Automobile panel profile detection method and system based on three-dimensional point cloud registration

Abstract

The present application belongs to the field of computer vision, and particularly relates to a kind of automobile covering contour detection method and system based on three-dimensional point cloud registration.The method of the present application comprises: obtaining the three-dimensional point cloud of the workpiece to be measured, and the CAD model of the workpiece to be measured;By matching the preset reference geometric feature in the CAD model with the corresponding point set in the three-dimensional point cloud, initial alignment is performed on the three-dimensional point cloud;Determine the initial weight map and perform iterative registration on the three-dimensional point cloud after initial alignment to obtain the preliminary registration result and the preliminary deviation data;Determine the high deviation area of the deviation value in the three-dimensional point cloud exceeding the preset tolerance threshold, and determine the adaptive weight map according to the position and deviation amplitude of the high deviation area;According to the adaptive weight map, the three-dimensional point cloud after initial alignment is executed to correct registration;According to the target registration result, the contour detection report of the workpiece to be measured is generated.The present application effectively improves the accuracy of contour detection.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of computer vision, and particularly relates to a three-dimensional point cloud registration-based automobile panel contour detection method and system. BACKGROUND

[0002] Automobile panels, such as doors, hoods, fenders, etc., have the characteristics of complex free-form surface shape and high dimensional accuracy requirements. In the production process, accurate contour detection of these panels is a key link to ensure the assembly quality and appearance quality of the whole vehicle.

[0003] Traditionally, contour detection is mostly carried out by contact measurement using a three-coordinate measuring machine (CMM). This method has high accuracy, but the measurement speed is slow, the sampling points are sparse, and it is difficult to realize global detection of the entire surface, which cannot meet the needs of modern manufacturing for high efficiency and full-size detection.

[0004] With the development of optical three-dimensional scanning technology, non-contact measurement methods based on structured light or laser scanning have been widely applied. This method can quickly obtain high-density three-dimensional point cloud data of the workpiece surface, providing a data basis for global contour analysis. In the detection process, the three-dimensional point cloud of the measured workpiece collected is registered with the standard CAD (Computer Aided Design) model, and the accuracy of the registration directly determines the accuracy of the subsequent contour deviation calculation.

[0005] However, existing point cloud registration algorithms, such as the Iterative Closest Point (ICP) algorithm and its variants, still face challenges when applied to complex surfaces such as automobile panels. On the one hand, the standard ICP algorithm treats all point cloud data equally and does not consider the influence of the geometric characteristics of different regions of the surface on the stability of the registration. For example, the point cloud noise in high-curvature regions is large and easily interferes with the registration process. On the other hand, when the workpiece has local manufacturing deviations that exceed the tolerance range, the traditional registration algorithm tends to average the overall error, which may lead to inaccurate evaluation of the true appearance of the high-deviation region, and even affect the registration accuracy of other qualified regions, ultimately reducing the reliability and accuracy of the detection.

[0006] Therefore, how to improve the accuracy of point cloud registration algorithms when dealing with complex surfaces and local deviations has become a technical problem to be solved in the field of automatic detection of automobile panels. SUMMARY

[0007] In order to solve the above-mentioned problems in the prior art, i.e., the conflict between multiple model prediction results leading to reduced early warning accuracy, the present application proposes, in a first aspect, a three-dimensional point cloud registration-based automobile panel contour detection method, comprising:

[0008] acquire a three-dimensional point cloud of a workpiece to be measured and a CAD model of the workpiece to be measured;

[0009] perform initial alignment on the three-dimensional point cloud by matching a preset reference geometric feature in the CAD model with a corresponding point set in the three-dimensional point cloud, so that the three-dimensional point cloud and the CAD model are aligned in the same coordinate system;

[0010] determine an initial weight map according to the curvatures of each region of the surface of the CAD model, and perform iterative registration on the three-dimensional point cloud after initial alignment by applying the initial weight map to obtain a preliminary registration result and preliminary deviation data;

[0011] determine a high-deviation region in the three-dimensional point cloud where the deviation value exceeds a preset tolerance threshold according to the preliminary deviation data, and determine an adaptive weight map according to the position and deviation amplitude of the high-deviation region, the adaptive weight map being used to assign a higher weight value to the high-deviation region than the initial weight map;

[0012] perform modified registration on the three-dimensional point cloud after initial alignment according to the adaptive weight map to obtain a target registration result;

[0013] determine the profile deviation between the three-dimensional point cloud and the CAD model according to the target registration result, and generate a profile detection report of the workpiece to be measured.

