Modular building flatness detection method, system and device based on unmanned vehicle

Abstract

The invention provides a modular building flatness detection method, system and device based on an unmanned vehicle, and relates to the technical field of modular building wall flatness detection, and the method comprises the steps: carrying out the scanning operation of a detected outer wall through a three-dimensional laser scanner carried by the unmanned vehicle, and obtaining the point cloud data of the outer wall; performing preprocessing and grid division on the outer wall point cloud data; based on the range line, extracting a facade point cloud file from the outer wall point cloud data; and calculating the flatness based on the facade point cloud file to obtain a detection result. According to the scheme, through multi-interval dynamic filtering, small noise interference can be effectively filtered out, and the detection precision is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of modular building wall flatness detection, in particular to a modular building flatness detection method, system and device based on unmanned vehicles. BACKGROUND

[0002] As one of the core indicators for measuring construction quality, the flatness of modular integrated building walls directly affects the appearance, waterproof performance and structural durability of buildings. According to the provisions of Article 12.3.8 of the "Standard for Acceptance of Quality of Building Decoration and Decoration Engineering" GB50210-2018, the allowable deviation of the flatness of the building exterior wall should be controlled within 4mm. In the context of large-scale development of high-rise buildings, manual detection methods such as using a ruler have been difficult to meet the dual demands of efficiency and accuracy in modern engineering, and technical upgrading is urgently needed.

[0003] The manual ruler detection method generally refers to the manual detection of the flatness of the building exterior wall using a ruler and a plug gauge. Single measurement can only cover an area of about 4 square meters. According to the calculation, 10,000 square meters of project requires about 2500 working days, which seriously restricts the construction progress and delivery cycle, and is difficult to meet the rapid detection needs of large-scale projects. Moreover, the measurement data is strongly subjective, has poor repeatability and low accuracy due to factors such as operator visual angle error, plug gauge insertion force, and ruler contact angle. In addition, due to the limitations of scaffold erection range and manual accessibility, the detection capability for complex parts such as exterior wall curved surfaces and cantilever structures is insufficient, and the overall detection coverage rate is often less than 60%, leaving many detection blind spots.

[0004] Although the non-contact technology based on image processing such as close-range photogrammetry has improved in visualization and data collection efficiency, it still faces constraints in practical application: artificial marker points need to be preset on the wall, which greatly increases the safety risk of high-altitude work and may damage the original wall decoration, which is not suitable for fine decoration or completion acceptance stage; the image stitching and reconstruction algorithm is easily disturbed by factors such as light and shooting angle, resulting in the accumulation of point position restoration errors, and the identification accuracy of local concave-convex parts of the wall is not high, and the overall interpretation accuracy is often less than 75%.

[0005] Due to the environmental interference such as scaffolding shielding and glass reflection, the original point cloud data contains a large amount of noise, and the traditional filtering algorithm (such as the curvature threshold method) cannot effectively distinguish the real wall surface features from the noise, resulting in a deviation of the flatness calculation of more than ±2mm, which cannot meet the control requirement of ±1mm level precision in the Unified Standard for Construction Engineering Quality Acceptance GB50300; the existing specification (GB50300 / GB50210) requires to take 2m ruler as the basis to evaluate the flatness, and the point cloud data generated by the three-dimensional scanning lacks a spatial scale conversion mechanism, can only output an overall model, and cannot automatically generate a ruler scale analysis result meeting the specification requirements, resulting in insufficient legal effect of the detection result; the existing software (such as Leica Cyclone) relies on manual interpretation of the heat map to locate the defect area, cannot automatically identify the micro concave-convex (±0.5mm level) and generate a quantitative repair scheme, and the manual analysis consumes more than 70% of the whole detection process, and the repair suggestion lacks accurate coordinate positioning and engineering quantity calculation basis. SUMMARY

[0006] The purpose of the present application is to provide a module integrated building wall flatness detection method, system and device based on unmanned vehicle, to solve at least one of the above technical problems in the prior art.

[0007] In a first aspect, to solve the above technical problems, the present application provides a module integrated building wall flatness detection method based on unmanned vehicle, comprising the following steps: Step 1, a three-dimensional laser scanner is carried on the unmanned vehicle to scan the measured outer wall and obtain outer wall point cloud data.

[0008] In a feasible embodiment, the unmanned vehicle comprises a positioning module and a computing workstation; the positioning module is used to realize the SLAM (simultaneous localization and mapping) system of the unmanned vehicle; and the computing workstation is used to analyze and calculate the point cloud data to obtain the flatness of the building outer wall.

