Auxiliary processing quality assessment method, medium and equipment based on three-dimensional point cloud data

Abstract

The present application provides an auxiliary processing quality assessment method, medium and equipment based on three-dimensional point cloud data, which relates to the technical field of workpiece quality inspection. The method first obtains the three-dimensional point cloud corresponding to the workpiece to be inspected, and compares it with the limit model corresponding to the workpiece to be inspected to determine the out-of-tolerance area for key inspection to determine whether the workpiece to be inspected is qualified. In the case that the workpiece to be inspected is qualified, the OBB directional encirclement method is used to determine the target coordinate system corresponding to the target surface of the workpiece to be inspected, and at least two target evaluation indicators of the target surface are combined to establish a multidimensional feature tensor corresponding to the target point cloud, so as to represent the quality characteristics of the target point cloud through a multidimensional data structure. Finally, the multidimensional feature tensor corresponding to the target point cloud is input into the preset quality assessment model to predict the quality level of the target surface, thereby assisting in the comprehensive and accurate assessment of the quality of the workpiece machining surface.

Description

Technical Field

[0001] The present application relates to the technical field of workpiece quality inspection, and in particular to an auxiliary processing quality assessment method, medium and equipment based on three-dimensional point cloud data. Background Art

[0002] Most current methods for inspecting finished workpieces rely on checking multiple local indicators one by one to determine the acceptability of the entire surface. This approach often has the following problems: 1. It is impossible to predict in advance the key inspection areas that may exceed the tolerance, resulting in low inspection efficiency; 2. It is difficult to determine the overall quality of the inspected workpiece by checking multiple local indicators one by one. Summary of the invention

[0003] The present application provides an auxiliary processing quality assessment method, medium and equipment based on three-dimensional point cloud data, which can assist in achieving a comprehensive and accurate assessment of the surface quality of processed parts.

[0004] A first aspect of an embodiment of the present application provides an auxiliary processing quality assessment method based on three-dimensional point cloud data, the method comprising:

[0005] Acquire a three-dimensional point cloud corresponding to the workpiece to be detected, and compare the three-dimensional point cloud with a limit model corresponding to the workpiece to be detected to determine an out-of-tolerance area of ​​the workpiece to be detected; the limit model is a simulation design model corresponding to the workpiece to be detected constructed based on a preset tolerance range, and the out-of-tolerance area is composed of point clouds in the three-dimensional point cloud that exceed the range of the limit model;

[0006] Based on the out-of-tolerance area, determining whether the workpiece to be inspected is qualified;

[0007] If the workpiece to be inspected is qualified, determining a roughness value, a waviness value, and a surface profile value corresponding to a target surface of the workpiece to be inspected;

[0008] Constructing an OBB bounding box of the workpiece to be inspected, and determining a target coordinate system corresponding to the target surface from a plurality of two-dimensional coordinate systems of the OBB bounding box;

[0009] Projecting the target point cloud on the target surface into the target coordinate system to obtain the two-dimensional coordinates of the projected target point cloud;

[0010] Determine at least two target evaluation indicators from the roughness value, the waviness value and the surface profile value;

[0011] Based on the two-dimensional coordinates of the projected target point cloud and at least two target evaluation indicators corresponding to the target point cloud, a multidimensional feature tensor corresponding to the target point cloud is obtained; the multidimensional feature tensor is used to characterize the quality characteristics of the target point cloud;

[0012] The multidimensional feature tensor corresponding to the target point cloud is input into a preset quality assessment model to predict the quality level of the target surface.

[0013] Optionally, comparing the three-dimensional point cloud with a limit model corresponding to the workpiece to be inspected to determine an out-of-tolerance area of ​​the workpiece to be inspected includes:

[0014] Roughly aligning the three-dimensional point cloud with a standard CAD model corresponding to the workpiece to be inspected;

[0015] Optimizing and adjusting the result of the rough registration to obtain the position of the three-dimensional point cloud after fine registration;

[0016] Based on the preset tolerance range and the standard CAD model, determining a maximum simulation model and a minimum simulation model corresponding to the workpiece to be inspected;

[0017] The point cloud in the three-dimensional point cloud that exceeds the maximum simulation model or the minimum simulation model is determined as an out-of-tolerance point cloud, and an out-of-tolerance area of ​​the workpiece to be inspected is determined based on the out-of-tolerance point cloud.

