Hardware size error detection method and system based on point cloud processing

Abstract

The present invention relates to the technical field of hardware dimension measurement and detection, and in particular to a method and system for hardware dimension error detection based on point cloud processing. The present invention first obtains the point cloud data of the hardware and performs preliminary density clustering on the point cloud data; further analyzes from multiple angles the distribution similarity characteristics of the target data point and other data points in the preliminary density clustering cluster to which it belongs, the repeated characteristics of the curvature, and the similar characteristics of the data density of the target data point, and obtains the hardware membership of the target data point; further adjusts the distance between the target data point and the center point in the preliminary density clustering cluster to which it belongs according to the hardware membership, performs secondary density clustering and obtains denoised point cloud data; finally, performs dimension error detection on the hardware based on the denoised point cloud data. The present invention analyzes the possibility that the data point belongs to the hardware area from multiple angles, optimizes the density clustering process, accurately removes the noise interference in the point cloud data, and obtains reliable hardware processing product dimension error detection results.

Description

Technical Field

[0001] The present invention relates to the technical field of dimensional metrology detection, and in particular to a hardware dimensional error detection method and system based on point cloud processing. Background Art

[0002] In construction projects, the dimensional accuracy of hardware components is crucial to ensuring the safety and stability of building structures. These components, such as rebar, embedded parts, connectors, and gears, require high dimensional and shape positional accuracy to achieve precise connections and load-bearing within the building structure. Dimensional error detection can identify deviations in key parameters such as length, diameter, and angle of these critical components, enabling timely adjustments to the manufacturing and installation processes to ensure that each component meets design requirements and ensures its safety and reliability in various critical applications.

[0003] Traditional methods for inspecting the dimensions of construction hardware rely primarily on manual measurement and simple measuring tools, such as tape measures and calipers. These methods are inefficient, have limited accuracy, and are susceptible to human influence. In recent years, with the rapid development of three-dimensional scanning technology, dimensional error detection methods based on 3D scanning have gradually been applied to the construction industry. By using a 3D scanner to perform a comprehensive surface scan of construction hardware, and then using data processing software to analyze the hardware's dimensional errors and shape deviations, this method can quickly acquire data on complex shapes and perform comprehensive geometric analysis. This method significantly improves inspection efficiency and accuracy, and is particularly advantageous when inspecting complex-shaped components.

[0004] However, 3D scanning technology also faces challenges in practical applications. For example, when converting point cloud data into a geometric model, noise and useless data must be removed. However, construction hardware often has complex geometries and subtle surface features, such as the ribs of rebar, the anchoring sections of embedded parts, and tiny holes. These features can be mistakenly identified as noise or deleted by filtering algorithms. This can lead to incomplete or inaccurate scan results, affecting the measurement accuracy of construction hardware product dimensions and causing inaccurate error detection results.

[0005] Therefore, it is necessary to develop more accurate and reliable detection methods and systems for dimensional error detection of hardware used in construction projects. These methods must be able to effectively process complex point cloud data, preserve key geometric features, and accurately identify and measure dimensional errors. Furthermore, the on-site construction environment, such as the impact of factors like light and dust on scanning results, must be considered to ensure the stability and reliability of the detection system in practical applications. By continuously improving and innovating detection technologies, the quality and safety of construction projects can be better guaranteed. Summary of the Invention

[0006] In order to solve the technical problem of inaccurate hardware point cloud data denoising, which affects hardware dimensional error detection, the purpose of the present invention is to provide a hardware dimensional error detection method and system based on point cloud processing. The technical solutions adopted are as follows:

[0007] A hardware dimension error detection method based on point cloud processing, the method comprising:

[0008] Acquire point cloud data of the hardware; perform preliminary density clustering on the point cloud data; select any data point in any preliminary density clustering cluster as a target data point;

[0009] The hardware performance degree of the target data point is obtained based on the uniform distribution characteristics of the data points in the local neighborhood of the target data point; the hardware possibility of the target data point is obtained based on the similarity characteristics of the hardware performance degrees of the target data point and other data points in the preliminary density clustering cluster to which it belongs, combined with the repetitive characteristics of the curvature of the target data point and other data points in the point cloud data; the density uniformity parameter of the target data point is obtained based on the similarity characteristics of the data density of the target data point and the data density of all data points within the preset neighborhood range of the target data point;

[0010] fusing the hardware probability and the density uniformity parameter of the target data point to obtain the hardware membership of the target data point; the hardware probability and the density uniformity parameter are both positively correlated with the hardware membership; adjusting the distance between the target data point and the center point of the initial density cluster to which it belongs based on the hardware membership to perform secondary density clustering; and obtaining denoised point cloud data based on the secondary density clustering result;

[0011] Dimensional error detection is performed on hardware based on the denoised point cloud data.