[0009] In a feasible embodiment, the scanning operation comprises the following steps: Step 11, based on the preset target of the building peripheral ground, an environmental map is constructed by the SLAM system of the unmanned vehicle; Step 12, in the environmental map, a plurality of detection positions are sequentially arranged on the circumference with the measured outer wall as the center and with a preset scanning distance as the radius; and the preset target is circumscribed within the circumference. Step 13, after receiving the detection instruction, the unmanned vehicle autonomously drives to each detection position; at each detection position, the three-dimensional laser scanner is used to scan the measured outer wall in blocks according to the preset scanning floor interval, and the outer wall point cloud data at different scanning floor heights is obtained, which is numbered according to the scanning floor height; then the outer wall point cloud data collected at each detection position is numbered and summarized according to the detection position, so that the outer wall point cloud data containing the scaffold shielding part can be collected; Step 14, based on the fixed preset target position in the outer wall point cloud data, the outer wall point cloud data at different detection positions at the same scanning floor height is summed to obtain the complete outer wall point cloud data E at the scanning floor height; Step 15, in E, based on the preset scanning distance, the measured outer wall surface measuring point is determined; the point cloud with an elevation greater than the preset distance threshold from the measured outer wall surface measuring point is filtered out in whole to obtain the outer wall point cloud data after preliminary screening. In this way, the scaffold point cloud can be quickly and effectively preliminarily filtered out, and the outer wall point cloud shielded by the scaffold can be retained.

[0010] Step 2, pre-processing and grid division of the outer wall point cloud data.

[0011] In a feasible implementation, the pre-processing includes conventional noise filtering and coordinate unification.

[0012] In a feasible implementation, the coordinate unification refers to setting a scale factor of 0.001 when the unit of the CAD (Computer Aided Design) tool is millimeter, so as to ensure that the coordinate system of the outer wall point cloud data is consistent with the coordinate system of the CAD tool.

[0013] In a feasible implementation, the grid division includes: dividing a regular grid with a preset unit size; defining a grid row number (Y direction) in a north-to-south increasing manner and a grid column number (X direction) in a west-to-east increasing manner, numbering the grid according to the numbering format; and marking the three-dimensional coordinates of the grid center point as the grid coordinates of the grid.

[0014] Step 3, extracting the facade point cloud file from the outer wall point cloud data based on the range line.

[0015] In a feasible implementation, the step 3 specifically includes: Step 31, in the CAD tool, a line segment containing only two end points is generated as a range line: the starting point is placed on the left side of the facade, i.e. P point; and the ending point is placed on the right side of the facade, i.e. Q point. Step 32, through the property panel, a global width W is set, and W≥the maximum value of the Y direction range of the outer wall point cloud data; Step 33, after removing the external reference (i.e. target coordinate system) in the outer wall point cloud data, load it into the range line according to different scanning height, save as several range line files (DXF format); Step 34, through the conventional spatial clipping method, get several facade point cloud files (pts format) with RGB (color space) information from the range line files.

[0016] Step 4, based on the facade point cloud file, calculate the flatness to get the detection result.

[0017] In a feasible implementation, the step 4 specifically includes: Step 41, based on each facade point cloud file, construct the reference plane equation through the least square method, so as to calculate the flatness in sequence, and the specific equation includes: ; Wherein, 、 、 and respectively represent the plane equation constant; represents the X direction coordinate of the measuring point (in the facade point cloud file); represents the Y direction coordinate of the measuring point; represents the elevation of the measuring point; Step 42, construct the calculation formula of the distance of the measuring point to the reference plane and the nominal concave-convex degree , and the specific formula includes: ; ; Step 43, take as the virtual observation value, construct the concave-convex error equation of each measuring point, and the specific formula includes: ; Wherein, represents the concave-convex error; represents the serial number of the measuring point, and has ; represents the total number of the measuring points; Step 44, make the square sum of the concave-convex error of all measuring points minimum, solve the plane equation constant, and get the reference plane; Step 45, based on the reference plane, calculate the actual concave-convex value of each measuring point, and the specific formula includes: ; Wherein, represents the The actual concave-convex value of the measuring point, when the actual concave-convex value is 0, it means that the measuring point is located on the reference plane (i.e. no concave-convex), the larger the actual concave-convex value, the more serious the measuring point deviates from the reference plane, and the more uneven the outer wall surface is; 、 and respectively represent the normal vector components of the reference plane, used to describe the spatial direction of the reference plane; represents the constant term of the reference plane, used to reflect the offset amount of the reference plane and the coordinate origin; represents the X-direction coordinate of the first measuring point; represents the Y-direction coordinate of the first measuring point; represents the elevation of the first measuring point; represents the square root function; Step 46, by multi-interval dynamic filtering, the measuring points in the facade point cloud file are screened, so as to realize the precise screening of the measuring points.

[0018] In a feasible implementation, the step 46 specifically includes: Step 461, calculating the mean square error of the actual concave-convex value , used to quantify the degree of deviation of all measuring points from the reference plane; Step 462, by the first interval, the effective measuring points are screened out from all measuring points; the data fluctuation in the interval belongs to the normal construction error, so as to maintain the reliability of the overall analysis; the specific screening formula of the interval is: ; Wherein, represents the first dynamic screening threshold, the specific value range is 2 to 3 times of , which can be dynamically set based on the wall surface curvature change amount of the current measuring point in the preset radius neighborhood; Step 463, by the second interval, the defect measuring points are screened out from all measuring points and marked with color, so that the defect measuring points such as bulges or depressions can be marked; the specific screening formula of the interval is: ; Wherein, represents the second dynamic screening threshold, the specific value range is to + ; Step 464, by the third interval, the out-of-limit measuring points are screened out from all measuring points, such as sensor mis-detection, bird obstruction, glass reflection and other abnormal interference, so as to avoid the out-of-limit measuring points distorting the statistical results; the specific screening formula of the interval is: .

