Steel bar component quality inspection method based on UAV image point cloud data

Abstract

The present invention relates to the technical field of bridge engineering quality inspection, and specifically to a steel bar component quality inspection method based on drone image point cloud data, comprising obtaining image data of a steel bar component at a certain height from the steel bar component by a camera carried by a drone, performing aerial triangulation calculation and point cloud data reconstruction on the image data, and constructing a three-dimensional point cloud model of the steel bar component; locating and extracting the top surface steel bars of the steel bar component; filtering and denoising the extracted steel bars e; clustering the steel bar components after denoising into single steel bars; performing parameter fitting on a single steel bar to locate the top surface center of the steel bar and calculate the diameter; and performing steel bar component quality inspection based on the top surface center as basic data. It can realize high-precision, automated, and rapid inspection of steel bar components, and effectively reduce the labor intensity of construction workers.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge engineering quality detection, and in particular to a steel bar component quality detection method based on unmanned aerial vehicle image point cloud data. Background Art

[0002] Steel bar components are a new construction technology proposed for traditional in-situ steel bar binding. This technology transfers the steel bar binding construction of bridge towers to the factory, conducts industrial production of steel mesh on site, and uses assembly to assemble tower columns, which improves the quality of tower column construction, reduces construction safety risks, and increases construction speed. The assembled steel bar components are large in size and have a large number of single steel bars.

[0003] The traditional method of inspecting steel bar quality is to measure it manually with a steel ruler. This method is inefficient and labor-intensive, increases the construction period to a certain extent, and is somewhat dangerous. Summary of the invention

[0004] The purpose of the present invention is to address the defects of the prior art and provide a steel bar component quality inspection method based on drone image point cloud data, which can achieve high-precision, automated and rapid inspection of steel bar components and effectively reduce the labor intensity of construction workers.

[0005] The present invention provides a steel bar component quality detection method based on drone image point cloud data, comprising:

[0006] The camera carried by the drone is used to obtain image data of the steel bar component at a certain height from the steel bar component, and the image data is subjected to aerial triangulation calculation and point cloud data reconstruction to construct a three-dimensional point cloud model of the steel bar component.

[0007] Locate and extract the top surface steel bars of steel bar parts;

[0008] Filter and remove noise from the extracted steel bar e;

[0009] Clustering of single steel bars for the steel bar parts after noise removal;

[0010] Perform parameter fitting on a single steel bar to locate the center of the top surface of the steel bar and calculate its diameter;

[0011] The quality of steel bar components is inspected based on the center of the top surface.

[0012] Preferably, the altitude of the drone is determined by the following method:

[0013] The flying height of the UAV is determined based on the height of the steel bar component, the focal length of the lens, the pixel size and the ground resolution. The flying height of the UAV = the height of the steel bar component + the focal length of the lens × the steel bar resolution ÷ the pixel size.

[0014] Preferably, the method of locating and extracting the top surface steel bars of the steel bar component includes:

[0015] Slice the steel bars at a certain interval along the elevation direction of the steel bar component, and number the point cloud slice data from top to bottom along the elevation, and the number is recorded as γi, where the highest point corresponds to the number γ1;

[0016] After each point cloud slice is completed, the number of slice data is counted, and the number is recorded as σi, where the number of point clouds corresponding to the slice numbered γ1 is σ1;

[0017] Dynamically calculate the increment ratio of slice data When the number of slices is ≥ 2, the calculation process is as follows:

[0018]

[0019] Perform point cloud detection to locate the top surface of the steel bar:

[0020] when When γi+1 is located as the steel bar slice to be detected, the steel bar is extracted and the slicing is stopped. Otherwise, the steel bar part point cloud data is sliced ​​along the elevation direction to dynamically analyze the point cloud data increment.

[0021] When all point cloud slices do not satisfy When γ1 is located as the detected steel bar slice, the slice data is extracted and the extracted steel bar slice data is recorded as e.