[0012] Furthermore, the method for obtaining the hardware performance level includes:

[0013] A surface formed by the data points in the intersection of the first preset neighborhood of the target data point and the preliminary density clustering cluster to which it belongs is used as the target surface of the target data point; and the hardware performance level of the target data point is obtained based on the uniform distribution characteristics of the data points in the target surface.

[0014] Furthermore, the method for obtaining the hardware performance level of the target data point according to the uniform distribution characteristics of the data point in the target curved surface includes:

[0015] Within the target surface, with the target data point as the origin, data points to be analyzed are selected in a preset direction away from the origin; based on the consistency of distances between all adjacent data points to be analyzed in the same preset direction, the hardware performance level of the target data point is obtained; the distance consistency and the hardware performance level are positively correlated.

[0016] Furthermore, the method for obtaining the curvature includes:

[0017] The maximum curvature of the target data point in all the preset directions is taken as the curvature of the target data point.

[0018] Furthermore, the method for obtaining the hardware possibility includes:

[0019] Obtaining a first possibility parameter of the target data point based on similarity characteristics of the hardware performance between the target data point and other data points in the preliminary density clustering cluster to which it belongs; the similarity characteristics of the hardware performance are positively correlated with the first possibility parameter;

[0020] In the point cloud data, a second possibility parameter of the target data point is obtained according to the number of other data points having the same curvature as the target data point; the number of other data points having the same curvature is positively correlated with the second possibility parameter;

[0021] The hardware possibility of the target data point is obtained according to the first possibility parameter and the second possibility parameter of the target data point; the first possibility parameter and the second possibility parameter are both positively correlated with the hardware possibility.

[0022] Furthermore, the method for obtaining the density uniformity parameter includes:

[0023] Obtaining a data density of each data point based on a quantity characteristic of data points within a second preset neighborhood of each data point;

[0024] The density uniformity parameter of the target data point is obtained based on similar features between the data density of the target data point and the data density of all data points within a preset neighborhood of the target data point; the similar features of the data density are positively correlated with the density uniformity parameter.

[0025] Furthermore, the method of performing secondary density clustering by adjusting the distance between the target data point and the center point of the primary density clustering cluster to which it belongs according to the hardware membership includes:

[0026] Performing negative correlation mapping and normalizing on the hardware membership of the target data point to obtain a distance correction coefficient of the target data point;

[0027] The product of the distance between the target data point and the center point of the preliminary density clustering cluster to which it belongs and the distance correction coefficient is used as the corrected distance between the target data point and the center point of the preliminary density clustering cluster to which it belongs, and secondary density clustering is performed according to the corrected distance.

[0028] Furthermore, the method for detecting hardware size errors based on the denoised point cloud data includes:

[0029] The denoised point cloud data is converted into a hardware surface using Poisson surface reconstruction. The reconstructed hardware surface is converted into a mesh model using a triangulated meshing method. The measurement data includes at least the diameter, number of teeth, tooth shape and tooth pitch of the hardware. The dimensional error of the hardware is obtained by comparing the measured data with the design specification data.

[0030] Furthermore, the preliminary density clustering and the secondary density clustering both use the DBSCAN clustering algorithm.

[0031] The present invention also proposes a hardware dimension error detection system based on point cloud processing, which includes a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, it implements any one of the steps of the hardware dimension error detection method based on point cloud processing.

[0032] The present invention has the following beneficial effects:

[0033] The present invention first obtains the point cloud data of the hardware and performs preliminary density clustering on the point cloud data, so as to facilitate the subsequent analysis of the preliminary density clustering results and optimize the clustering process; further obtains the degree of hardware performance from the perspective of uniform distribution, evaluates the possibility that the target data point belongs to the hardware, and prepares for the hardware possibility of the subsequent target data point; further makes full use of the structural symmetry and local structural consistency of the hardware to obtain the hardware possibility of the target data point, and provides more basis for the subsequent optimization clustering to improve the denoising effect; further obtains the density of the target data point based on the similarity between the data density of the target data point and the data density of all data points within the preset neighborhood range of the target data point. Uniform parameters provide a basis for distinguishing defective data points from noise data points; further integrate the hardware possibility and density uniformity parameters of the target data points, and from the distribution similarity angle, the hardware structure symmetry angle and the density similarity angle, the possibility of the target data point being the point cloud data of the hardware is obtained, and the hardware membership of the target data point is obtained, which provides a reliable basis for subsequent optimization of clustering and accurate denoising of the point cloud data; further optimize the clustering process according to the hardware membership, perform secondary density clustering and obtain denoised point cloud data, better reflect the real local density changes, improve the accuracy of the clustering results, and enhance the accuracy of dimensional error detection; finally, perform dimensional error detection on the hardware based on the denoised point cloud data. The present invention analyzes the possibility of data points belonging to the hardware area from multiple angles, optimizes the density clustering process, accurately removes noise interference in the point cloud data, and obtains reliable dimensional error detection results for hardware processing products. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 A flowchart of a method for detecting hardware size errors based on point cloud processing provided by one embodiment of the present invention;

