieves high-precision elevation correction by introducing a dynamic deformation field and an adjustable deformation transfer coefficient, thereby improving the authenticity and comparability of the overall flatness assessment.

[0046] This invention proposes a vibration trace recognition algorithm that integrates height gradient and normal vector deviation. By combining multi-scale geometric features and spatial distribution density to construct a vibration quality evaluation model, it can intelligently identify process problems such as insufficient or excessive vibration. It also restores the true surface morphology through high-order spline interpolation. This method not only improves the detection sensitivity and diagnostic interpretability of vibration defects, but also outputs targeted parameter adjustment suggestions through the reporting module, realizing a complete closed loop from surface error identification to construction process optimization. This provides a new technical path for digital concrete construction quality control. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;

[0049] Figure 2 This is a feature analysis diagram of an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0051] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0052] like Figures 1-2 As shown, the machine vision-based method for detecting the flatness of concrete surfaces includes the following steps:

[0053] S1, Integrated Sensor Acquisition: Controls the movement of the inspection vehicle equipped with electromagnetic sensors and a 3D camera to simultaneously acquire point cloud data of the concrete surface and spatial coordinate data of the steel mesh;

[0054] S1 specifically includes:

[0055] S11, Path Planning Control: Based on the reinforcement mesh layout drawings, plan the movement path of the inspection vehicle to ensure comprehensive coverage of the entire concrete surface area. To avoid omissions or blind spots during scanning, set the path spacing H to match the maximum design spacing D of the reinforcement mesh. max A proportional relationship is expressed as:

[0056] H = η·D max ;

[0057] Among them, D max The maximum design spacing of the reinforcing mesh is 100-300 mm, determined according to the common reinforcement layout specifications in floor slabs or foundation components, ensuring coverage of all reinforcement directions. η is the path overlap coefficient, ranging from 1.2 to 1.5, set to overlap and cover to avoid detection blind spots. 1.0 indicates perfect coverage, greater than 1 indicates overlap, and 1.2-1.5 are empirically recommended values. H is the path spacing of the inspection vehicle, ranging from 120-450 mm.

[0058] S12, Pose Synchronous Acquisition: During the movement of the detection vehicle, it is necessary to acquire pose data at time point t.i Simultaneous acquisition of visual and electromagnetic signals enables the fusion of surface morphology and rebar distribution information. When the inspection vehicle moves at a speed v (v∈[0.2,0.5], balancing accuracy and efficiency, excessive speed can lead to image blurring or sensor sampling delay, while excessively slow speed can affect construction progress), at each moment t... i Two operations are performed simultaneously, specifically including:

[0059] (1) The point cloud data of the concrete surface at the current vehicle position is acquired by a 3D camera and represented as follows:

[0060] P c (t i )={(x j ,y j ,z j |j=1,2,...,N c};

[0061] Each point represents the spatial coordinates of the concrete surface, N. c This represents the number of point clouds, ranging from 1000 to 10000, and is related to the camera resolution and scanning range. It needs to cover the complete surface features of the scanned area in each frame. j ,y j ,z j These are the surface point cloud coordinates, acquired in real time by a 3D camera;

[0062] (2) The magnetic field intensity distribution map of the reinforcing steel area is obtained by an electromagnetic sensor array and is represented as follows:

[0063] M(t i )={I k (x k ,y k |k=1,2,...,N m};

[0064] Among them, I k (x k ,y k ) indicates at position (x k ,y k The magnetic field strength value at point () ranges from -100 to 100, which is the commonly used detection range for electromagnetic sensors. The magnetic induction intensity varies significantly near the reinforcing bars, requiring full-area induction coverage. (N) m This represents the number of sensor response points, ranging from 100 to 1000. It is related to the sensor array density and the distribution of reinforcing bars. Too low a value will result in missed detections, while too high a value will increase the computational burden.

[0065] S13, Reinforcement Coordinate Calculation: To achieve spatial positioning of each reinforcement bar in the reinforcement mesh, the magnetic field strength distribution M(t) is calculated. iFiltering and feature extraction are performed, and a two-dimensional Gaussian kernel function is used to perform smooth convolution on the original magnetic field data, as shown below:

[0066]

[0067] Where σ is the kernel width, which is proportional to the diameter of the rebar. σ = 0.7d is taken to balance smoothness and positioning sensitivity. The unit is mm. d is the diameter of the rebar, which can enhance the signal at the center of the rebar and suppress background noise. The value range is 6-32. The unit is mm. G(x,y) is the Gaussian convolution result, which represents the response value after smoothing the local magnetic field and is used for subsequent Laplace detection.