[0019] Step 47, construct a virtual ruler, calculate the flatness of each facade point cloud file relative to the virtual ruler and make a judgment.

[0020] In a feasible implementation, the step 47 specifically includes: Step 471, on the surface of the facade point cloud, a dynamic analysis grid of a preset size is constructed, and the grid node density is dynamically adjusted according to the amount of change in the curvature of the wall surface; Step 472, in each grid unit of the dynamic analysis grid, an ideal plane is fitted as a virtual ruler by using the least square method; Step 473, the virtual distance value of each measuring point to the virtual ruler is calculated as the flatness, the maximum deviation and the root mean square error (RMSE) of the flatness are counted and judged: when the maximum deviation and the root mean square error both meet the preset qualified threshold, the exterior wall flatness is judged to be qualified; thus the detection result meeting the current specification can be directly output, and the specific formula includes: ; Wherein, represents the maximum deviation value; represents the elevation of the th measuring point relative to the virtual ruler; represents the design elevation of the current measuring point (or ).

[0021] Step 48, identify the defect area of each facade point cloud by using the gradient descent method; combine edge detection to mark the boundary of the defect area and generate a defect distribution heat map; generate a corresponding repair scheme through a defect processing strategy.

[0022] In a feasible implementation, the step 48 specifically includes: Step 481, for each facade point cloud file, connect each measuring point to the center of the defect measuring point along the steepest direction, then take the center as the starting point to mark a continuous area of a preset length, and obtain the defect area; Step 482, scan each measuring point along the edge of the defect area by using an electronic chalk, and when the elevation difference between adjacent measuring points is greater than a preset elevation difference threshold, connect them into a closed loop as the boundary of the defect area; Step 483, after converting each facade point cloud file into a three-dimensional graph, mark the center coordinates, defect area boundary and maximum deviation value of each defect area; based on the preset single building floor height and the joint extension distance of the three-dimensional graph, fit and splice the three-dimensional graphs of different scanning floor heights in the vertical direction according to the floor sequence in the display frame to obtain an overall three-dimensional graph; color render the overall three-dimensional graph to obtain a defect distribution heat map; Step 484, based on the maximum deviation value, a corresponding repair scheme is obtained according to the defect processing strategy; The defect processing strategy comprises: For the convex defect: The maximum deviation value is in the range of 4-5mm, marked as a first-level defect, and the repair scheme is to polish to flatness by machine, and the polishing process parameters are calculated; the specific calculation formula of the polishing process parameters comprises: grinding head loss = polishing area x (maximum deviation value / 3); When the maximum deviation value is greater than 5mm, it is marked as a second-level defect, and the repair scheme is to first chisel off the convex surface layer, and then polish to flatness by machine, and then perform anti-cracking treatment; For the concave defect: The maximum deviation value is in the range of 4-5mm, marked as a first-level defect, and the repair scheme is to fill and repair by spraying mortar, and the filling process parameters are calculated; the calculation formula of the filling process parameters comprises: spraying thickness = maximum deviation value + 1 (mm); filling amount = defect area x average deviation value x 1.1 (10% loss coefficient); When the maximum deviation value is greater than 5mm, it is marked as a second-level defect, and the repair scheme is to process by hanging glass fiber mesh combined with layered filling.

[0023] In a second aspect, based on the same inventive concept, the application also provides a modular building flatness detection system based on an unmanned vehicle, comprising a data receiving module, a data processing module and a result generating module; The data receiving module is configured to receive external wall point cloud data. The data processing module comprises a preprocessing unit, a facade point cloud unit and a flatness unit. The preprocessing unit is configured to preprocess and grid divide the external wall point cloud data. The facade point cloud unit extracts a facade point cloud file from the external wall point cloud data based on a range line. The flatness unit calculates the flatness based on the facade point cloud file to obtain a detection result. The result generating module is configured to send the detection result.

[0024] In a third aspect, based on the same inventive concept, the application also provides a modular building flatness detection device based on an unmanned vehicle, comprising a processor, a memory and a bus, the memory stores instructions and data readable by the processor, and the processor is configured to call the instructions and data in the memory to execute the building external wall flatness detection method based on the unmanned vehicle as described above, and the bus is connected between the functional components for transmitting information.

[0025] In a feasible implementation, the device further comprises an unmanned vehicle carrying a three-dimensional laser scanner.

[0026] In an implementable embodiment, the unmanned vehicle comprises a positioning module and a computing workstation; the computing workstation comprises the processor, the memory and the bus.

[0027] In an implementable embodiment, the processor is a Jetson AGX Orin processor, which is a high-performance processor module for the field of edge AI (artificial intelligence) and autonomous machines, is based on an Ampere architecture and an ARM Cortex-A78 AEC CPU central processor, and has a server-level computing power and an energy efficiency optimization design.