[0036] Figure 2 A flowchart of a method for obtaining gear possibilities provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0037] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of the hardware dimensional error detection method and system based on point cloud processing proposed by the present invention. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0038] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0039] The following describes in detail the hardware dimension error detection method and system based on point cloud processing provided by the present invention with reference to the accompanying drawings.

[0040] See also Figure 1 , which shows a flowchart of a hardware size error detection method based on point cloud processing provided by an embodiment of the present invention, specifically comprising:

[0041] In construction projects, there are many types of hardware components, such as steel bars, embedded parts, connectors, gears, etc., which can all be detected using point cloud methods. This application takes gear hardware as an example.

[0042] Step S1: Acquire point cloud data of the gear; perform preliminary density clustering on the point cloud data; and select any data point in any preliminary density clustering cluster as a target data point.

[0043] In an embodiment of the present invention, during the gear error detection process, a blackout curtain and a reflective plate are used to control the ambient light, a 3D scanner laser is used to scan the gear, the gear is firmly placed on a turntable, the angle and distance between the scanner and the gear are adjusted to ensure that the entire surface of the gear can be scanned, and the point cloud data of the gear is obtained. The scanned data is then imported into dedicated software for fusion processing to generate a complete gear point cloud model.

[0044] Considering that the density clustering algorithm performs preliminary classification on the point cloud data of the gear, it can cluster points with similar distances into clusters based on the density relationship between data points, effectively distinguishing potential gear structures from non-gear data. After the preliminary density clustering, the various characteristics of the data points are further analyzed, the distribution of the data points is adjusted, the density clustering is re-executed, and the cluster division is optimized, so as to more accurately capture the structural characteristics of the gear and eliminate those noise points that obviously deviate from the gear model to ensure the final gear surface reconstruction quality. Therefore, the point cloud data is first subjected to preliminary density clustering; any data point in any preliminary density clustering cluster is selected as the target data point.

[0045] Preferably, in one embodiment of the present invention, the DBSCAN clustering algorithm is used for preliminary density clustering. In the point cloud data, the Euclidean distance between any two points is obtained, the Euclidean distance corresponding to each data point is sorted from large to small, and the kth nearest Euclidean distance corresponding to each data point is taken as the k distance of each data point; the k distances of all data points in the point cloud data are sorted from small to large and then a k distance change curve is fitted; the k distance corresponding to the maximum bending point on the k distance change curve is used as the neighborhood distance; k+1 is used as the minimum number of points to perform DBSCAN clustering.

[0046] As an example, The maximum bending point on the k-distance variation curve is directly obtained by the maximum curve bending point (L-curve method), which is an existing technology and will not be described in detail.

[0047] It should be noted that the initial density clustering may mistakenly isolate the data points in the gear area as noise points. To avoid this situation, the isolated points after the initial density clustering are divided into the nearest cluster for further analysis.

[0048] Step S2: Based on the uniform distribution characteristics of data points in the local neighborhood of the target data point, the gear performance degree of the target data point is obtained; based on the similar characteristics of the gear performance degree of the target data point and other data points in the preliminary density clustering cluster to which it belongs, combined with the repetitive characteristics of the curvature of the target data point and other data points in the point cloud data, the gear possibility of the target data point is obtained; based on the similar characteristics of the data density of the target data point and the data density of all data points within the preset neighborhood range of the target data point, the density uniformity parameter of the target data point is obtained.

[0049] Ideal gear point cloud data has a uniform density distribution, while noise appears as scattered outliers or low-density areas. At the same time, the tooth surface of the gear should appear as a smooth and continuous point cloud structure. Therefore, we first obtain the gear performance of the target data point based on the uniform distribution characteristics of the data points in the local neighborhood of the target data point, and evaluate the possibility that the target data point belongs to the gear from the perspective of uniform distribution. This is the gear possibility for subsequent target data points, and finally adjusts the distance between the target data point and the center point of the initial density clustering cluster to which it belongs, in preparation for secondary density clustering.