[0022] Preferably, the filtering and noise removal of the extracted steel bar e includes:

[0023] Point cloud data projection, that is, projecting each point ei in the extracted steel bar e onto the XOY plane, and the projected point cloud data is pe;

[0024] Coincident point elimination, that is, performing valid digital extraction on the point cloud data, performing coincident point detection on the point cloud data after valid digital extraction, and retaining unique points for points with coincident coordinate positions in the projected point cloud pe;

[0025] Nearest neighbor search, that is, traverse all points in the projected point cloud pe, remove the coincident points for any point pei, and search the nearest neighbor for the removed point cloud data. The nearest neighbor of each point is Ni, and the distance Di between each point and its nearest neighbor is recorded;

[0026] Noise point judgment, that is, compare each point of Di with the steel bar resolution. When Di is greater than the steel bar resolution, pei is judged as a noise point, and the ei corresponding to pei is a noise point, and the pei and ei points are eliminated; when Di is less than or equal to the steel bar resolution, calculate pei and Ni corresponding to the ei and ei* points of the extracted steel bar e, and calculate the spacing Dei between ei and ei*; when Dei is greater than the steel bar resolution, pei is judged as a noise point, and the ei corresponding to pei is a noise point, and the pei and ei points are eliminated, and the extracted steel bar e point cloud data with the noise points eliminated is recorded as e_n.

[0027] Preferably, clustering the de-noised steel bar components into single steel bars comprises:

[0028] Point cloud data projection, projecting each point in the point cloud data e_n onto the XOY plane, forming the projected point cloud data e_pe;

[0029] For clustering of a single steel bar, select any point in the point cloud data e_pe, and execute the algorithm according to the Euclidean distance clustering method with the steel bar resolution as the threshold to form multiple point cloud clustering blocks. Extract the point cloud data in e_n corresponding to each point cloud clustering block to complete the clustering of a single steel bar.

[0030] Preferably, the parameter fitting of a single steel bar to achieve the center positioning and diameter calculation of the steel bar top surface includes:

[0031] Determine the fitting model, including fitting the point cloud data with a cylindrical model to obtain the diameter of the steel bar and the center of the top surface of the steel bar. The formula is as follows;

[0032]

[0033] In the formula, (xyz) is the surface point of the steel bar; r is the radius of the steel bar; (lmn) is the direction vector of the steel bar axis; (x0 y0 z0) is a point on the steel bar axis.

[0034] Preferably, the steel bar component quality inspection based on the top surface center as basic data includes:

[0035] Determine the top center of the steel bar, including obtaining the top center of all single steel bars in the point cloud data e_n, that is, obtaining the maximum point of a single steel bar on the Z axis, projecting the point to the steel bar axis direction vector as the top center of the single steel bar, and recording the data set of the top center of all steel bars as O_top;

[0036] Segment the center point of the top surface of the steel bar, including clustering the data set O_top of the center point of the top surface into three categories: O_top inside, O_top middle, and O_top outside;

[0037] Determine the positions of points in the point cloud data of O_top inside, O_top middle, and O_top outside, calculate the center points of the three point cloud classes of O_top inside, O_top middle, and O_top outside, record them as O_top inside-o, O_top middle-o, and O_top outside-o, sort the position of the point cloud data of O_top inside, O_top middle, and O_top outside, and obtain the re-sorted data sets O_top′inside-XY, O_top′inside, O_top′middle-XY, O_top′middle, O_top′outside-XY, and O_top′outside;

[0038] Calculate the spacing between steel bars, including calculating the adjacent sequential points in O_top′ inner-XY, O_top′ middle-XY, and O_top′ outer-XY, and obtain the spacing between adjacent points as the direct spacing between steel bars;

[0039] The steel bar position deviation is calculated to obtain the offset of all steel bars in O_top′inner-XY, O_top′middle-XY, and O_top′outer-XY, and the steel bar quality inspection is completed.

[0040] Preferably, the calculation of the steel bar position deviation includes:

[0041] Perform steel bar straight line fitting, including arbitrarily selecting a point O_top′inside_i in O_top′inside-XY and calculating the direction of O_top′inside_i Vector and O_top′ inner_i points to Vector Calculating vectors and When the angle is acute, the point O_top′ is inner_i, Gather into one category, otherwise the points O_top′ inner_i, to be classified into one category;

[0042] Traverse the above process until all points in -XY in O_top′ are clustered;

[0043] Fit each category of point cloud data to obtain a set of straight line models of all categories, recorded as L′inner-XY, where L′inner-XY contains 4 straight lines, recorded as L′inner-XY_1, L′inner-XY_2, L′inner-XY_3, and L′inner-XY_4;

[0044] Calculate the angles and intersections between straight lines, including determining the intersection of all straight lines in L′inner-XY, and calculating the angles and intersections of intersecting straight lines. When L′inner-XY_1, L′inner-XY_2, L′inner-XY_3, and L′inner-XY_4 are arranged counterclockwise on the XOY plane, the angle between L′inner-XY_1 and L′inner-XY_2 is calculated as follows: The intersection is The angle between L′inner-XY_2 and L′inner-XY_3 is The intersection is The angle between L'inner-XY_3 and L'inner-XY_4 is The intersection is The angle between L′inner-XY_4 and L′inner-XY_1 is The intersection is