[0068] The second derivative of the Gaussian response image is calculated using the Laplacian operator and used to detect zero-crossing points to identify the center of the rebar. The judgment condition is expressed as follows:

[0069]

[0070] in, It is the Laplacian operator for Gaussian images, with a value range of [-1, 0], used to detect the location of curvature extrema and help identify magnetic field minimum points (center of the steel bar);

[0071] When the condition is met When the current position is determined to be the center point of the rebar, the set of rebar coordinates is output, represented as:

[0072] R(t i )={(x m ,y m ,z0)|m=1,2,...,N r};

[0073] Where z0 is the preset embedment depth of the reinforcing bar, with a value ranging from 10 to 80, determined by the construction design drawings. It is generally not measured directly by sensors but is set as a constant for coordinate splicing. T zero This is the Laplace zero-crossing threshold, with a value range of [-0.05, -0.02]. It controls the sensitivity for extracting extreme values ​​at the control center; a lower value helps suppress noise interference while still retaining a clear minimum value. x m ,y m It is the center coordinate (plane) of the reinforcing bar, which is determined by the result of the magnetic field extreme point extraction and depends on the accuracy of the electromagnetic sensor and the algorithm processing;

[0074] S14, Spatiotemporal Alignment: To unify the representation of the outputs of the 3D camera and electromagnetic sensors, a spatiotemporal alignment is constructed at the same time t. i The fused data matrix below includes information on both surface point clouds and rebar center points, distinguished by identifier bits, and is represented as follows:

[0075]

[0076] in, The point represents a point on the concrete surface. Used to distinguish the type of data point, ψ indicates that the point is a rebar coordinate point, ψ = 1, used to distinguish the type of data point.

[0077] S2, Rebar Deformation Compensation: Based on the acquired rebar mesh coordinate data, calculate the rebar deflection deformation, perform reverse compensation on the point cloud data, and generate a corrected surface point cloud of the rebar mesh;

[0078] S2 specifically includes:

[0079] S21, Establishing an elastic foundation beam model: To simulate the deformation behavior of steel reinforcement under the constraints of concrete and external vibration pressure, based on the obtained steel reinforcement coordinate data R(t)... i )={(x m ,y m Using a single steel bar as the calculation unit, an elastic foundation beam model is constructed, and its deflection deformation δ(x) is estimated, expressed as:

[0080]

[0081] Where δ(x) is the deflection deformation of the reinforcing bar at position x, ranging from 0 to 2, representing the vertical displacement of the reinforcing bar under the constraint of concrete and vibration pressure. It depends on the stiffness of the concrete and the vibration intensity, and is usually a millimeter-level offset. q is the concrete fluid pressure, ranging from 500 to 5000, in kPa, characterizing the fluid impact force on the reinforcing bar during concrete construction. It increases with the increase of slump SL, q = k1·SL + k2, where SL is the concrete slump, ranging from 100 to 230, characterizing the confinement capacity of the concrete on the reinforcing bar. Its value range is determined by the elastic modulus of the concrete; the larger the value, the "harder" the concrete. k1 is the empirical coefficient of the pressure-slump linear term, ranging from 10 to 20, used to fit the pressure changes under different construction flow conditions, and is generally calibrated based on historical construction data. k2 is the empirical coefficient of the pressure-slump constant term, ranging from 0 to 500, characterizing the initial pressure term when the slump is zero. c This is the subgrade coefficient, with units of kN / m. 3 The subgrade coefficient characterizes the degree of elastic constraint of concrete on the deformation of reinforcing steel, and has a power-law relationship with the elastic modulus of concrete. E c The elastic modulus of concrete is taken as [2.0 × 10⁻⁶]. 4 4.0×10 4 β is a model characteristic coefficient, representing the ratio of foundation stiffness to steel reinforcement stiffness, used to adjust the attenuation characteristics of deflection distribution. The value range is 0.01-1.0, and EI is the bending stiffness of the steel reinforcement, with a value range of