[0014] In some preferred embodiments, the initial alignment on the three-dimensional point cloud by matching a preset reference geometric feature in the CAD model with a corresponding point set in the three-dimensional point cloud comprises:

[0015] identify and extract an RPS reference defined on the CAD model, the RPS reference including at least one reference main hole, one reference long circular groove, and one reference surface;

[0016] search and segment a point set corresponding to the RPS reference in topological structure and size in the three-dimensional point cloud by a preset geometric feature recognition algorithm;

[0017] determine and apply a rigid transformation matrix to transform the coordinate system of the three-dimensional point cloud to a reference coordinate system defined by the RPS reference to align the three-dimensional point cloud by using a least squares optimal fitting algorithm.

[0018] In some preferred embodiments, the determination of the initial weight map according to the curvatures of each region of the surface of the CAD model comprises:

[0019] discretely sample the triangular mesh surface of the CAD model to determine the Gaussian curvature value of each sampling point or each triangular facet;

[0020] mapping a Gaussian curvature value of each sampling point or each triangular patch to an initial weight value between 0 and 1 by using a weight assignment function to form the initial weight map, wherein the initial weight value is inversely related to the Gaussian curvature value, and a region with a lower curvature value is assigned a higher initial weight value, and a region with a higher curvature value is assigned a lower initial weight value.

[0021] In some preferred embodiments, the determining the high deviation region with a deviation value exceeding a preset tolerance threshold in the three-dimensional point cloud according to the preliminary deviation data comprises:

[0022] According to the preliminary deviation data, marking points with a deviation value greater than the preset tolerance threshold in the three-dimensional point cloud as candidate points;

[0023] Based on the three-dimensional spatial coordinates of all candidate points, using a preset spatial clustering algorithm, by setting a neighborhood radius and a minimum point number parameter, spatially adjacent candidate points are aggregated into one or more high deviation region clusters to form the high deviation region.

[0024] In some preferred embodiments, for each point in the high deviation region, its adaptive weight value in the adaptive weight map is calculated by a gain function based on its weight value in the initial weight map and its deviation amplitude;

[0025] The gain function is a positively related nonlinear function, the nonlinear function takes the excess amount of each point as input and outputs a corresponding weight gain factor, the adaptive weight value of each point is equal to the sum of its initial weight and the gain factor, the growth rate of the gain factor increases with the excess amount, and the excess amount is the difference between the deviation absolute value and the tolerance threshold.

[0026] In some preferred embodiments, the gain function is specifically defined as an n-th order polynomial function or an exponential function with the excess amount as the independent variable to ensure that the growth rate of the weight gain factor accelerates with the increase of the excess amount.

[0027] In some preferred embodiments, the applying the initial weight map to perform iterative registration on the initial aligned three-dimensional point cloud comprises:

[0028] For each point in the three-dimensional point cloud, finding its nearest neighbor point on the surface of the CAD model, and taking the normal vector of the triangular patch where the nearest neighbor point is located as its corresponding normal vector;

[0029] Constructing an objective function for minimizing the weighted sum of squares of distances between each point and its corresponding tangent plane on the CAD model, wherein the weight of each point is determined by the initial weight map;

[0030] determining a rigid transformation matrix that minimizes the weighted sum of squares by iteratively solving the objective function, and applying the rigid transformation matrix to the three-dimensional point cloud until the registration error converges.

[0031] In some preferred embodiments, the profile detection report comprises:

[0032] a three-dimensional color map of profile deviations rendered on the CAD model, the three-dimensional color map being used to indicate the deviation distribution across the surface of the workpiece under test;

[0033] a statistical list of profile indicators including maximum positive deviation, maximum negative deviation, average deviation, and root mean square error; and

[0034] a location, area, and average amount of over-tolerance of each high deviation region.