[0045] Make straight line adjustments, including converting the calculated Compared with the design angle of the steel bar component, the calculated angle difference is recorded as Take any straight line in the linear model L′in-XY as the reference line, assuming that the reference line is L′in-XY_1, and take the intersection points Adjust the straight lines L′inner-XY_2 and L′inner-XY_4 to the following angles: and The adjusted L′inner-XY_2 and L′inner-XY_4 are recorded as L″inner-XY_2 and L″inner-XY_4; L″inner-XY_4 is used as the reference line again, and the angles of L′inner-XY_3 and L″inner-XY_4 are calculated again. and intersection Computational design perspective and The difference by As the reference point, the angle Adjust the straight line L′nei-XY_3, and record the adjusted straight line as L″nei-XY_3; thus, the adjustment of the four straight lines is completed;

[0046] Calculate the offset of the steel bars, including taking points from the straight lines L′nei-XY_1, L″nei-XY_2, L″nei-XY_3, and L″nei-XY_4 according to the designed spacing of the steel bars, and arranging all the points in sequence, recording the arranged point data as O_L′nei-XY_1, and subtracting each point in O_top′nei-XY in sequence from the data in O_L′nei-XY_1 to obtain the offset of all the steel bars in O_top′nei-XY;

[0047] The same steps as those of O_top′ inner-XY are performed on O_top′ inner-XY and O_top′ outer-XY respectively to obtain the offset of all steel bars in O_top′ inner-XY and O_top′ outer-XY, and complete the steel bar quality inspection.

[0048] The beneficial effects of the present invention are:

[0049] 1. Use the camera carried by the drone to obtain the image data of the steel bar parts at a certain height from the steel bar parts, perform aerial triangulation calculation and point cloud data reconstruction on the image data, and build a three-dimensional point cloud model of the steel bar parts; locate and extract the top surface steel bars of the steel bar parts; filter and remove noise on the extracted steel bars; cluster the steel bar parts after noise removal into single steel bars; perform parameter fitting on the single steel bar to locate the top surface center of the steel bar and calculate the diameter; use the top surface center as the basic data to perform quality inspection of the steel bar parts. High-precision, automated, and rapid inspection of steel bar parts is achieved, effectively reducing the labor intensity of construction workers.

[0050] 2. This method proposes a steel bar component quality inspection method based on drone image point cloud data. The steel bar positioning and extraction technology based on drone image reconstruction point cloud data realizes the automatic collection of steel bar data and the automatic identification of steel bar parameters, which effectively solves the problems of long measurement time, low efficiency and high labor intensity of traditional manual steel ruler measurement, improves the construction efficiency and reduces the labor intensity of workers.

[0051] 3. This method proposes a method for constructing a benchmark reference for steel bar quality inspection. One of the major difficulties in the quality inspection of steel bar parts is how to determine the measurement reference for the quality inspection of steel bar parts. The traditional method is based on random point selection by the total station. This method lacks theoretical feasibility and can only qualitatively evaluate the construction quality of steel bar parts. This method uses original technology for steel bar quality inspection and proposes a method for constructing a benchmark reference for steel bar part inspection based on methods such as straight line fitting and straight line direction adjustment. It realizes the calculation of the offset of steel bars in steel bar parts and completes the quality inspection of steel bar parts. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A schematic diagram of a process of a steel bar component quality detection method based on drone image point cloud data of the present invention;

[0053] Figure 2 It is a schematic diagram of the top surface of the steel bar extracted by the present invention;

[0054] Figure 3 This is a schematic diagram of point cloud sorting-angle calculation of the present invention;

[0055] Figure 4 It is a schematic diagram of adjusting the straight line direction of the present invention. DETAILED DESCRIPTION

[0056] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0057] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0058] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0059] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0060] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0061] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways. "Multiple" means "two or more".

[0062] like Figure 1 The present invention provides a method for detecting the quality of steel bar components based on UAV image point cloud data, which includes:

[0063] The camera carried by the drone is used to obtain image data of the steel bar component at a certain height from the steel bar component, and the image data is subjected to aerial triangulation calculation and point cloud data reconstruction to construct a three-dimensional point cloud model of the steel bar component.