[0028] By adopting the technical scheme, the present application has the following beneficial effects: The modularized building flatness detection method, system and device based on an unmanned vehicle provided by the present application can effectively filter out small noise interference and significantly improve detection accuracy, so that the noise misjudgment rate is reduced from 23.7% to 4.1%, and the flatness calculation deviation is controlled within ±0.5 mm, thereby meeting the control requirement of ±1 mm level accuracy in the current specification; The point cloud processing time can be compressed from 15 minutes to 90 seconds in combination with edge computing hardware; the positioning speed can reach 1000 square meters per second by using a gradient descent algorithm to identify continuous defect areas; the decision-making time can be shortened by 60% by using a quantitative repair scheme; and the daily detection area of the device can reach 50,000 square meters, thereby effectively improving detection and construction efficiency; Only 2 people are needed to operate the present application, which can reduce 80% of the detection labor input; the repair material consumption can be reduced by 40% by using a precise calculation repair scheme, and the repair material waste rate is only 3%; the original data volume is compressed by 90% by using edge computing, thereby effectively reducing cloud storage and transmission costs; The present application can perform non-contact flatness detection on various complex building outer wall structures, thereby reducing the safety risk of detection personnel; The present application can generate a flatness detection report in accordance with the current specification, and has industry compliance and authority. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0030] Figure 1A flow chart of a modular building flatness detection method based on an unmanned vehicle is provided for the embodiment of the present application. Figure 2 An overhead explanatory view about detection position setting is provided for the embodiment of the present application. Figure 3 A single analog gauge block diagram is provided for the embodiment of the present application. Figure 4 An analog gauge block application diagram is provided for the embodiment of the present application. Figure 5 A defect distribution thermodynamic diagram is provided for the embodiment of the present application. Figure 6 A modular building flatness detection system diagram based on an unmanned vehicle is provided for the embodiment of the present application. Figure 7 A modular building flatness detection device diagram based on an unmanned vehicle is provided for the embodiment of the present application. Reference signs: 1-three-dimensional laser scanner; 2-unmanned vehicle. DETAILED DESCRIPTION

[0031] The technical solutions of the present application will be described in detail below with reference to the drawings. Obviously, the described embodiments are part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0032] In the description of the present application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore cannot be understood as a limitation on the present application. In addition, the terms "first", "second", "third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0033] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0034] The present invention will be further explained below with reference to specific embodiments.

[0035] It should also be noted that the specific embodiments or implementation methods described below are a series of optimized settings listed by the present invention to further explain the specific content of the invention, and these settings can be combined or used in conjunction with each other.

[0036] Example 1: like Figure 1 As shown in the figure, this embodiment provides a modular building flatness detection method based on unmanned vehicles, which includes the following steps: Step 1: Using an unmanned vehicle equipped with a 3D laser scanner, the exterior wall to be measured is scanned to obtain point cloud data of the exterior wall.

[0037] Furthermore, the model of the 3D laser scanner is Faro Focus X130 (dot pitch 5mm@20m).

[0038] Furthermore, the unmanned vehicle includes a positioning module and a computing workstation; the positioning module is used to implement the SLAM (Simultaneous Localization and Mapping) system of the unmanned vehicle; the computing workstation is used to analyze and calculate the point cloud data to obtain the flatness of the building's exterior wall.

[0039] Furthermore, the scanning operation includes the following steps: Step 11: Based on the pre-set targets on the ground around the building, construct an environmental map using the SLAM system of the unmanned vehicle; Step 12: In the environmental map, with the exterior wall to be measured as the center and a preset scanning distance as the radius, set several detection positions (e.g., 90-degree angle directly facing the exterior wall, 45-degree angle side to the exterior wall, etc.) at preset angle intervals on a circle; the circle encloses the preset target. Step 13: After receiving the detection command, the unmanned vehicle autonomously drives to each detection location; at each detection location, the 3D laser scanner scans the exterior wall to be measured in blocks according to the preset scanning floor interval (e.g., every three floors), and obtains the exterior wall point cloud data at different scanning floor heights, and then numbers them according to the scanning floor height; then the exterior wall point cloud data collected at each detection location is summarized according to the detection location number, so that the exterior wall point cloud data including the parts covered by scaffolding can be collected; Preferably, the preset scanning floor interval k is not an integer multiple of the modular building unit floor m, and when scanning to the height of k×(m-1) floors, the value of k is changed. This ensures that the vertical joints between adjacent modular building units (which are most prone to flatness defects) appear only within a single block of exterior wall point cloud data, and not at the vertical splicing points between adjacent blocks of exterior wall point cloud data. Similarly, the horizontal width of the exterior wall point cloud data can be greater than the width of the measured exterior wall. When the horizontal width cannot be greater than the width of the measured exterior wall, a similar setting method as k can be used to ensure that the horizontal joints between adjacent modular building units appear only within a single block of exterior wall point cloud data, and not at the horizontal splicing points between adjacent blocks of exterior wall point cloud data. Step 14: Based on the fixed preset target position in the external wall point cloud data, sum the external wall point cloud data at different detection positions at the same scanned floor height to obtain the complete external wall point cloud data E for that scanned floor height; Step 15: In E, based on the preset scanning distance, determine the measuring points on the exterior wall surface to be measured; filter out point clouds whose elevation is greater than the preset distance threshold (e.g., 100mm) from the measuring points on the exterior wall surface to be measured, and obtain the preliminary filtered exterior wall point cloud data. This allows for a quick and effective initial screening of scaffolding point clouds while retaining point clouds of exterior walls obscured by scaffolding.