[0035] The present application proposes in a second aspect a profile detection system for automobile coverings based on three-dimensional point cloud registration, comprising:

[0036] a data acquisition module configured to acquire a three-dimensional point cloud of a workpiece under test and a CAD model of the workpiece under test;

[0037] an initial alignment module configured to perform initial alignment on the three-dimensional point cloud by matching a preset reference geometric feature in the CAD model with a corresponding point set in the three-dimensional point cloud, so that the three-dimensional point cloud and the CAD model are aligned in the same coordinate system;

[0038] a preliminary registration module configured to determine an initial weight map according to the curvature of each region of the surface of the CAD model, and perform iterative registration on the three-dimensional point cloud after initial alignment using the initial weight map to obtain a preliminary registration result and preliminary deviation data;

[0039] a weight calculation module configured to determine a high deviation region in the three-dimensional point cloud where the deviation value exceeds a preset tolerance threshold according to the preliminary deviation data, and determine an adaptive weight map according to the location and deviation amplitude of the high deviation region, the high deviation region of the adaptive weight map being assigned a higher weight value than the initial weight map;

[0040] a correction registration module configured to perform correction registration on the three-dimensional point cloud after initial alignment according to the adaptive weight map to obtain a target registration result;

[0041] a detection output module configured to determine profile deviations of the three-dimensional point cloud and the CAD model according to the target registration result, and generate a profile detection report of the workpiece under test.

[0042] The present application has the following beneficial effects:

[0043] Compared with the prior art, the automobile panel profile detection method based on three-dimensional point cloud registration provided by the application reduces the interference of high curvature areas (usually with larger noise and unstable features) on the registration result in the iterative registration stage by introducing an initial weight map based on the curvature of the CAD model, and preferentially uses flat and feature stable areas for alignment, thereby improving the stability of the preliminary registration and the robustness to the initial position.

[0044] Meanwhile, the application identifies high deviation areas with deviation out of tolerance after obtaining the preliminary registration result through a two-step registration strategy, and generates an adaptive weight map; in the correction registration stage, by significantly increasing the weight of the high deviation area, the registration algorithm focuses on the alignment of these areas, avoiding the problem of averaging local large deviations in traditional algorithms, so that the local topographic defects of the workpiece can be more truly reflected, and the accuracy of the profile detection is significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0045] Other features, objects and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments made with reference to the accompanying drawings:

[0046] Figure 1 is a flowchart of an automobile panel profile detection method based on three-dimensional point cloud registration according to an embodiment of the application;

[0047] Figure 2 is a framework diagram of an automobile panel profile detection system based on three-dimensional point cloud registration according to an embodiment of the application;

[0048] Figure 3 is a structural diagram of a computer system according to an embodiment of the application. DETAILED DESCRIPTION

[0049] The application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to explain the related application, and not to limit the application. In addition, it should be noted that only parts related to the application are shown in the drawings for ease of description.

[0050] It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other without conflict. The application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0051] Please refer to Figure 1 The embodiment of the application provides an automobile panel profile detection method based on three-dimensional point cloud registration, comprising:

[0052] Step S10, obtaining a three-dimensional point cloud of a workpiece to be measured and a CAD model of the workpiece to be measured;

[0053] Step S20, performing initial alignment on the three-dimensional point cloud by matching the preset reference geometric features in the CAD model with the corresponding point set in the three-dimensional point cloud, so that the three-dimensional point cloud and the CAD model are aligned in the same coordinate system;

[0054] Step S30, determining an initial weight map according to the curvature of each region of the CAD model surface, and applying the initial weight map to perform iterative registration on the three-dimensional point cloud after initial alignment to obtain a preliminary registration result and preliminary deviation data;

[0055] Step S40, determining a high deviation region in the three-dimensional point cloud according to the preliminary deviation data, wherein the deviation value of the high deviation region exceeds a preset tolerance threshold, and determining an adaptive weight map according to the position and deviation amplitude of the high deviation region, wherein the adaptive weight map is used to assign a higher weight value to the high deviation region compared with the initial weight map;

[0056] Step S50, performing correction registration on the three-dimensional point cloud after initial alignment according to the adaptive weight map to obtain a target registration result;

[0057] Step S60, determining the profile deviation of the three-dimensional point cloud and the CAD model according to the target registration result, and generating a profile detection report of the workpiece to be measured.