[0064] Locate and extract the top surface steel bars of steel bar parts;

[0065] Filter and remove noise from the extracted steel bar e;

[0066] Clustering of single steel bars for the steel bar parts after noise removal;

[0067] Perform parameter fitting on a single steel bar to locate the center of the top surface of the steel bar and calculate its diameter;

[0068] The quality of steel bar components is inspected based on the center of the top surface.

[0069] Embodiment 1

[0070] This embodiment provides a preferred embodiment of a method for detecting steel bar component quality based on drone image point cloud data, and the execution steps are as follows:

[0071] Step S1: Determine the flight altitude of the drone, which includes step S11.

[0072] Step S11: Determine the flight altitude of the UAV based on the height of the steel bar component, the focal length of the lens, the pixel size and the ground resolution. The flight altitude of the UAV = the height of the steel bar component + the focal length of the lens × the steel bar resolution ÷ the pixel size. Taking the Zenmuse P1 UAV onboard camera as an example, the pixel size of the camera is 4.4um, a 35mm camera lens is selected, and the steel bar resolution is set to 1mm. The flight altitude of the UAV is 7 + (35 × 1 ÷ 4.4) ≈ 14.95 meters, and the flight altitude can be set to 15m.

[0073] Step S2: Using the UAV 3D reconstruction software, perform aerial triangulation calculation on the image data and reconstruct the point cloud data to obtain a 3D point cloud model.

[0074] Step S3: Detecting the location and extraction of the steel bars, which includes steps S31 to S34.

[0075] Step S31: Slice the steel bar component at a certain interval along the elevation direction of the steel bar component, and number the point cloud slice data from top to bottom along the elevation, the number is recorded as γi, the highest point corresponds to the number γ1, and so on.

[0076] Step S32: After each point cloud slice is completed in step S31, the number of slice data is counted, and the number is recorded as σi. For example, the number of point clouds corresponding to the slice numbered γ1 is σ1.

[0077] Step S33: Dynamically calculate the increment ratio of slice data When the number of slices is ≥ 2, the calculation process is as follows:

[0078]

[0079] Step S34: Point cloud detection of the top surface of the steel bar, the judgment conditions are as follows:

[0080] when When , that is, the number of point clouds of two adjacent slices in elevation has an order of magnitude difference, then γi+1 is positioned as the detected steel bar slice, the steel bar is extracted, and the slicing in step S31 is stopped; otherwise, step S31 is executed to continue slicing the steel bar component point cloud data along the elevation direction and dynamically analyze the point cloud data increment. When all point cloud slices do not meet When γ1 is located as the steel bar slice, the slice data is extracted. The extracted steel bar slice data is recorded as e. Figure 2 Shown is a schematic diagram of the top surface of the extracted steel bars.

[0081] Step S4: filtering and denoising the extracted steel bar e, which includes steps S41 to S44.

[0082] Step S41: Projecting point cloud data. Project each point ei in e onto the XOY plane, and denote the projected point cloud data as pe;

[0083] Step S42: Overlapping points are eliminated. Valid digits are extracted from the point cloud data, and the accuracy of the valid data is maintained to the next order of magnitude of the steel bar resolution. For example, if the steel bar resolution is 10 -3 m, the number of significant digits needs to be kept to 10 -4 m, for example, if the coordinates of a point are (3.124568 3.124351 3.124248), after extracting valid digits, the coordinates of the point are (3.1245 3.1243 3.1242). Perform coincidence point detection on the point cloud data after valid digit extraction, and retain the unique points of the points whose coordinates overlap in the projected point cloud pe.

[0084] Step S43: nearest neighbor search. Traverse all points in the projected point cloud pe, remove overlapping points for any point pei, search for nearest neighbor points for the removed point cloud data, record the nearest neighbor point of each point as Ni, and record the distance Di between each point and its nearest neighbor point.

[0085] Step S44: Noise point judgment. Compare each point of Di in step S42 with the steel bar resolution. When Di is greater than the steel bar resolution, pei is judged to be a noise point, and ei corresponding to pei is a noise point, and pei and ei points are removed; when Di is less than or equal to the steel bar resolution, calculate pei and Ni corresponding to the ei and ei* points of the extracted steel bar e, calculate the spacing Dei between ei and ei*, and when Dei is greater than the steel bar resolution, pei is judged to be a noise point, and ei corresponding to pei is a noise point, and pei and ei points are removed. The extracted steel bar e point cloud data with noise points removed is recorded as e_n.

[0086] Step S5: clustering the denoised steel bar components into single steel bars, which includes steps S51 to S52.

[0087] Step S51: point cloud data projection. Step S41 is executed on the point cloud data e_n to project the points onto the XOY plane, forming the projected point cloud data e_pe.