[0040] Furthermore, such as Figure 2 As shown, the specific method for setting the detection position includes: On the ground plane of the environmental map, take the central measuring point at the bottom of the measured exterior wall as the center point. Draw a circle with a preset scanning distance as the radius. ; Facing the exterior wall being measured, Top Directly The location is used as the first detection location; The formula for calculating the preset angle interval includes: ; in, Indicates the interval angle; This indicates the length of a horizontal unit of the scaffolding (i.e., the distance between scaffolding columns). This indicates the distance between the end of the scaffolding closest to the measured exterior wall and the measured exterior wall. Taking the first detection position as the 90-degree orientation, and moving towards any 0-degree orientation, As intervals, sequentially from Towards Draw several rays until the 0-degree orientation is reached; The first ray and the intersection of the second ray and the plane as a second detection position; the intersection of the second ray and the plane as a second detection position; the intersection of the second ray and the plane as a second detection position; and so on, to obtain all detection positions; In this way, the scanning laser of the three-dimensional laser scanner can pass through the frame structure of the scaffold by a shorter path and scan the entire measured outer wall surface, so as to be processed subsequently.

[0041] Step 2, preprocessing and grid division of the outer wall point cloud data.

[0042] Further, the preprocessing includes conventional noise filtering and coordinate unification.

[0043] Further, the coordinate unification refers to setting a scale factor as 0.001 when the unit of the CAD (Computer Aided Design) tool is millimeter, so as to ensure that the outer wall point cloud data and the coordinate system of the CAD tool are dimensionally uniform.

[0044] Further, the grid division includes: dividing a regular grid with a preset unit size (for example, 2m×2m); defining a grid row number (Y direction) in a self-north-to-south increasing manner and a grid column number (X direction) in a self-west-to-east increasing manner, numbering the grid according to a numbering format (for example, G-row number-column number), and marking the three-dimensional coordinates of the center point of the grid as the grid coordinates of the grid.

[0045] Step 3, extracting the facade point cloud file from the outer wall point cloud data based on the range line.

[0046] Further, the step 3 specifically includes: Step 31, generating a line segment containing only two end points in the CAD tool as a range line: the starting point is placed on the left side of the facade, that is, the P point; and the ending point is placed on the right side of the facade, that is, the Q point. Step 32, setting a global width W through a property panel, and W≥the maximum value of the Y direction range of the outer wall point cloud data. Step 33, after removing the external reference (that is, the target coordinate system) in the outer wall point cloud data, loading each according to different scanning heights into the range line, and saving as a plurality of range line files (DXF format); Step 34, obtaining a plurality of facade point cloud files (pts format) with RGB (color space) information through a conventional spatial clipping method on the range line file.

[0047] Step 4, calculating the flatness based on the facade point cloud file to obtain a detection result.

[0048] Further, the step 4 specifically includes: Step 41, based on each facade point cloud file, construct the reference plane equation by least square method, so as to calculate the flatness in sequence, the specific equation includes: ; Wherein, , , and respectively represent the plane equation constant; represents the X direction coordinate of the measuring point (in the facade point cloud file); represents the Y direction coordinate of the measuring point; represents the elevation of the measuring point; Step 42, construct the calculation formula of the distance from the measuring point to the reference plane and the nominal concave-convex degree , which specifically includes: ; ; Step 43, take as the virtual observation value, construct the concave-convex error equation of each measuring point, which specifically includes: ; Wherein, represents the concave-convex error; represents the serial number of the measuring point, and has ; represents the total number of measuring points; Step 44, minimize the square sum of the concave-convex error of all measuring points, solve the plane equation constant, and obtain the reference plane; For example, if there are 100 measuring points, from 0 to 99, according to the conditions (indirect adjustment method): , the coordinates of each measuring point are brought into the reference plane equation, and each plane equation constant can be solved, so as to determine the reference plane; Step 45, based on the reference plane, calculate the actual concave-convex value of each measuring point, which specifically includes: ; Wherein, represents the actual concave-convex value of the th measuring point, when the actual concave-convex value is 0, it means that the measuring point is located on the reference plane (i.e. no concave-convex), the larger the actual concave-convex value, the more serious the deviation of the measuring point from the reference plane, the more uneven the outer wall surface, and when the concave-convex direction needs to be defined, only the sign of the absolute value symbol needs to be judged; , and respectively represent the normal vector component of the reference plane, which is used to describe the spatial direction of the reference plane; The constant term representing the reference plane is used to reflect the offset of the reference plane from the origin of the coordinate system. Indicates the first The X-coordinates of each measuring point; Indicates the first The Y-coordinate of each measuring point; Indicates the first Elevation of each measuring point; Represents the square root function; Step 46: Through multi-interval dynamic filtering, the measurement points in the facade point cloud file are screened to achieve precise screening of the measurement points.