[10] . 6 10 9 The unit is N·mm 2 EI represents the bending resistance of reinforcing steel bars, which depends on the cross-section of the steel bars and the strength of the material. It is an important mechanical performance indicator. EI = E s ·I,E s This is the elastic modulus of the steel reinforcement, with a value of 2.0 × 10⁻⁶. 5 MPa is a standard material property of steel reinforcement in building structures, usually taken as a fixed value in engineering. I is the moment of inertia of the steel reinforcement section. d is the diameter of the reinforcing bar;

[0082] S22, Deformation Field Construction: After obtaining the deflection expression for each steel bar, the discrete steel bar deformation needs to be extended to the entire concrete surface point cloud region. This involves applying the obtained concrete surface point cloud P... c (t i )={(x j ,y j ,z j For each point in}, calculate the deflection value corresponding to its vertical projection position in the direction of the reinforcing bar. Generate a continuous deformation field through four-neighborhood bilinear interpolation, expressed as:

[0083]

[0084] Where, δ j It is a point (x) j ,y j The continuous deformation at point δ(x) ranges from 0 to 2, representing the vertical deformation of a point on the concrete surface affected by the deflection of the reinforcing steel. Common indentation deformation is on the order of millimeters. m ) is the rebar node x m Deflection at (x) j ,y j ,z j (x) represents the spatial coordinates of surface points in the original concrete point cloud, obtained from the concrete surface morphology captured by a 3D camera, serving as the base data before compensation. m ,y m ) is the distance (x) j ,y j The four nearest rebar nodes, selected through neighborhood search, are used as reference rebar nodes for interpolation calculation of the local deformation effect, Δx. jm =|x j -x m |,Δy jm =|y j -y m |,Δx jmΔy is the distance in the x-direction between a point in the point cloud and a rebar node, used to measure local weight; the closer the point, the greater the weight. Its value range does not exceed the rebar mesh spacing. jm This is the distance in the y-direction between the cloud point and the rebar node, used for bilinear interpolation calculations to ensure the locality of the interpolation region. The value range does not exceed the rebar mesh spacing. (D) x D y It refers to the spacing of the steel mesh in the x and y directions, with a value range of 100-300.

[0085] S23, Reverse Compensation Calculation: To correct the concrete surface error caused by steel reinforcement deformation, after obtaining the continuous deformation field δ j Then, a reverse correction is performed on each point cloud height value to compensate for the surface depression caused by the settlement of the reinforcing bars. The corrected height calculation is expressed as follows:

[0086] z j ′=z j +λ·δ j ;

[0087] Among them, z j These are the original height coordinates of the point cloud, representing the elevation value of the concrete surface before compensation, z. j ' is the corrected height after point cloud compensation, representing the true surface elevation value after reverse compensation for steel reinforcement deformation. λ is the deformation transfer coefficient, ranging from 0.6 to 0.8, used to quantify the intensity of the influence of steel reinforcement deformation transferred to the surface. A value less than 1 indicates partial transfer, and a larger value indicates more complete compensation. δ j >0 indicates a localized low-lying area caused by the settlement of reinforcing bars;

[0088] S24, Generate corrected point cloud: The corrected elevation value z... j ′ and the planar coordinates (x) of the original point cloud j ,y j The combination of these parameters outputs a point cloud dataset after deformation compensation, as follows:

[0089] P c ′(t i )={(x j ,y j ,z j ′)|j=1,2,...,N c};

[0090] Where, x j ,y j These are the original planar coordinates of the j-th point in the point cloud, acquired by a 3D camera, used to locate the two-dimensional spatial position of each surface point, and do not change with deformation compensation. P c ′(t i () is the corrected point cloud dataset of the concrete surface, representing time t.i The spatial morphology of the concrete surface, after deformation field compensation, serves as the basis for quality evaluation.

[0091] S3, Vibration Feature Analysis: Identify vibration mark features on the point cloud of the corrected surface of the steel mesh, associate construction process parameters with the geometric parameters of the vibration marks, and output the true flatness evaluation and vibration quality report.