[0088] Step S52: Single steel bar clustering, select any point in the point cloud data e_pe, and execute the algorithm according to the Euclidean distance clustering method with the steel bar resolution as the threshold to form multiple point cloud clustering blocks, that is, the distance between any point in each clustering block and its nearest neighbor is less than or equal to the steel bar resolution, and any two points between different clustering blocks are greater than the steel bar resolution. Extract the point cloud data in e_n corresponding to each point cloud block to complete the single steel bar clustering.

[0089] Step S6: Perform parameter fitting on a single steel bar, and denote any single steel bar data set as S. It includes steps S61 and S62.

[0090] Step S61: Determine the fitting model. Fit the point cloud data with a cylindrical model to obtain the diameter of the steel bar and the center of the top surface of the steel bar. The formula is as follows.

[0091]

[0092] In the formula, (xyz) is the surface point of the steel bar; r is the radius of the steel bar; (lmn) is the direction vector of the steel bar axis; (x0 y0 z0) is a point on the steel bar axis.

[0093] Step S62: Use a random consistency sampling algorithm to fit the model parameters involved in step S61 to obtain the radius r of the steel bar; the direction vector (lmn) of the steel bar axis, and a point on the steel bar axis (x0 y0 z0).

[0094] Step S7: steel bar component quality inspection, which includes steps S71 to S75.

[0095] Step S71: Determine the top center of the steel bar. Obtain the top centers of all individual steel bars in the point cloud data e_n of step S44. The maximum point of an individual steel bar on the Z axis is projected onto the steel bar axis direction vector as the top center of the individual steel bar. The data set of the top centers of all steel bars is recorded as O_top.

[0096] Step S72: Segmentation of the center point of the top surface of the steel bar. The steel bar parts are composed of three layers of steel bars: inner, middle and outer. Using the Spectral clustering algorithm, the clustering parameter is set to 3, and the top surface center data set O_top is clustered and segmented into three categories: O_top inner, O_top middle and O_top outer.

[0097] Step S73: Position the points in the O_top inner, O_top middle and O_top outer point cloud data (ordering). Calculate the center points of the three point cloud classes O_top inner, O_top middle and O_top outer, and record them as O_top inner-o, O_top middle-o and O_top outer-o.

[0098] Step S73-1: Sort the point cloud data in O_top. Project the point cloud data in O_top from O_top-o to the XOY plane, record them as O_topXY and O_topXY-O respectively, calculate the vector from O_topXY-O to each point in O_topXY, calculate the angle ω between each vector and the positive direction of the X-axis, reorder the points in O_topXY and O_top in descending order of angle, and record the sorted point cloud data as O_top′-XY and O_top′. Figure 3 The schematic diagram of point cloud sorting-angle calculation is shown.

[0099] Step S73 - 2 : Process the two data sets O_top center and O_top outside according to the method of step S73 - 1 to obtain the reordered data sets O_top′ center-XY, O_top′ center, O_top′ outside-XY, O_top′ outside.

[0100] Step S74: Calculate the spacing between steel bars. Calculate the adjacent sequential points in O_top'inner-XY, O_top'middle-XY, and O_top'outer-XY in steps S73-1 and S73-2, and obtain the spacing between adjacent points as the direct spacing between steel bars.

[0101] Step S75: Calculation of steel bar position deviation. Process the point cloud data of O_top' inner-XY, O_top' middle-XY, and O_top' outer-XY in steps S73-1 and S73-2. It includes steps S75-1 to S75-5.

[0102] Step S75-1: Rebar straight line fitting. Taking O_top′in-XY as an example, arbitrarily select a point O_top′in-i in O_top′in-XY and calculate the direction of O_top′in-i. Vector Calculate the value of O_top′ in_i pointing to Vector Calculating vectors and When the angle is acute, the point O_top′ is inner_i, Gather into one category, otherwise the points O_top′ inner_i, Classify them into one category. Repeat the above process until all points in O_top′in-XY are clustered. Perform the least squares fitting on each category of point cloud data to obtain the set of straight line models of all categories, recorded as L′in-XY, where L′in-XY contains 4 straight lines, recorded as L′in-XY_1, L′in-XY_2, L′in-XY_3, and L′in-XY_4.