[0049] Furthermore, step 46 specifically includes: Step 461: Calculate the root mean square error of the actual concavity / convexity values. This is used to quantify the degree to which all measuring points deviate from the reference plane; Step 462: Select valid measuring points from all measuring points within the first interval. Data fluctuations within this interval are considered normal construction errors (such as fluctuations in mortar filling thickness) to maintain the reliability of the overall analysis. The specific selection formula for this interval is: ; in, This represents the first dynamic filtering threshold, specifically ranging from 2 to 3 times. It can be based on the current measuring point within a preset radius (e.g. The change in wall curvature within the neighborhood is dynamically set (e.g., when the change in wall curvature is greater than 100%). When, the value is 2 times. ); Step 463: Through the second interval, filter out defect measurement points from all measurement points and mark them with a color (e.g., black). This will mark defect measurement points such as bulges or dents. The specific filtering formula for this interval is: ; in, This represents the second dynamic filtering threshold, with a specific value range of [value missing]. to + ; Step 464: Through the third interval, filter out out-of-limit measuring points from all measuring points, such as sensor false detections, bird obstruction, glass reflections, and other abnormal interferences, thereby avoiding distortion of statistical results by out-of-limit measuring points; the specific screening formula for this interval is: .

[0050] Step 47: Construct a virtual straightedge, calculate and determine the flatness of the measurement points in each facade point cloud file relative to the virtual straightedge.

[0051] Furthermore, step 47 specifically includes: Step 471: On the surface of the facade point cloud, construct a dynamic analysis grid of a preset size (e.g., 2*2 meters). The grid node density (e.g., the initial value can be set to 50 points / square meter) is dynamically adjusted according to the change in wall curvature. Preferably, the specific calculation method for the change in wall curvature includes: Step 4711: Taking the current measurement point as the center, select measurement points within a preset range as the neighboring point set; fit the local geometric surface using the least squares method; Step 4712: Based on the local geometric surface, calculate the curvature quantization value of the local geometric surface using differential geometric formulas (e.g., the rate of change of the normal vector between measurement points); the curvature quantization value includes principal curvature and / or Gaussian curvature and / or mean curvature. Step 4713: Calculate the dispersion of the curvature quantification value as the change in wall curvature; the dispersion includes the standard deviation, the larger the standard deviation, the greater the change in wall curvature. Step 472: Within each grid cell of the dynamic analysis grid, fit an ideal plane using the least squares method, as a virtual guide, such as... Figure 3 The single black line shown in the image and Figure 4 As shown by the multiple gray straight lines in the image; Step 473: Calculate the virtual distance from each measuring point to the virtual straightedge as the flatness value; statistically analyze the maximum deviation and root mean square error (RMSE) of the flatness and make a judgment: when both the maximum deviation and the RMSE meet the preset qualified threshold, the flatness of the exterior wall is judged to be qualified; thus, the test results that comply with the current specifications can be directly output. The specific formula includes: ; in, Indicates the maximum deviation value (e.g.) Figure 3 In ); Indicates the first The elevation of each measuring point relative to the virtual straightedge; Indicates the design elevation of the current measuring point (or...) ).

[0052] Step 48: Identify the defect regions of each facade point cloud using the gradient descent method; combine edge detection to mark the boundaries of the defect regions and generate a defect distribution heatmap; generate corresponding repair schemes through defect handling strategies.

[0053] Furthermore, step 48 specifically includes: Step 481, for each facade point cloud file, a continuous area of a preset length (e.g. 4 mm) is marked from the center of the defect measurement point to the center of the defect measurement point after connecting the center of the defect measurement point to the center of the defect measurement point in the steepest direction from each measurement point, and the defect area is obtained; Step 482, by means of an electronic pencil, scanning is performed along the measurement points of the edge of the defect area until the elevation difference between adjacent measurement points is greater than a preset elevation difference threshold (e.g. 2 mm), and then the adjacent measurement points are connected to form a closed loop, which is taken as the boundary of the defect area; Step 483, after converting each facade point cloud file into a three-dimensional graph, the center coordinates, defect area boundary and maximum deviation value of each defect area are marked; based on a preset single floor height (e.g. 3.5 meters for 1-4 floors, and 2.9 meters for 5-50 floors) and a three-dimensional graph joint extension distance (e.g. 1.5 meters), the three-dimensional graphs of different scanning floor heights are vertically fitted and spliced into a display frame in sequence according to the floor, and a whole three-dimensional graph is obtained; color rendering is performed on the whole three-dimensional graph to obtain a defect distribution heat map, as shown in Figure 5 ; Step 484, based on the maximum deviation value, a corresponding repair scheme is obtained according to a defect treatment strategy; The defect treatment strategy includes: For convex defects: When the maximum deviation value is within the range of 4-5 mm, it is marked as a first-level defect, and the repair scheme is to polish it to be flat by a machine and calculate the polishing process parameters; the specific calculation formula of the polishing process parameters includes: grinding head loss = polishing area x (maximum deviation value / 3); When the maximum deviation value is greater than 5 mm, it is marked as a second-level defect, and the repair scheme is to remove the convex surface layer by chiseling, then polish it to be flat by a machine, and then perform anti-cracking treatment; For concave defects: When the maximum deviation value is within the range of 4-5 mm, it is marked as a first-level defect, and the repair scheme is to fill and repair by spraying special mortar, and to calculate the filling process parameters; the calculation formula of the filling process parameters includes: spraying thickness = maximum deviation value + 1 (mm); filling amount = defect area x average deviation value x 1.1 (10% loss coefficient); When the maximum deviation value is greater than 5 mm, it is marked as a second-level defect, and the repair scheme is to fill and repair by hanging glass fiber mesh combined with layering (the thickness of each filling layer is ≤3 mm).