[0050] Preferably, in one embodiment of the present invention, considering that the data points in the neighborhood of the target data point do not necessarily belong to the same initial density clustering cluster as the target data point, the distance between the data points of different clustering clusters is relatively far, and the distribution characteristics themselves are different, which cannot be used as the basis for analyzing the gear performance degree of the target data point. Therefore, it is necessary to use the surface formed by the data points in the intersection of the first preset neighborhood of the target data point and the initial density clustering cluster to which it belongs as the target surface of the target data point; limit the analysis range to the target surface, regard the target surface as the local neighborhood of the target data point, and obtain the gear performance degree of the target data point based on the uniform distribution characteristics of the data points in the target surface.

[0051] As an example, the first preset neighborhood is an area formed by a sphere with a radius of 5, the center of the sphere is the target data point, and the unit length is 1 mm.

[0052] In other embodiments of the present invention, the implementer may also set other first preset neighborhoods.

[0053] Preferably, in one embodiment of the present invention, within the target surface, with the target data point as the origin, data points to be analyzed are selected in a preset direction away from the origin; the gear performance degree of the target data point is obtained based on the distance consistency between all adjacent data points to be analyzed in the same preset direction; only by analyzing the distribution characteristics of the data points in the preset direction, the data uniform distribution characteristics of the entire surface are represented, the analysis process is simplified, and the gear detection efficiency is improved, wherein the distance consistency and the gear performance degree are positively correlated.

[0054] As an example, there are 9 preset directions, which are evenly distributed. By obtaining the surface normal passing through the target data point, a plane perpendicular to the normal is constructed on the normal, and 9 preset directions are evenly selected within the plane and projected back onto the surface. At this time, it is possible to select the data points to be analyzed in the target surface with the target data point as the origin and in the preset direction away from the origin. The calculation formula for the gear performance degree includes:

[0055] ;

[0056] in, Indicates the The gear performance level of each target data point; Expressed as a natural constant is the exponential function of the base; Indicates the total number of preset directions; The serial number indicating the preset direction; Indicates the The number of data points to be analyzed in a preset direction; Indicates the sequence number of the data point to be analyzed; Indicates the The target data point The first The distance parameter corresponding to the data points to be analyzed; Indicates the The target data point The average value of all distance parameters in a preset direction; Indicates the The target data point The variance of all distance parameters in a preset direction.

[0057] In the calculation formula of the gear performance degree, the distance parameter corresponding to the target data point in the preset direction, the overall difference from the average value of the distance parameter, and the variance of the distance parameter are used to express the consistency of the distance between adjacent data points to be analyzed in the same preset direction. The smaller the difference between the distance parameter and the average value of the distance parameter, the better the performance. The smaller it is, the more consistent the distance parameters are, the stronger the consistency of the distances between adjacent data points to be analyzed in the preset direction, the more uniform the distribution, and the greater the gear performance. The larger it is, the stronger the fluctuation of the distance parameter is, the weaker the distance consistency between adjacent data points to be analyzed in the preset direction is, the more uneven the distribution is, and the smaller the gear performance is.

[0058] It should be noted that the distance parameter of the data point to be analyzed is the Euclidean distance between the data point to be analyzed and the data point close to the origin in the same preset direction. For example, the distance parameter corresponding to the first data point to be analyzed in a preset direction is the Euclidean distance from the origin to the first data point to be analyzed; the distance parameter corresponding to the second data point to be analyzed is the Euclidean distance from the first data point to be analyzed to the second data point to be analyzed.

[0059] In other embodiments of the present invention, the implementer may also select data points to be analyzed on the surface by means of CAD models or finite element analysis networks; or adopt a weighted summation method to analyze the data points. and A weighted summation is performed, for example, the weights are 0.6 and 0.4 respectively, to calculate the gear performance degree; other numbers of preset directions can also be set, which are technical means well known to those skilled in the art and will not be described in detail here.

[0060] In an embodiment of the present invention, in order to ensure that noise can be identified as a separate outlier rather than mistakenly classified as a gear area, it is necessary to accurately determine whether each data point in the point cloud data belongs to the gear area. Taking into account the structural symmetry and local structural consistency of the gear, the possibility that the data point belongs to the gear can be evaluated by analyzing the similarity of the gear performance between the data points and the repeatability of the geometric properties of the data points. Therefore, based on the similarity of the gear performance between the target data point and other data points in the preliminary density clustering cluster to which it belongs, combined with the repetitive characteristics of the curvature of the target data point and other data points in the point cloud data, the gear possibility of the target data point is obtained, providing more basis for the subsequent adjustment of the distance between the target data point and the center point in the preliminary density clustering cluster to which it belongs.