[0103] Step S75-2: Calculate the angles and intersections between the straight lines. Determine the intersection of all the straight lines in L'inner-XY, and calculate the angles and intersections of the intersecting straight lines. Assume that L'inner-XY_1, L'inner-XY_2, L'inner-XY_3, and L'inner-XY_4 are arranged counterclockwise on the XOY plane, and calculate the angle between L'inner-XY_1 and L'inner-XY_2 as The intersection is The angle between L′inner-XY_2 and L′inner-XY_3 is The intersection is The angle between L'inner-XY_3 and L'inner-XY_4 is The intersection is The angle between L′inner-XY_4 and L′inner-XY_1 is The intersection is

[0104] Step S75-3: Adjust the straight line direction. The angle in the figure is compared with the design angle of the steel bar component, and the calculated angle difference is recorded as Take any straight line in the linear model L′in-XY obtained in step S75-1 as the reference line, assuming that the reference line is L′in-XY_1, and take the intersection points Adjust the straight lines L′inner-XY_2 and L′inner-XY_4 to the following angles: and The adjusted L′inner-XY_2 and L′inner-XY_4 are recorded as L″inner-XY_2 and L″inner-XY_4; L″inner-XY_4 is used as the reference line again to calculate the angle between L′inner-XY_3 and L″inner-XY_1 again. and intersection Computational design perspective and The difference by As the reference point, the angle Adjust the straight line L′nei-XY_3, and record the adjusted straight line as L″nei-XY_3; thus, the adjustment of the four straight lines is completed. Figure 4 The figure shows the schematic diagram of the straight line direction adjustment.

[0105] Step S75-4: Calculate the offset of the steel bars. Take points from the straight lines L′nei-XY_1, L″nei-XY_2, L″nei-XY_3, and L″nei-XY_4 according to the designed spacing of the steel bars, and arrange all the points in sequence according to step S73. The arranged point data is recorded as O_L′nei-XY_1. Subtract each point in O_top′nei-XY from the data in O_L′nei-XY_1 in sequence to obtain the offset of all the steel bars in O_top′nei-XY.

[0106] Step S75-5: Execute steps S75-1 to S75-4 for O_top′-XY and O_top′-XY respectively, obtain the offset of all steel bars in O_top′-XY and O_top′-XY, and complete the steel bar quality inspection.

[0107] It should be understood that the specific order or hierarchy of steps in the disclosed process is an example of an exemplary method. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the process can be rearranged without departing from the scope of protection of the present disclosure. The attached method claims present the elements of the various steps in an exemplary order and are not intended to be limited to the specific order or hierarchy described.

[0108] In the above detailed description, various features are grouped together in a single embodiment to simplify the disclosure. This method of disclosure should not be interpreted as reflecting an intention that the embodiments of the claimed subject matter require more features than are clearly stated in each claim. On the contrary, as reflected in the appended claims, the invention is in a state of having less than all the features of the disclosed individual embodiments. Therefore, the appended claims are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the invention.

[0109] The disclosed embodiments are described above to enable any person skilled in the art to implement or use the present invention. Various modifications of these embodiments are obvious to those skilled in the art, and the general principles defined herein may also be applied to other embodiments without departing from the spirit and scope of the present disclosure. Therefore, the present disclosure is not limited to the embodiments given herein, but is consistent with the broadest scope of the principles and novel features disclosed in this application.

[0110] The above description includes examples of one or more embodiments. Of course, it is impossible to describe all possible combinations of components or methods for the purpose of describing the above embodiments, but it should be recognized by those skilled in the art that the various embodiments may be further combined and arranged. Therefore, the embodiments described herein are intended to cover all such changes, modifications and variations that fall within the scope of protection of the appended claims. In addition, with respect to the term "comprising" used in the specification or claims, the word is covered in a manner similar to the term "including", just as "including" is explained as a transitional word in the claims. In addition, any term "or" used in the specification of the claims is intended to mean "non-exclusive or".

[0111] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A steel bar component quality detection method based on drone image point cloud data, characterized in that: include: The camera carried by the drone is used to obtain image data of the steel bar component at a certain height from the steel bar component, and the image data is subjected to aerial triangulation calculation and point cloud data reconstruction to construct a three-dimensional point cloud model of the steel bar component. Locate and extract the top surface steel bars of steel bar parts; Filter and remove noise from the extracted steel bar e; Clustering of single steel bars for the steel bar parts after noise removal; Perform parameter fitting on a single steel bar to locate the center of the top surface of the steel bar and calculate its diameter; Use the top surface center as the basic data to inspect the quality of steel bar components; The method of locating and extracting the top surface steel bars of the steel bar component comprises: Slice the steel bars at a certain distance along the elevation direction of the steel bar parts, and number the point cloud slice data from top to bottom along the elevation, and the number is recorded as γ i , where the highest point corresponds to number γ1; After each point cloud slice is completed, the number of slice data is counted, and the number is recorded as σ i , where the number of point clouds corresponding to the slice numbered γ1 is σ1; Dynamically calculate the increment ratio of slice data When the number of slices is ≥ 2, the calculation process is as follows: Perform point cloud detection to locate the top surface of the steel bar: when , that is, the number of point clouds of two adjacent slices in elevation differs by an order of magnitude, then γ i+1 Positioning is to detect steel bar slices, extract the steel bar, stop slicing, otherwise continue slicing the steel bar component point cloud data along the elevation direction, and dynamically analyze the point cloud data increment; When all point cloud slices do not satisfy When γ1 is located as the detected steel bar slice, the slice data is extracted and the extracted steel bar slice data is recorded as e.