[0058] The embodiment aims to solve the problem that the traditional registration method is prone to registration deviation when facing a workpiece containing manufacturing deviation or deformation (especially local large deviation), thereby causing distortion of the deviation detection result. The embodiment strengthens the influence of the local deviation region in the registration process through an adaptive weighting strategy, thereby obtaining a more accurate and more realistic alignment result.

[0059] In a specific application scenario, the workpiece to be measured can be a freshly produced automobile door outer panel, which can have local depressions or protrusions caused by stamping springback or improper transportation.

[0060] Specifically, the specific process of the embodiment method is as follows:

[0061] First, data acquisition. A high-precision non-contact measurement device, such as a structured light scanner, is used to scan the automobile door outer panel to be measured to obtain three-dimensional point cloud data of its surface. The point cloud data is usually in the order of magnitude of millions, with high density and accuracy (for example, average point distance 0.1 mm, accuracy better than 0.05 mm).

[0062] At the same time, the standard CAD model of the door outer panel is called from the product design library. The model is usually a standard three-dimensional model designed by UG NX or CATIA software.

[0063] Next, the collected three-dimensional point cloud data and the CAD model are initially aligned. Since the point cloud coordinate system obtained by the scanner is inconsistent with the design coordinate system of the CAD model, the embodiment first performs initial alignment so as to facilitate subsequent precise registration.

[0064] Specifically, the initial alignment can be performed based on a preset reference point system (RPS). The RPS reference is predefined on the CAD model of the vehicle door outer panel, for example, a 10mm round hole as a main positioning, a 20mm*10mm long circular groove as a rotation constraint, and a planar point as a pose constraint. During the initial alignment, the point cloud is first roughly placed near the CAD model by manual or semi-automatic means, and then an algorithm such as RANSAC (random sample consensus) is used to segment the geometric features (circle, long circular groove, plane) corresponding to the RPS reference from the point cloud. The geometric center or fitting parameters of these segmented point cloud features are calculated, and are aligned with the theoretical parameters of the corresponding RPS reference on the CAD model.

[0065] Then, an initial weight map is generated based on the surface curvature information of the CAD model. After the initial alignment, the position and pose of the point cloud and the CAD model are basically consistent, but there is still a slight deviation. In order to focus more on the shape stable area in the subsequent precise registration, the embodiment assigns a weight to each point on the CAD model.

[0066] Specifically, the CAD model is first subjected to a triangular meshing process. Then, the local curvature of each mesh vertex on the model surface is calculated, for example, the Gaussian curvature, which reflects the degree of bending of the surface. For the vehicle door outer panel, the curvature of the large-area flat area is close to zero, while the areas with distinct features such as the waist line of the vehicle body and the edges of the vehicle window have a larger curvature. The principle of weight distribution is inversely proportional to the absolute value of the curvature.

[0067] For example, the calculation formula of the weight w is:

[0068] w = 1 / (1 + a * |K|);

[0069] where K is the Gaussian curvature, and a is a positive coefficient (for example, a = 10). The vertices of the flat area (such as the large surface of the vehicle door) will obtain a higher weight (close to 1.0), while the vertices of the feature line or corner area with high curvature will have a lower weight (for example, 0.1-0.3). Further, during the preliminary registration, the algorithm is more dependent on the stable and less likely to deform flat surface, avoiding being excessively affected by the slight deviation of the high curvature area, so as to obtain a robust global optimal alignment.

[0070] Subsequently, a preliminary registration of the point cloud and the CAD model is performed using the generated initial weight map. A weighted Iterative Closest Point (Weighted ICP) algorithm is employed in this step. In each iteration, for each point in the point cloud, its closest point on the CAD model is found. Then, based on these pairs of points and considering the weights of the corresponding points on the CAD model, a rigid transformation (rotation and translation) is computed that minimizes the weighted sum of squared errors. This transformation is applied to the entire point cloud, and one iteration is completed. This process is repeated until a convergence condition is met, for example, the number of iterations reaches an upper limit (e.g., 100), or the difference in the norm of the pose transformation matrix between two iterations is less than a pre-set threshold (e.g., 1e-6). After the preliminary registration is completed, the point cloud and the CAD model achieve a good fit globally.