[0054] Further, the method of the embodiment is applied to the flatness detection of the outer facade of a 50-story office building: The global width W is set to 35.6 meters (completely covering the facade in the Y direction); The first dynamic screening threshold is set to 2.5; The second dynamic screening threshold is set to 3.0; The monomer floor height is set to: 1-30 floors: 4.2 meters; 31-50 floors: 3.8 meters; The calculation of the flatness result includes: the mean square error of the actual concave-convex value = 4 mm; The largest defect area is located at the 32nd floor, and the maximum deviation value = 12.7 mm.

[0055] Example Two: As shown in the figure, the embodiment provides a modular building flatness detection system based on unmanned vehicles, which includes a data receiving module, a data processing module, and a result generating module. Figure 6 The data receiving module is used to receive external wall point cloud data. The data processing module includes a preprocessing unit, a facade point cloud unit, and a flatness unit. The preprocessing unit is used to preprocess and grid divide the external wall point cloud data. The facade point cloud unit extracts facade point cloud files from the external wall point cloud data based on range lines. The flatness unit calculates the flatness based on the facade point cloud files to obtain the detection result. The result generating module is used to send out the detection result.

[0056] Example Three: As shown in the figure, the embodiment provides a modular building flatness detection device based on unmanned vehicles, which includes a processor, a memory, and a bus. The memory stores instructions and data that can be read by the processor. The processor is used to call the instructions and data in the memory to execute the building external wall flatness detection method based on unmanned vehicles as described above. The bus connects each functional component for transmitting information. Figure 7 Further, the device further includes an unmanned vehicle 2 equipped with a three-dimensional laser scanner 1.

[0057] Further, the unmanned vehicle 2 includes a positioning module and a computing workstation; the computing workstation includes the processor, the memory, and the bus.

[0058]

[0059] ​​Further, the processor is a Jetson AGX Orin processor, which is a high-performance processor module for the field of edge AI (artificial intelligence) and autonomous machines, is based on an Ampere architecture and an ARM Cortex-A78 AECPU central processor, and has a server-level computing power and an energy efficiency optimization design.

[0060] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A modular building flatness detection method based on unmanned vehicle, characterized in that, The method comprises the following steps: Step 1, scanning the measured module integrated building wall by a three-dimensional laser scanner carried by an unmanned vehicle to obtain point cloud data of the outer wall; Step 2, preprocessing and grid division of the point cloud data of the outer wall; Step 3, extracting facade point cloud files from the point cloud data of the outer wall based on a range line; Step 4, calculating flatness based on the facade point cloud files to obtain a detection result.

2. The method of claim 1, wherein, The preprocessing comprises noise filtering and coordinate unification.

3. The method of claim 1, wherein, The grid division comprises: dividing a regular grid with a preset unit size; defining a grid row number in a north-to-south increasing manner and a grid column number in a west-to-east increasing manner, numbering the grid according to a numbering format, and marking the three-dimensional coordinates of the center point of the grid as the grid coordinates of the grid.

4. The method of claim 1, wherein, Step 3 specifically comprises: Step 31, generating a line segment containing only two end points as a range line in a CAD tool: the start point is placed on the left side of the facade, and the end point is placed on the right side of the facade; Step 32, setting a global width W through a property panel, and W is greater than or equal to the maximum value of the Y-direction range of the point cloud data of the outer wall; Step 33, after removing the external reference in the point cloud data of the outer wall, loading each scanning height into the range line one by one and saving as a plurality of range line files; Step 34, obtaining a plurality of facade point cloud files with RGB information through a conventional spatial clipping method on the range line files.