2. The steel bar component quality detection method based on drone image point cloud data according to claim 1 is characterized in that: The altitude of the drone is determined by the following method: The flying height of the UAV is determined based on the height of the steel bar component, the focal length of the lens, the pixel size and the ground resolution. The flying height of the UAV = the height of the steel bar component + the focal length of the lens × the steel bar resolution ÷ the pixel size.

3. The steel bar component quality detection method based on drone image point cloud data according to claim 1 is characterized in that: The filtering and noise removal of the extracted steel bar e includes: Point cloud data projection, that is, to extract each point e in the steel bar e i Project to the XOY plane, and record the projected point cloud data as pe; Coincident point elimination, that is, performing valid digital extraction on the point cloud data, performing coincident point detection on the point cloud data after valid digital extraction, and retaining unique points for points with coincident coordinate positions in the projected point cloud pe; Nearest neighbor search, that is, traverse all points in the projected point cloud pe, for any point pe i Execute the elimination of overlapping points, search for the nearest neighbor points of the eliminated point cloud data, and record the nearest neighbor point of each point as N i , record the distance D between each point and its nearest neighbor i ; Noise point judgment, that is, D i Each point of D is compared with the steel bar resolution. i When it is greater than the steel bar resolution, it is judged that pe i is the noise point, pe i The corresponding e i is the noise point, for pe i 、e i Points are removed; when D i When it is less than or equal to the steel bar resolution, calculate pe i 、N i Corresponding to the extracted steel bar e i 、e i* point, calculate e i 、e i* The spacing between i ;Dang De i When it is greater than the steel bar resolution, it is judged that pe i is the noise point, pe i The corresponding e i is the noise point, for pe i 、e i The points are removed and the extracted steel bar e point cloud data after removing the noise points is recorded as e_n.

4. The steel bar component quality detection method based on drone image point cloud data according to claim 3 is characterized in that: The clustering of single steel bars of the steel bar components after noise removal includes: Point cloud data projection, projecting each point in the point cloud data e_n onto the XOY plane, forming the projected point cloud data e_pe; For clustering of a single steel bar, select any point in the point cloud data e_pe, and execute the algorithm according to the Euclidean distance clustering method with the steel bar resolution as the threshold to form multiple point cloud clustering blocks. Extract the point cloud data in e_n corresponding to each point cloud clustering block to complete the clustering of a single steel bar.

5. The steel bar component quality detection method based on drone image point cloud data according to claim 3 is characterized in that: The parameter fitting of a single steel bar to realize the center positioning and diameter calculation of the top surface of the steel bar includes: Determine the fitting model, including fitting the point cloud data with a cylindrical model to obtain the diameter of the steel bar and the center data of the top surface of the steel bar. The formula is as follows; In the formula, (xyz) is the surface point of the steel bar; r is the radius of the steel bar; (lmn) is the direction vector of the steel bar axis; (x0y0z0) is a point on the steel bar axis.