[0071] After the preliminary registration is completed, high deviation regions with large manufacturing deviations are identified. By calculating the normal distance from each point in the point cloud to the surface of the CAD model after preliminary registration, a deviation value is obtained. This deviation value is compared with a pre-set tolerance threshold, for example, the manufacturing tolerance of an automobile door is usually set to ±0.5mm. All points with deviation absolute values exceeding 0.5mm are identified as out-of-tolerance candidate points. In order to avoid misjudging isolated noise points as deviation regions, spatial clustering analysis needs to be performed on these out-of-tolerance candidate points. For example, the DBSCAN (Density-Based Spatial Clustering) algorithm can be used, with a suitable neighborhood radius ε (e.g., 5mm) and a minimum number of points MinPts (e.g., 20), to aggregate spatially adjacent out-of-tolerance candidate points into one or more continuous high deviation regions.

[0072] Next, for the identified high deviation regions, an adaptive weight map is generated, i.e., the weights of the high deviation regions are significantly increased, so that in the subsequent correction registration, an alignment pose that better explains these deviations is tried to be found.

[0073] Specifically, the embodiment updates the initial weight map through a gain function. For the vertices of the CAD model located in the high deviation region, the new weight W_new can be obtained by multiplying the original weight W_old by a gain factor. The gain factor is positively related to the size of the deviation, i.e., the larger the deviation, the more the weight is increased. For example, the gain factor can be a quadratic function of the deviation value d:

[0074] Gain factor = 1 + β * d^2;

[0075] where β is the gain coefficient (e.g., β = 0.5). For example, the weight gain of a region with a deviation of 1mm will be much larger than that of a region with a deviation of 0.6mm. In this way, the embodiment obtains an adaptive weight map that is sensitive to local large deviation regions.

[0076] Then, the adaptive weight map is used to perform a revised registration of the point cloud and the CAD model. Specifically, the same Weighted Iterative Closest Point (Weighted ICP) algorithm as in the preliminary registration is repeatedly run in this process, but this time using the adaptive weight map described above.

[0077] It should be noted that, since the weights of the high deviation areas are significantly increased, the revised registration process will pay more attention to these areas, forcing the algorithm to focus on optimizing the point pair distances in these areas when calculating the pose transformation. Ultimately, the algorithm will converge to a new registration pose that more accurately reflects the actual deformation state of the workpiece, rather than averaging the local deviations to accommodate the global minimum error.

[0078] Finally, based on the results of the revised registration, the embodiment generates a deviation detection report. Under the final determined registration pose, the normal distances of all points in the point cloud to the surface of the CAD model are recalculated to obtain the final deviation distribution. To visually display the detection results, these deviation values are visualized as a three-dimensional color map and superimposed on the model. For example, green can be set to represent the qualified area within the ±0.2mm tolerance band, yellow and orange to represent 0.2mm to 0.5mm positive deviation, blue and purple to represent -0.2mm to -0.5mm negative deviation, and red to prominently identify the serious out-of-tolerance area exceeding ±0.5mm. This graphically illustrated detection report can clearly show the size quality state of the workpiece under test to the quality inspection engineer, accurately locating the deviation areas and their severity.

[0079] At this point, through the above steps, the embodiment successfully achieves accurate and reliable adaptive registration and deviation detection of a three-dimensional point cloud and a CAD model.

[0080] More specifically, in the above embodiment, after the preliminary registration is completed, the shortest normal distance of each point in the three-dimensional point cloud to the surface of the CAD model, i.e., the deviation value, is calculated.

[0081] In a preferred embodiment, an engineering-acceptable tolerance threshold T is pre-set, for example, for a car door outer panel, T can be set to 0.5mm. All points with deviation absolute value greater than T are marked as out-of-tolerance candidate points.