5. The method of claim 1, wherein, Step 4 specifically comprises: Step 41, constructing a reference plane equation based on each facade point cloud file by a least square method, and the specific equation comprises: ； wherein, , , and represent plane equation constants, respectively; represents the X coordinate of the survey point; represents the Y coordinate of the survey point; represents the elevation of the survey point; Step 42, constructing the distance from the measuring point to the reference plane and the nominal degree of concave-convex The calculation formula specifically includes: ； ； Step 43, with As a virtual observation value, the concave-convex error equation of each measuring point is constructed, and the specific formula includes: ； wherein, represents the concave-convex degree error; represents the measuring point serial number, and has ; represents the total number of measuring points; Step 44, minimizing the square sum of concave-convex error of all measurement points to solve the constant of the plane equation and obtain a reference plane; Step 45, calculating the actual concave-convex value of each measurement point based on the reference plane, and the specific formula comprises: ； wherein, represents the actual relief value of the th measurement point; , and respectively represent the normal vector components of the reference plane; represents the constant term of the reference plane; represents the X-direction coordinate of the th measurement point; represents the Y-direction coordinate of the th measurement point; represents the elevation of the th measurement point; represents the square root function; Step 46, screening the measurement points in the facade point cloud file through multi-interval dynamic filtering; Step 47, constructing a virtual ruler, calculating the flatness of the measurement points in each facade point cloud file relative to the virtual ruler, and performing judgment; Step 48, identifying the defect area of each facade point cloud through a gradient descent method; combining edge detection to mark the boundary of the defect area to generate a defect distribution heat map; and generating a corresponding repair scheme through a defect processing strategy.

6. The method of claim 5, wherein, Step 46 specifically comprises: Step 461, calculating mean square deviation of actual concavo-convex value ; Step 462, screening effective measurement points from all measurement points through a first interval; the specific screening formula of the interval is: ； wherein, represents the first dynamic screening threshold value, and the specific value range is 2 to 3 times of ; Step 463, screening defect measurement points from all measurement points and marking the color through a second interval; the specific screening formula of the interval is: ； wherein, represents the second dynamic screening threshold, and the specific value range is to + ; Step 464, screening out out-of-limit measurement points from all measurement points through a third interval; the specific screening formula of the interval is: 。 7. The method of claim 5, wherein, Step 47 specifically comprises: Step 471, constructing a dynamic analysis grid of a preset size on the surface of the facade point cloud, and the grid node density is dynamically adjusted according to the change amount of the wall surface curvature; Step 472, fitting an ideal plane as a virtual ruler in each grid unit of the dynamic analysis grid through a least square method. Step 473, calculate the virtual distance value of each measuring point to the virtual ruler as the flatness; calculate the maximum deviation and root mean square error of the flatness and make a judgment: when the maximum deviation and root mean square error meet the preset qualified threshold, the judgment is that the flatness of the outer wall is qualified; the specific formula includes: ； wherein, represents the maximum deviation value; represents the elevation of the th measurement point relative to the virtual straight edge; represents the design elevation of the current measurement point.

8. The method of claim 5, wherein, The step 48 specifically includes: Step 481, for each facade point cloud file, connect the center of the defect measuring point from each measuring point along the steepest direction, mark the continuous area of a preset length taken from the center as the starting point, and obtain the defect area; Step 482, scan the measuring points along the edge of the defect area by an electronic chalk until the elevation difference between adjacent measuring points is greater than a preset elevation difference threshold, then connect to form a closed loop as the defect area boundary; Step 483, after converting each facade point cloud file into a three-dimensional graph, mark the center coordinates, defect area boundary and maximum deviation value of each defect area; based on the preset single floor height and three-dimensional graph joint extension distance, the three-dimensional graphs of different scanning floor heights are vertically fitted and spliced into a display frame in order of floor to obtain an overall three-dimensional graph; color rendering is performed on the overall three-dimensional graph to obtain a defect distribution heat map; Step 484, based on the maximum deviation value, a corresponding repair scheme is obtained according to the defect treatment strategy; The defect treatment strategy includes: For protruding defects: When the maximum deviation value is in the range of 4-5mm, it is marked as a first-level defect, and the repair scheme is to polish to flatness by machine, and the polishing process parameters are calculated; the specific calculation formula of the polishing process parameters includes: grinding head loss = polishing area x (maximum deviation value / 3); When the maximum deviation value is greater than 5mm, it is marked as a second-level defect, and the repair scheme is to remove the protruding surface layer by chiseling, then polish to flatness by machine, and then perform anti-cracking treatment; For concave defects: When the maximum deviation value is in the range of 4-5mm, it is marked as a first-level defect, and the repair scheme is to fill and repair by spraying mortar, and the filling process parameters are calculated; the calculation formula of the filling process parameters includes: spraying thickness = maximum deviation value + 1; filling amount = defect area x average deviation value x 1.1; When the maximum deviation value is greater than 5mm, it is marked as a second-level defect, and the repair scheme is to treat by hanging glass fiber mesh combined with layered filling.

9. A modular building flatness detection system based on unmanned vehicles, characterized by, It includes a data receiving module, a data processing module and a result generating module; The data receiving module is used to receive the outer wall point cloud data; The data processing module includes a preprocessing unit, a facade point cloud unit and a flatness unit; The preprocessing unit is used to preprocess and grid divide the outer wall point cloud data; The facade point cloud unit extracts the facade point cloud file from the outer wall point cloud data based on the range line; The flatness unit calculates the flatness based on the facade point cloud file to obtain the detection result; The result generating module is used to send out the detection result.

10. A modular building flatness detection device based on unmanned vehicle, characterized in that, It includes a processor, a memory and a bus, the memory stores instructions and data read by the processor, the processor is used to call the instructions and data in the memory to execute the method in any one of claims 1-8, and the bus is connected between each functional component for transmitting information.

Images