[0061] Preferably, in one embodiment of the present invention, the maximum curvature of the target data point in all preset directions is used as the curvature of the target data point.

[0062] Preferably, in one embodiment of the present invention, the method for obtaining the gear possibility includes:

[0063] See also Figure 2 , which shows a flow chart of a method for obtaining gear possibilities provided by an embodiment of the present invention, specifically comprising:

[0064] Step S201: obtaining a first possibility parameter of the target data point based on similar characteristics of the gear performance between the target data point and other data points in the preliminary density clustering cluster to which it belongs.

[0065] Among them, the similarity characteristics of the gear performance degree are positively correlated with the first possibility parameter.

[0066] Considering that the more similar the gear performance of a target data point is to other data points in the same cluster, the smaller the difference, indicating that the target data point is more likely to be a data point in the gear region and less likely to be a noise point, the first probability parameter of the target data point is obtained based on the similarity characteristics of the gear performance between the target data point and other data points in the initial density cluster. The similarity characteristics of the gear performance are positively correlated with the first probability parameter.

[0067] As an example, the gear performance degree of the target data point and the gear performance degree of other data points in the same preliminary density cluster are averaged by the negative correlation function. After negative correlation mapping, it is used as the first possibility parameter of the target data point. Expressed as a natural constant is the exponential function of the base; Represents the independent variable.

[0068] The difference characteristics of the gear performance degree of the target data point and the gear performance degrees of other data points in the same preliminary density cluster are expressed by the absolute value of the difference. The larger the absolute value of the difference, the more obvious the difference characteristics. After negative correlation mapping, the similar characteristics of the gear performance degrees are expressed, and the first possibility parameter is obtained. The greater the similar characteristics of the gear performance degrees, the larger the first possibility parameter.

[0069] In other embodiments of the present invention, the implementer may also perform negative correlation mapping by taking the inverse, converting the difference features of the gear performance levels into similar features, and obtaining the first possibility parameter.

[0070] Step S202: In the point cloud data, obtain a second possibility parameter of the target data point based on the number of other data points having the same curvature as the target data point.

[0071] The number of other data points with the same curvature is positively correlated with the second likelihood parameter.

[0072] Considering that the more data points in the point cloud data that have the same curvature as the target data point, the stronger the symmetry of the gear presented by the target data point, the more likely it is a gear data point, and the greater the possibility of the gear, the second possibility parameter of the target data point is obtained according to the number of other data points with the same curvature as the target data point in the point cloud data; the number of other data points with the same curvature is positively correlated with the second possibility parameter.

[0073] As an example, the number of data points in the point cloud data with the same curvature as the target data point is mapped to a positive correlation normalization function In the function, the mapping value is used as the second likelihood parameter. The number of data points in the point cloud data that share the same curvature as the target data point reflects the repetitive characteristics of the curvature of the target data point and other data points in the point cloud data. The more data points that share the same curvature as the target data point, the more obvious the repetitive characteristics of the curvature, and the more likely it is a gear data point. The greater the likelihood of a gear, the larger the second likelihood parameter.

[0074] In other embodiments of the present invention, implementers may also use other positive correlation normalization functions such as Among the functions, the ones mentioned above are all existing commonly used functions and will not be described in detail.

[0075] Step S203: Obtain the gear likelihood of the target data point according to the first likelihood parameter and the second likelihood parameter of the target data point.

[0076] Among them, the first possibility parameter and the second possibility parameter are both positively correlated with the gear possibility.

[0077] After obtaining the first and second likelihood parameters, the first and second likelihood parameters can be fused to obtain the gear likelihood of the target data point. The first and second likelihood parameters are both positively correlated with the gear likelihood.

[0078] As an example, a multiplication method is used to represent positive correlation, and the product of the first possibility parameter and the second possibility parameter is used as the gear possibility of the target data point.

[0079] In other embodiments of the present invention, the implementer may also express positive correlation by weighted summation, for example, assigning weights of 0.6 and 0.4 to the first possibility parameter and the second possibility parameter respectively, and using the summation result as the gear possibility of the target data point.