[0082] In a preferred embodiment, in order to eliminate isolated pseudo-out-of-tolerance points caused by measurement noise or point cloud sparsity, clustering analysis needs to be performed on the out-of-tolerance candidate points. The embodiment preferably uses the DBSCAN (Density-Based Spatial Clustering with Noise) algorithm. The DBSCAN algorithm does not need to pre-specify the number of clusters and can discover clusters of arbitrary shape, making it very suitable for this scenario. Its main parameters are:

[0083] Neighborhood radius: define the neighborhood of a point, for example, set to 5mm;

[0084] Minimum points: the minimum number of points a core object must contain in its neighborhood, for example, set to 20;

[0085] By running the DBSCAN algorithm, spatially densely connected outlier candidate points will be aggregated into one or more high deviation regions (clusters), while sparse, isolated outlier candidate points are identified as noise and ignored, thus identifying the real manufacturing deviation or deformation region.

[0086] In a preferred embodiment, the core of the adaptive weight map is to enhance the weight of the identified high deviation region, and the enhancement amplitude is positively correlated with the deviation size.

[0087] Specifically, all vertices on the triangular mesh of the CAD model are traversed. For each vertex, it is determined whether it is located in the high deviation region identified in the previous step.

[0088] If the vertex is not in the high deviation region, its weight remains the value in the initial weight map;

[0089] If the vertex is in the high deviation region, the weight of the vertex needs to be dynamically adjusted.

[0090] First, calculate the average deviation or maximum deviation of the high deviation region, and design the gain function G(d), which is a monotonically increasing function of the deviation d. Preferably, the function can be nonlinear to amplify the influence of larger deviations. For example, a quadratic function or an exponential function can be used:

[0091] Quadratic gain function: G(d) = 1 + β * d^2;

[0092] Exponential gain function: G(d) = exp(γ * d);

[0093] Where β and γ are positive gain coefficients.

[0094] The new weight W_new of the vertex in the region is calculated as follows:

[0095] W_new = W_old * G(d_avg);

[0096] Where W_old is the initial weight of the vertex, and G(d_avg) is the gain factor calculated according to the average deviation of the region where the vertex is located, and its value is obtained by calculating the gain function G(d) selected for the region where the vertex is located. Based on this, the larger the deviation of the region, the higher the amplification factor of its weight, thus having a decisive influence on the subsequent correction registration.

[0097] More specifically, in both the initial registration and the modified registration, the present embodiment employs a weighted Iterative Closest Point algorithm. While the standard ICP algorithm aims to minimize the sum of squared distances between points in the point cloud and the model, the weighted ICP algorithm aims to minimize the weighted sum of squared distances.

[0098] The objective function of which is:

[0099] argmin(R, t)=Σ [ w_i * || (R*p_i + t) - q_i ||^2 ];

[0100] where p_i is a point in the point cloud, q_i is its corresponding closest point on the model, R and t are the rotation matrix and translation vector to be solved, and w_i is the weight assigned to q_i.

[0101] It is easy to understand that the initial weight map generated by the above embodiment is used in the initial registration, while the adaptive weight map generated by the above embodiment is used in the modified registration.

[0102] Further, referring to Figure 2 the second embodiment of the present application proposes a system for detecting the profile of an automobile panel based on three-dimensional point cloud registration, comprising:

[0103] a data acquisition module 210, configured to acquire a three-dimensional point cloud of a workpiece to be measured and a CAD model of the workpiece to be measured;

[0104] an initial alignment module 220, configured to perform initial alignment on the three-dimensional point cloud by matching a preset reference geometric feature in the CAD model with a corresponding point set in the three-dimensional point cloud, so that the three-dimensional point cloud and the CAD model are aligned in the same coordinate system;

[0105] an initial registration module 230, configured to determine an initial weight map according to the curvature of each region of the surface of the CAD model, and apply the initial weight map to perform iterative registration on the three-dimensional point cloud after initial alignment, to obtain an initial registration result and initial deviation data;

[0106] a weight calculation module 240, configured to determine a high-deviation region in the three-dimensional point cloud where the deviation value exceeds a preset tolerance threshold according to the initial deviation data, and determine an adaptive weight map according to the position and deviation amplitude of the high-deviation region, the high-deviation region of the adaptive weight map being assigned a higher weight value than the initial weight map;

[0107] a modified registration module 250, configured to perform modified registration on the three-dimensional point cloud after initial alignment according to the adaptive weight map, to obtain a target registration result;

[0108] The detection output module 260 is used to determine the contour deviation between the three-dimensional point cloud and the CAD model based on the target registration result, and generate a contour detection report of the workpiece to be tested.