[0092] S3 specifically includes:

[0093] S31, Vibration Mark Recognition: To identify surface defect areas caused by improper vibration, the obtained corrected surface point cloud data P of the reinforcing mesh is processed. c ′(t i )={(x j ,y j ,z j Vibration mark extraction was performed, and the vibration mark regions T that met the dual criteria were extracted using a region growing algorithm. k (k = 1, 2, ..., N) t The dual criterion condition is expressed as:

[0094] And n j -n0>θ th ;

[0095] Among them, T k It is the set of indented areas identified after the k-th identified vibration mark region meets the double threshold condition, used to analyze vibration anomalies. This is the point cloud height gradient magnitude, representing the rate of change of surface elevation. It is used to identify depression boundaries; a larger value indicates more pronounced local undulations. n j It is a point (x) j ,y j ,z j The normal vector at point (0, 0, 1) is given, and n0 is the normal vector of the reference plane, n0 = (0, 0, 1). th This is the gradient threshold, ranging from 0.25 to 0.35. It's an empirically set edge detection threshold; values ​​below this are considered gentle slopes, while values ​​above it indicate sharp edges. θ th This is the normal vector deviation angle threshold, ranging from 25° to 35°. An angle smaller than this is considered to be flush with the horizontal plane, while an angle larger than this indicates that the surface has a significant tilt. The normal vector deviation can be equivalently represented as:

[0096] ∥n j -n0∥=cos -1 (n jz );

[0097] Where, n jzThe z-component of the normal vector, ranging from 0.82 to 1, represents the degree to which a point is oriented vertically and is used to indirectly calculate the normal deviation angle. j is the unit normal vector at point j, a unit vector with a magnitude of 1, representing the orientation of the point cloud surface, used to determine whether a local area deviates from the horizontal plane, and n0 is the normal vector of the horizontal reference plane (0,0,1);

[0098] S32, Geometric parameter calculation: For each identified vibration mark area T k Calculate three geometric feature parameters, specifically including:

[0099] (1) Average depth h k : Measures the average degree of indentation between the point cloud inside the trace and its surrounding reference plane. The background reference height is taken from the annular region formed by extending 5 cm outward from the trace as the center, and is represented as:

[0100]

[0101] Among them, z p ′ is the height value of point p inside the region, used to calculate the depression depth of each trace region, h k The area marked with vibration marks is T. k The average depth, ranging from 0.2 to 1.5, characterizes the degree of indentation of the vibration mark. A larger value indicates concentrated vibration energy or excessively long vibration time. N k The area marked with tamping marks is T. k The number of point clouds within a given area is related to the point cloud resolution and the trace area, thus affecting the accuracy of average depth calculation. It is the average height of the point cloud within the ring area, which replaces the global plane as the background reference and can reduce interference from residual deformation of the reinforcing steel.

[0102] (2) Equivalent width w k : The area marked with vibration marks T k Projected area A k Equivalent to a circular region, converted to diameter, and used for normalization analysis, it is represented as:

[0103]

[0104] Among them, w k This is the equivalent width of the trace, ranging from 5 to 50. It converts irregular areas into equivalent circular diameters for easier normalization analysis. A k It is the projected area of ​​the tamping marks, in mm. 2 The value range is

[10] . 2 10 4 ];

[0105] (3) Aspect Ratio γ kDefined as the ratio of average depth to equivalent width, it is an important indicator for measuring the "sharpness" of a trace, expressed as:

[0106]

[0107] Where, γ k The aspect ratio, ranging from 0.1 to 1.0, reflects the "sharpness" of the mark and is used to judge the concentration of vibration. A larger value indicates a deeper and narrower mark. (N) k It is region T k The number of points in the cloud, ranging from 50 to 5000, is related to the point cloud resolution and trace area, and affects the accuracy of the average depth calculation. W is the average height of the point cloud in the outer ring of the trace. ring It is the width of the ring band, let W. ring =5·d, where d is the diameter of the steel bar;

[0108] S33, Process Parameter Correlation Analysis: Based on the extracted geometric features, construct the vibration quality evaluation function Q. k This is used to comprehensively reflect the severity of the traces and their distribution density per unit area, and is expressed as:

[0109]

[0110] Among them, Q k It is a comprehensive index of vibration quality, with a value range of 0-3. It comprehensively reflects the severity and distribution density of vibration marks and is a core indicator for evaluating vibration effect. k It is the aspect ratio of the tamping mark, N t This represents the total number of vibration marks, ranging from 0 to 50. S is the area of ​​the detection area (unit: m²). 2 ), α is the morphological feature weighting coefficient, α = 0.7, which is empirically set to emphasize the dominant influence of trace morphology on the vibration effect, β is the distribution density weighting coefficient, β = 0.3, which indicates the degree of influence of the number of traces per unit area, and is secondary to morphological features;

[0111] Based on Q k With vibration depth h k The results are used for construction diagnosis, and are expressed as follows:

[0112] (1) If Q k low The problem was determined to be insufficient vibration machine speed.