[0080] Gear processing products may have manufacturing defects during the production process, and these defects will have a significant impact on the products. In the process of denoising gear point cloud data, the defects are part of the gear and need to be retained, so it is necessary to distinguish between defective data points and noise data points; considering that the noise data points exist independently, the density difference between the noise points and the data points in the neighborhood is large, while the density of the defective data points and other data points in the neighborhood changes continuously or uniformly, and can be distinguished by the density change between the data points and the data points in the neighborhood. Therefore, the density uniformity parameter of the target data point is obtained based on the similarity between the data density of the target data point and the data density of all data points within the preset neighborhood range of the target data point.

[0081] Preferably, in one embodiment of the present invention, considering that the more data points there are in a fixed neighborhood of a data point, the greater the density of data points around the data point, the data density of each data point is obtained based on the quantity characteristics of the data points in the second preset neighborhood of each data point, and the data density is represented by means of the quantity characteristics; considering that noise points appear randomly, the data density of noise points is quite different from the data density of gear data points, so the more similar the data densities are, the smaller the difference is, and the less likely the target data point is to be a noise point. In order to limit the analysis scope, the target data point is compared with all data points within the preset neighborhood, and the density uniformity parameter of the target data point is obtained based on the similarity characteristics of the data density of the target data point and the data density of all data points within the preset neighborhood of the target data point; the similarity characteristics of the data density are positively correlated with the density uniformity parameter.

[0082] As an example, the second preset neighborhood of the target data point is an area formed by a sphere with a radius of 3, the center of the sphere is the target data point, and the unit length is 1mm; the preset neighborhood range of the target data point is the same as the second preset neighborhood of the target data point; the data density of all data points within the preset neighborhood range of the target data point is represented by the mean, and the average value of the data density of all data points within the preset neighborhood range of the target data point is taken as the comparison data density, which represents the data density of all data points within the preset neighborhood range of the target data point; the absolute value of the difference between the data density of the target data point and the comparison data density is calculated using the negative correlation function After negative correlation mapping, it is used as the density uniformity parameter of the target data point; the smaller the absolute value of the difference between the data density of the target data point and the comparison data density, the more similar the data density of the target data point is to all data points within the preset neighborhood range, which means that the target data point is less likely to be a noise point, and the larger the density uniformity parameter is.

[0083] In other embodiments of the present invention, the implementer may also set other second preset neighborhoods and preset neighborhood ranges, and the second preset neighborhoods and the preset neighborhood ranges may also be different.

[0084] Step S3: The gear probability and density uniformity parameters of the target data point are integrated to obtain the gear membership of the target data point; both the gear probability and the density uniformity parameters are positively correlated with the gear membership; the distance between the target data point and the center point of the initial density cluster to which it belongs is adjusted according to the gear membership, and secondary density clustering is performed; and denoised point cloud data is obtained based on the secondary density clustering results.

[0085] By analyzing the gear possibility and density uniformity parameters of the target data points, the possibility of the target data points being gear point cloud data is determined from the distribution similarity angle, gear structure symmetry angle and density similarity angle respectively. Therefore, the gear possibility and density uniformity parameters of the target data points are further integrated, and the gear membership of the target data points is comprehensively obtained by combining multiple analysis angles, providing a reliable basis for subsequent optimized clustering and accurate denoising of point cloud data; both the gear possibility and density uniformity parameters are positively correlated with the gear membership.

[0086] Preferably, in one embodiment of the present invention, since the gear probability and the density uniformity parameter are both in [0, 1], the product of the gear probability and the density uniformity parameter of the target data point is directly used as the gear membership of the target data point, and the gear membership is also distributed in [0, 1].

[0087] In other embodiments of the present invention, if the gear probability or density uniformity parameter is not in [0, 1], the product of the gear probability and the density uniformity parameter of the target data point can be linearly normalized and used as the gear membership.

[0088] Considering that the greater the gear membership, the more likely the data point is to belong to the gear area, and the smaller the gear membership, the more likely it is to be a noise point, the distance between the target data point and the center point of the preliminary density clustering cluster to which it belongs is adjusted according to the gear membership, and secondary density clustering is performed. This can better reflect the actual local density changes, improve the accuracy of the clustering results, better distinguish between noise points and real gear data points, and ultimately improve the accuracy of dimensional error detection.

[0089] Preferably, in one embodiment of the present invention, considering that the larger the gear membership, the more the target data point belongs to the gear region, the more it needs to be clustered into the cluster, and the closer the distance from the cluster center point needs to be; conversely, the smaller the gear membership, the further away from the cluster center point it needs to be; based on this, the gear membership of the target data point is negatively correlated and normalized to obtain the distance correction coefficient of the target data point;

[0090] The product of the distance between the target data point and the center point of the preliminary density clustering cluster to which it belongs and the distance correction coefficient is used as the corrected distance between the target data point and the center point of the preliminary density clustering cluster to which it belongs, and secondary density clustering is performed based on the corrected distance.