[0109] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the system described above can be found in the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0110] The following is for reference. Figure 3 It shows a schematic diagram of the structure of a computer system suitable for implementing the system and method embodiments of the present invention. Figure 3 The server shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0111] like Figure 3 As shown, the computer system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes based on programs stored in the read-only memory (ROM) 302 or programs loaded into the random access memory (RAM) 303 from the storage section 308. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0112] The following components are connected to the input / output I / O interface 305: an input section 306 including a keyboard, mouse, etc.; an output section 307 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the input / output I / O interface 305 as needed. A removable medium 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 310 as needed so that computer programs read from it can be installed into the storage section 308 as needed.

[0113] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program in accordance with embodiments of the present application. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for performing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 309, and / or installed from the removable media 311. When the computer program is executed by the central processing unit CPU 301, the above-described functions defined in the methods of the present application are performed. Note that the above-described computer readable medium of the present application can be either a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable storage medium may, for example, be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination of the foregoing.

[0114] More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, a computer readable storage medium can be any tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus or device. In the present application, a computer readable signal medium can include a computer readable program code carried by a data signal in a baseband or as part of a carrier wave. Such a propagated signal can take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium can be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable storage medium can be transmitted using any appropriate medium, including, but not limited to, wireless, wire line, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0115] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0116] The flow diagrams and the block diagrams in the drawings are illustrations of architectures, functionalities, and operations of possible implementations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams or block diagrams can represent a module, a segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flow diagrams, and combinations thereof, can be implemented by special purpose hardware-based systems that perform the specified functions or operations, or combinations of special purpose hardware and computer instructions.

[0117] The terms "first", "second", etc. are used to distinguish similar objects, not to describe or indicate a particular order or sequence.

[0118] The term "comprising" or any other similar term is intended to encompass the inclusion of one or more elements, steps, or components, but not to the exclusion of any other elements, steps, or components. It is noted that some embodiments can not include all of the elements or steps discussed above.

[0119] Thus far, the technical solutions of the present application have been described in conjunction with the preferred embodiments shown in the drawings.

[0120] The above merely illustrates the embodiments of the present application but should not be taken as limitations. Various changes and modifications can be made by those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application should be included in the scope of claims of the present application.

Claims

1. A method for detecting the profile of an automobile covering based on the registration of three-dimensional point clouds, characterized in that, The method comprises: acquiring a three-dimensional point cloud of a workpiece to be measured and a CAD model of the workpiece to be measured; performing initial alignment on the three-dimensional point cloud by matching a preset reference geometric feature in the CAD model with a corresponding point set in the three-dimensional point cloud, so that the three-dimensional point cloud and the CAD model are aligned in the same coordinate system; determining an initial weight map according to the curvatures of each region on the surface of the CAD model, and performing iterative registration on the three-dimensional point cloud after initial alignment by using the initial weight map to obtain a preliminary registration result and preliminary deviation data; determining a high-deviation region in the three-dimensional point cloud where the deviation value exceeds a preset tolerance threshold according to the preliminary deviation data, and determining an adaptive weight map according to the position and deviation amplitude of the high-deviation region, the adaptive weight map being used to assign a higher weight value to the high-deviation region than the initial weight map; performing correction registration on the three-dimensional point cloud after initial alignment according to the adaptive weight map to obtain a target registration result; determining the profile deviation between the three-dimensional point cloud and the CAD model according to the target registration result, and generating a profile detection report of the workpiece to be measured.

2. The method of claim 1, wherein, The method comprises: identifying and extracting an RPS reference defined on the CAD model, the RPS reference comprising at least one reference main hole, one reference long circular groove and one reference surface; searching for and segmenting a point set corresponding to the RPS reference in topological structure and size in the three-dimensional point cloud by using a preset geometric feature recognition algorithm; determining and applying a rigid transformation matrix to transform the coordinate system of the three-dimensional point cloud to a reference coordinate system defined by the RPS reference to align the three-dimensional point cloud.