[0113] (2) If Q k Q high =0.42 and h k h max This indicates that the vibration time exceeded the limit.

[0114] Among them, Q​low The threshold for judging insufficient vibration, Q low =0.18, determined based on engineering experience; a value lower than this usually indicates insufficient energy or inadequate vibration frequency. Q high The threshold for judging excessive vibration, Q high =0.42. A value higher than this indicates that the vibration energy or time is too high, which may cause destructive marks. h max It is the maximum reasonable depth of vibration, h max =0.8, in mm, indicates that marks exceeding this depth indicate poor vibration effect, which can lead to hollow concrete or honeycomb surface.

[0115] S34, Smoothness Correction: After identifying the areas with vibration defects, T all areas with vibration marks... k After removing data from the compensated point cloud data, the missing regions are reconstructed using interpolation. A bicubic spline interpolation method is employed, with the interpolation function calculated using 16 neighboring points, expressed as:

[0116]

[0117] Where z′(u,v) is the height of the interpolation function result, in mm, representing the surface height value fitted by the bicubic spline algorithm, used to repair missing regions, a ij These are the interpolation coefficients. B is a 16×16 spline basis matrix, used to solve the linear equation coefficient matrix for interpolation coefficients, defined by standard splines. Z is a neighborhood height column vector used to fit the elevation distribution of missing regions, generating true flatness point cloud data P. true P true ={(x q ,y q ,z q ′)} represents the ground point cloud after removing abnormal vibration marks and restoring smoothness;

[0118] S35, Vibration Quality Report Generation: Outputs a comprehensive report including a vibration defect distribution map, flatness index, and process adjustment suggestion table, specifically including:

[0119] Vibration defect distribution map: Mark all vibration mark areas T k The location, and the corresponding Q. k value;

[0120] Smoothness index: The overall elevation fluctuation of the surface after correction is quantified in the form of variance, and is expressed as:

[0121]

[0122] Wherein, σ is the standard deviation index of surface smoothness, with a value ranging from 0 to 5. It is used to quantify the fluctuation of surface elevation; the smaller the value, the smoother the surface. It is the overall average elevation of the point cloud, the overall reference height of the surface, used for calculating the flatness variance;

[0123] Process Adjustment Recommendation Form: For areas with problems, based on the scoring and depth indicators, adjustment suggestions are output, such as recommended modification values ​​for vibration time, frequency, or rotation speed.

[0124] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in detail for the sake of brevity.

Claims

1. A method for detecting the flatness of a concrete surface based on machine vision, characterized by, The method comprises the following steps: S1, fusion sensing collection: controlling a detection vehicle carrying electromagnetic sensors and a three-dimensional camera to move and synchronously acquiring concrete surface point cloud data and steel mesh spatial coordinate data; S2, steel deformation compensation: calculating steel deflection deformation based on the acquired steel mesh coordinate data, compensating the point cloud data in reverse, and generating a steel mesh corrected surface point cloud; specifically comprising: S21, establishing an elastic foundation beam model: based on the steel coordinate data, constructing an elastic foundation beam model according to a single steel bar, and estimating the deflection change of the steel bar under the constraint of concrete and the vibration and compaction pressure; S22, deformation field construction: extending the steel deflection to the entire point cloud area, calculating the deformation of each surface point by an interpolation method, and generating a continuous deformation field; S23, reverse compensation calculation: compensating and correcting the height of each point, repairing the surface depression caused by the sinking of the steel bar according to the deformation and the transfer coefficient; S24, generating a corrected point cloud: combining the corrected height value with the original plane coordinates to output the compensated point cloud data; S3, vibration feature analysis: identifying the vibration trace features on the steel mesh corrected surface point cloud, correlating the construction process parameters according to the vibration trace geometric parameters, and outputting the real flatness evaluation and the vibration quality report.