[0091] The calculation formula for the corrected distance includes: ;

[0092] in, Indicates the Corrected distance of target data points; Indicates the Gear membership of target data points; Indicates the The distance correction factor of each target data point; Indicates the The distance between a target data point and the center point of the preliminary density clustering cluster to which it belongs.

[0093] In the calculation formula of the corrected distance, the gear membership is negatively mapped and normalized in the form of a constant 1 minus the gear membership; the distance between the target data point and the center point of the preliminary density clustering cluster to which it belongs is the Euclidean distance; the distance between the target data point and the center point of the preliminary density clustering cluster to which it belongs is reduced to different degrees. The larger the gear membership, the greater the reduction, and the smaller the gear membership, the smaller the reduction. As a result, the data points with smaller gear membership are farther away from the cluster center, ensuring that the cluster center can reflect the local characteristics of the data point to the greatest extent, and ultimately improving the clustering effect of the entire point cloud data.

[0094] It should be noted that the secondary density clustering also uses the DBSCAN clustering algorithm, and k is also set to 4, which is the same as the preliminary density clustering process and will not be described in detail here.

[0095] In other embodiments of the present invention, the distance correction coefficient may be mapped to [0, 2] before performing distance correction. The implementer may also use other density clustering methods such as the OPTICS clustering algorithm for density clustering. These are all technical means well known to those skilled in the art and will not be described in detail here.

[0096] After optimized secondary density clustering based on the various distribution characteristics of the data, denoised point cloud data can be obtained according to the secondary density clustering results, providing reliable data support for dimensional error detection.

[0097] In one embodiment of the present invention, outliers in the secondary density clustering results are removed, and the clusters are retained as denoised point cloud data.

[0098] Step S4: Dimensional error detection of the gear is performed based on the denoised point cloud data.

[0099] The denoised point cloud data retains the effective geometric information of the gear surface, thereby better capturing the actual geometric shape of the gear. Finally, the dimensional error of the gear is detected based on the denoised point cloud data.

[0100] Preferably, in one embodiment of the present invention, the denoised point cloud data is converted into a gear surface using Poisson surface reconstruction. The reconstructed gear surface is then converted into a mesh model using a triangulated meshing method. Measurements are then made of at least the gear's diameter, number of teeth, tooth profile, and pitch. The gear's dimensional error is then determined by comparing the measured data with the design specification data. These techniques are well known to those skilled in the art and will not be elaborated upon here.

[0101] In one embodiment of the present invention, after obtaining the dimensional error of the gear, it also includes storing the detection data, and marking and eliminating the gears whose dimensional error exceeds the design specification. The design specification of the gear is related to the specific production process and can be obtained through the production manual or product design drawings, which will not be elaborated here.

[0102] One embodiment of the present invention also provides a hardware dimension error detection system based on point cloud processing, which includes a memory, a processor and a computer program, wherein the memory is used to store the corresponding computer program, and the processor is used to run the corresponding computer program. When the computer program runs in the processor, it can implement the hardware dimension error detection method based on point cloud processing described in steps S1-S4.

[0103] In summary, in order to solve the technical problem of inaccurate denoising of gear point cloud data, which affects the gear dimensional error detection, the present invention first obtains the gear point cloud data and performs preliminary density clustering on the point cloud data; further analyzes from multiple angles the distribution similarity characteristics of the target data point and other data points in the corresponding preliminary density clustering cluster, the repeated characteristics of the curvature, and the similar characteristics of the data density of the target data point to obtain the gear membership of the target data point; further adjusts the distance between the target data point and the center point in the corresponding preliminary density clustering cluster according to the gear membership, performs secondary density clustering and obtains denoised point cloud data; finally, performs dimensional error detection on the gear based on the denoised point cloud data. The present invention analyzes the possibility of data points belonging to the gear area from multiple angles, optimizes the density clustering process, accurately removes noise interference in the point cloud data, obtains reliable gear processing product dimensional error detection results, ensures the compliance of gear products, and ensures the reliability of gears in applications.