6. The steel bar component quality detection method based on drone image point cloud data according to claim 1 is characterized in that: The steel bar component quality inspection based on the top surface center as basic data includes: Determine the top center of the steel bar, including obtaining the top center of all single steel bars in the point cloud data e_n, that is, obtaining the maximum point of a single steel bar on the Z axis, projecting the point to the steel bar axis direction vector as the top center of the single steel bar, and recording the data set of the top center of all steel bars as O_top; Segment the center point of the top surface of the steel bar, including clustering the data set O_top of the center point of the top surface into three categories: O_top inside, O_top middle, and O_top outside; Determine the position of the points in the point cloud data of O_top inside, O_top middle, and O_top outside, and calculate the center points of the three point cloud classes of O_top inside, O_top middle, and O_top outside, which are recorded as O_top inside -o 、O_top -o 、O_top outside -o , sort the point cloud data inside, in, and outside O_top, and obtain the re-sorted data set O_top′ -XY , inside O_top′, in O_top′ -XY , O_top′inside, O_top′outside -XY , O_top′ outside; Among them, determining the position of points in the point cloud data in O_top includes: The point cloud data in O_top is converted into -o and projected into the XOY plane, respectively denoted as O_top XY 、O_top XY-O , calculate O_top XY-O Point to O_top XY The vector of each point in the X axis is calculated, and the angle ω between each vector and the positive direction of the X axis is calculated. The angles of O_top are calculated in descending order. XY 、O_top 内 The points are reordered, and the sorted point cloud data is recorded as O_top′ -XY , O_top′inside; The same position determination method as that in O_top is used to determine the positions of points in the point cloud data in O_top and outside O_top, respectively, and the positions of points in O_top′ are obtained. -XY , O_top′inside, O_top′outside -XY , O_top′ outside; Calculate the spacing of the reinforcement bars, including the spacing of the reinforcement bars within O_top′. -XY , O_top′ -XY 、O_top′ outside -XY Calculate the adjacent sequential points in the image, and obtain the spacing between adjacent points as the direct spacing of the steel bars; Calculate the steel bar position deviation and obtain the inner value of O_top′ -XY , O_top′ -XY 、O_top′ outside -XY The offset of all steel bars in the steel bar is measured to complete the steel bar quality inspection.

7. The steel bar component quality detection method based on drone image point cloud data according to claim 6 is characterized in that: The calculation of the steel bar position deviation comprises: Perform steel bar straight line fitting, including arbitrarily selecting O_top′ -XY A point in O_top′ _i , calculate O_top′ _i point to Vector and O_top′ _i point to Vector Calculating vectors and When the angle is acute, the point O_top′ is _i , Gather into one category, otherwise the point O_top′ _i , to be classified into one category; Traverse the above process until O_top′ is within -XY All points in the cluster are clustered; Fit each category of point cloud data to obtain the linear model set of all categories, recorded as L′ 内-XY , where L′ 内-XY It contains 4 straight lines, denoted as L′ 内-XY_1 , L′ 内-XY_2 , L′ 内-XY_3 , L′ 内-XY_4 ; Calculate the angle and intersection between straight lines, including determining L′ 内-XY All straight lines intersect, and calculate the angle and intersection point of the intersecting straight lines. When L′ 内-XY_1 , L′ 内-XY_2 , L′ 内-XY_3 , L′ 内-XY_4 When arranged counterclockwise on the XOY plane, the calculated L′ 内-XY_1 , L′ 内-XY_2 The angle between The intersection is L ′ 内-XY_2 , L′ 内-XY_3 The angle between The intersection is L′ 内-XY_3 , L′ 内-XY_4 The angle between The intersection is L′ 内-XY_4 , L′ 内-XY_1 The angle between The intersection is Make straight line adjustments, including converting the calculated Compared with the design angle of the steel bar component, the calculated angle difference is recorded as Using the straight line model L′ 内-XY Any straight line in is the reference line, assuming that the reference line is L′ 内-XY_1 , respectively, with the intersection point For the straight line L′ 内-XY_2 , L′ 内-XY_4 Adjust the angle to and After adjustment, L′ 内-XY_2 , L′ 内-XY_4 Note: L″ 内-XY_2 , L″ 内-XY_4 ; again with L″ 内-XY_4 As the baseline, calculate L' again 内-XY_3 With L″ 内-XY_4 Angle and intersection Calculation design angle and The difference by As the reference point, the angle For the straight line L′ 内-XY_3 Make adjustments and record the straight line after adjustment as L″ 内-XY_3 ; Thus completing the adjustment of 4 straight lines; Calculate the steel bar offset, including the straight line L′ 内-XY_1 , L″ 内-XY_2 , L″ 内-XY_3 , L″ 内-XY_4 Select points according to the designed spacing of the steel bars, and arrange all the points in order. The arranged point data is recorded as O_L′. -XY_1 , and the inner part of O_top′ -XY Each point in O_L′ is sequentially -XY_1 Subtract the data in O_top′ -XY The offset of all reinforcements in; For O_top′ -XY 、O_top′ outside -XY Execute the same as in O_top′ -XY The same steps as above are used to obtain O_top′ -XY 、O_top′ outside -XY The offset of all steel bars in the steel bar is measured to complete the steel bar quality inspection.

Images