2. The machine vision-based concrete surface evenness detection method of claim 1, wherein, The S1 comprises: S11, path planning control: planning the moving path of the detection vehicle according to the design spacing of the steel mesh, and setting the path spacing as a magnification multiple of the design spacing; S12, pose synchronous collection: when the detection vehicle moves at a set speed, the system synchronously collects two types of data at each time point, including the surface point cloud acquired by the three-dimensional camera and the magnetic field intensity collected by the electromagnetic sensor; S13, steel coordinate calculation: performing Gaussian filtering and Laplace operator processing on the magnetic field data, extracting the steel center position, and generating the spatial coordinates of the steel bar in combination with the preset embedding depth; S14, space-time alignment and fusion: unifying the point cloud and the steel coordinates to the same coordinate system, constructing a fusion data set, and distinguishing the surface points and the steel points through identifiers.

3. The machine vision-based concrete surface evenness detection method of claim 2, wherein, The S12 comprises: S121, surface point cloud acquisition: during the movement of the detection vehicle according to the preset path, acquiring the spatial point cloud data of the concrete surface in real time through the three-dimensional camera, and recording the surface structure including the height information; S122, magnetic field data collection: acquiring the magnetic field intensity distribution map of the detection area through the electromagnetic sensor array.

4. The machine vision-based concrete surface evenness detection method of claim 2, wherein, The S13 comprises: S131, Gaussian smoothing processing: performing Gaussian kernel convolution on the magnetic field intensity distribution collected by the electromagnetic sensor, smoothing the noise and enhancing the response characteristics of the steel signal area; S132, center point identification and coordinate generation: using the Laplace operator to detect the minimum value points in the convolution result, identifying the steel center position, and outputting the steel coordinate point set in combination with the design embedding depth information.

5. The machine vision-based concrete surface evenness detection method of claim 1, wherein, The S21 comprises: S211, steel deformation modeling: based on the acquired steel coordinate data, regarding each steel bar as an independent calculation unit, and establishing an elastic foundation beam model; S212, deflection calculation expression: calculating the deflection deformation of the steel bar in the length direction by using the concrete elastic modulus, the steel stiffness and the construction vibration and compaction pressure parameters.

6. The machine vision-based concrete surface evenness detection method of claim 1, wherein, The S22 comprises: S221, projection matching and neighborhood search: for each point in the concrete surface point cloud, find its vertical projection position in the direction of the steel bar, and determine its four adjacent steel bar nodes; S222, value generation continuous deformation field: based on the deflection value of the four-neighborhood steel bar node, the deformation variable of each point cloud point is calculated by the bilinear interpolation method, forming a continuous deformation field covering the entire surface.

7. The machine vision-based concrete surface evenness detection method of claim 1, wherein, The S23 comprises: S231, compensation coefficient setting: set the deformation transfer coefficient according to the material properties and engineering experience; S232, surface elevation correction: using the constructed deformation variable field, the height value of each point in the point cloud is compensated, and the local surface depression area caused by the sinking of the steel bar is repaired.

8. The machine vision-based concrete surface evenness detection method of claim 7, wherein, The S3 comprises: S31, vibration trace identification: in the corrected point cloud, the suspected vibration trace area is identified by analyzing the height mutation and surface inclination; S32, geometric parameter calculation: calculate the average depth, equivalent width and depth-width ratio of each vibration trace area identified; S33, process parameter correlation analysis: based on the trace morphology and distribution density, a scoring model is constructed to determine whether the vibration has a problem; S34, flatness correction: remove the vibration trace area and use interpolation algorithm to complete the missing data, generate a more true surface point cloud; S35, report generation: output the vibration defect map, flatness index and construction parameter adjustment suggestion, and form a complete quality evaluation report.

9. The machine vision-based concrete surface evenness detection method of claim 8, wherein, The S32 comprises: S321, average depth calculation: calculate the average depression depth of the whole area by calculating the elevation difference between each vibration trace area and its surrounding reference area; S322, equivalent width calculation: the projection area of the vibration trace area is equivalent to a circle, which is converted into an equivalent diameter; S323, depth-width ratio calculation: the average depth and equivalent width are calculated by ratio, and the depth-width ratio is obtained.

Images