3. The method of claim 1, wherein, The method comprises: discretely sampling the triangular mesh surface of the CAD model to determine the Gaussian curvature value of each sampling point or each triangular facet; constructing a weight assignment function to map the Gaussian curvature value of each sampling point or each triangular facet to an initial weight value between 0 and 1 by using the weight assignment function to form the initial weight map, wherein the initial weight value is inversely related to the Gaussian curvature value, and a region with a lower curvature value is assigned a higher initial weight value, and a region with a higher curvature value is assigned a lower initial weight value.

4. The method of claim 1, wherein, The method comprises: marking points in the three-dimensional point cloud with a deviation value greater than a preset tolerance threshold as candidate points according to the preliminary deviation data; based on the three-dimensional space coordinates of all candidate points, using a preset spatial clustering algorithm to aggregate spatially adjacent candidate points into one or more high-deviation region clusters by setting a neighborhood radius and a minimum number of points parameter to form the high-deviation region.

5. The method of claim 1, wherein, For each point in the high deviation region, its adaptive weight value in the adaptive weight map is calculated based on its weight value in the initial weight map and its deviation amplitude by a gain function; The gain function is a positive correlation nonlinear function, the nonlinear function takes the super deviation of each point as input, outputs the corresponding weight gain factor, the adaptive weight value of each point is equal to the sum of its initial weight and the gain factor, the growth rate of the gain factor increases with the super deviation, and the super deviation is the difference between the deviation absolute value and the tolerance threshold.

6. The method of claim 5, wherein, The gain function is specifically defined as an n-th order polynomial function or an exponential function with the super deviation as the independent variable to ensure that the growth rate of the weight gain factor accelerates with the increase of the super deviation.

7. The method of claim 1, wherein, The application of the initial weight map to the initial alignment of the three-dimensional point cloud includes: For each point in the three-dimensional point cloud, find its nearest neighbor on the surface of the CAD model, and take the normal vector of the triangular patch where the nearest neighbor is located as its corresponding normal vector; Construct a target function for minimizing the weighted sum of squares of the distances between each point and its corresponding tangent plane on the CAD model, wherein the weight of each point is determined by the initial weight map; Determine the rigid transformation matrix that minimizes the weighted sum of squares by iteratively solving the target function, and apply the rigid transformation matrix to the three-dimensional point cloud until the registration error converges.

8. The method of claim 1, wherein, The profile detection report includes: A three-dimensional color spectrum map of profile deviation rendered on the CAD model, which is used to indicate the deviation distribution of the entire workpiece surface to be measured; A statistical list of profile indicators, including maximum positive deviation, maximum negative deviation, average deviation, and root mean square error; and The location, area, and average super deviation of each high deviation region.

9. A three-dimensional point cloud registration based automotive panel profile detection system, comprising: The system includes: A data acquisition module for acquiring a three-dimensional point cloud of a workpiece to be measured and a CAD model of the workpiece to be measured; An initial alignment module for performing initial alignment on the three-dimensional point cloud by matching the pre-set reference geometric features in the CAD model with the corresponding point set in the three-dimensional point cloud, so that the three-dimensional point cloud and the CAD model are aligned in the same coordinate system; A preliminary registration module for determining an initial weight map according to the curvature of each region on the surface of the CAD model, and applying the initial weight map to the initial alignment of the three-dimensional point cloud to perform iterative registration and obtain a preliminary registration result and preliminary deviation data; A weight calculation module for determining high deviation regions in the three-dimensional point cloud where the deviation value exceeds a pre-set tolerance threshold according to the preliminary deviation data, and determining an adaptive weight map according to the location and deviation amplitude of the high deviation regions, wherein the high deviation regions of the adaptive weight map are assigned higher weight values than the initial weight map; A correction registration module for performing correction registration on the initial alignment of the three-dimensional point cloud according to the adaptive weight map to obtain a target registration result; The detection output module is configured to determine a profile deviation between the three-dimensional point cloud and the CAD model according to the target registration result, and generate a profile detection report of the workpiece to be detected.

Images