[0104] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0105] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A hardware dimension error detection method based on point cloud processing, characterized in that: The method comprises: Acquire point cloud data of the hardware; perform preliminary density clustering on the point cloud data; select any data point in any preliminary density clustering cluster as a target data point; The hardware performance degree of the target data point is obtained based on the uniform distribution characteristics of the data points in the local neighborhood of the target data point; the hardware possibility of the target data point is obtained based on the similarity characteristics of the hardware performance degrees of the target data point and other data points in the preliminary density clustering cluster to which it belongs, combined with the repetitive characteristics of the curvature of the target data point and other data points in the point cloud data; the density uniformity parameter of the target data point is obtained based on the similarity characteristics of the data density of the target data point and the data density of all data points within the preset neighborhood range of the target data point; fusing the hardware probability and the density uniformity parameter of the target data point to obtain the hardware membership of the target data point; the hardware probability and the density uniformity parameter are both positively correlated with the hardware membership; adjusting the distance between the target data point and the center point of the initial density cluster to which it belongs based on the hardware membership to perform secondary density clustering; and obtaining denoised point cloud data based on the secondary density clustering result; Performing dimensional error detection on the hardware based on the denoised point cloud data; The method for obtaining the hardware performance level includes: A surface formed by the data points in the intersection of the first preset neighborhood of the target data point and the preliminary density clustering cluster to which it belongs is used as a target surface for the target data point; and a hardware performance level of the target data point is obtained based on the uniform distribution characteristics of the data points in the target surface; The method for obtaining the hardware possibility includes: Obtaining a first possibility parameter of the target data point based on similarity characteristics of the hardware performance between the target data point and other data points in the preliminary density clustering cluster to which it belongs; the similarity characteristics of the hardware performance are positively correlated with the first possibility parameter; In the point cloud data, a second possibility parameter of the target data point is obtained according to the number of other data points having the same curvature as the target data point; the number of other data points having the same curvature is positively correlated with the second possibility parameter; Obtaining the hardware likelihood of the target data point according to the first likelihood parameter and the second likelihood parameter of the target data point; the first likelihood parameter and the second likelihood parameter are both positively correlated with the hardware likelihood; The method for obtaining the hardware membership includes: The product of the gear likelihood and density uniformity parameter of the target data point is linearly normalized and used as the gear membership.

2. The method for detecting hardware size errors based on point cloud processing according to claim 1, characterized in that: The method for obtaining the hardware performance level of the target data point according to the uniform distribution characteristics of the data points within the target curved surface includes: Within the target surface, with the target data point as the origin, data points to be analyzed are selected in a preset direction away from the origin; based on the consistency of distances between all adjacent data points to be analyzed in the same preset direction, the hardware performance level of the target data point is obtained; the distance consistency and the hardware performance level are positively correlated.

3. The hardware dimension error detection method based on point cloud processing according to claim 2 is characterized in that: The method for obtaining the curvature includes: The maximum curvature of the target data point in all the preset directions is taken as the curvature of the target data point.

4. The method for detecting hardware size errors based on point cloud processing according to claim 1, characterized in that: The method for obtaining the density uniformity parameter includes: Obtaining a data density of each data point based on a quantity characteristic of data points within a second preset neighborhood of each data point; The density uniformity parameter of the target data point is obtained based on similar features between the data density of the target data point and the data density of all data points within a preset neighborhood of the target data point; the similar features of the data density are positively correlated with the density uniformity parameter.

5. The method for detecting hardware size errors based on point cloud processing according to claim 1, characterized in that: The method of performing secondary density clustering by adjusting the distance between the target data point and the center point of the primary density clustering cluster according to the hardware membership includes: Performing negative correlation mapping and normalizing on the hardware membership of the target data point to obtain a distance correction coefficient of the target data point; The product of the distance between the target data point and the center point of the preliminary density clustering cluster to which it belongs and the distance correction coefficient is used as the corrected distance between the target data point and the center point of the preliminary density clustering cluster to which it belongs, and secondary density clustering is performed according to the corrected distance.

6. The method for detecting hardware size errors based on point cloud processing according to claim 1, characterized in that: The method for detecting hardware size errors based on the denoised point cloud data includes: The denoised point cloud data is converted into a hardware surface using Poisson surface reconstruction. The reconstructed hardware surface is converted into a mesh model using a triangulated meshing method. The measurement data includes at least the diameter, number of teeth, tooth shape and tooth pitch of the hardware. The dimensional error of the hardware is obtained by comparing the measured data with the design specification data.

7. The method for detecting hardware size errors based on point cloud processing according to claim 1, characterized in that: The DBSCAN clustering algorithm is used for both the preliminary density clustering and the secondary density clustering.

8. A hardware dimension error detection system based on point cloud processing, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the hardware dimension error detection method based on point cloud processing as described in any one of claims 1 to 7 are implemented.

Images