[0018] Optionally, determining whether the workpiece to be inspected is qualified based on the out-of-tolerance area includes:

[0019] Acquire test data of the out-of-tolerance area in a simulation test or physical test scenario;

[0020] Based on the test data, determine whether the workpiece to be inspected is qualified.

[0021] Optionally, when it is determined that the workpiece to be inspected is unqualified, the method further includes:

[0022] Based on the test data, tracing back the process parameters in the production process of the workpiece to be tested;

[0023] Based on the process parameters, the reasons for the deviation are determined, and based on the reasons for the deviation, alarm information is generated; the alarm information is used to remind the staff to optimize and adjust the process parameters.

[0024] Optionally, after obtaining the three-dimensional point cloud corresponding to the workpiece to be inspected, the method further includes:

[0025] Determining a plurality of continuous smooth regions and discontinuous smooth regions in the workpiece to be inspected according to the curvature or normal change rate of the three-dimensional point cloud;

[0026] Among them, the continuous smooth area is used to reflect the quality characteristics of the roughness and waviness of the surface of the workpiece to be inspected, and the deformation information in the discontinuous smooth area is used to reflect the deformation error characteristics of the surface of the workpiece to be inspected, and is used to provide a detection direction for determining the out-of-tolerance area.

[0027] Optionally, when the workpiece to be inspected is qualified, determining a roughness value and a waviness value corresponding to a target surface of the workpiece to be inspected includes:

[0028] For any of the continuous smooth regions in the target surface, multi-scale filtering is performed on the continuous smooth region to obtain a high-frequency feature point cloud and a low-frequency feature point cloud corresponding to the continuous smooth region;

[0029] Dividing the high-frequency feature point cloud into regions based on a first minimum adjacent distance to obtain a plurality of first sub-regions, and fitting the plurality of high-frequency feature point clouds in the first sub-regions to obtain a first reference surface corresponding to each of the first sub-regions;

[0030] Based on the high-frequency feature point cloud and the first reference surface, determining a roughness value corresponding to each of the first sub-regions; wherein in the first sub-regions, the roughness values ​​of all high-frequency feature point clouds are the same;

[0031] Dividing the low-frequency feature point cloud into regions based on a second minimum adjacent distance to obtain a plurality of second sub-regions, and fitting the plurality of low-frequency feature point clouds in the second sub-regions to obtain a second reference surface corresponding to each of the second sub-regions; wherein the first minimum adjacent distance is smaller than the second minimum adjacent distance;

[0032] Based on the low-frequency feature point cloud and the second reference surface, a waviness value corresponding to each of the second sub-regions is determined; wherein, in the second sub-regions, the waviness values ​​of all the low-frequency feature point clouds are the same.

[0033] Optionally, when the workpiece to be inspected is qualified, determining a surface profile value corresponding to a target surface of the workpiece to be inspected includes:

[0034] Based on the result of precise registration of the three-dimensional point cloud with the standard CAD model corresponding to the workpiece to be inspected, determining the Euclidean distance between any of the point clouds and the corresponding point in the standard CAD model;

[0035] Dividing the continuous smooth area into regions based on the second minimum adjacent distance to obtain a plurality of third sub-regions, any of the third sub-regions including a plurality of the high-frequency feature point clouds and / or the low-frequency feature point clouds;

[0036] For any of the third sub-regions, determine the mean of the Euclidean distances of all point clouds in the third sub-region, and use the mean as the surface contour value of the third sub-region; wherein, in the third sub-region, the surface contour values ​​of all point clouds are the same.

[0037] Optionally, when the target evaluation index is a roughness value, a waviness value, and a surface profile value, a multidimensional feature tensor corresponding to the target point cloud is obtained based on the two-dimensional coordinates after the projection of the target point cloud and at least two target evaluation indexes corresponding to the target point cloud, including:

[0038] Determine the roughness value, the waviness value, and the surface profile value corresponding to the target point cloud as z values ​​of different channels of the target point cloud;

[0039] Based on the projected two-dimensional coordinates (x, y) of the target point cloud and the z value, a three-dimensional feature tensor (x, y, 3) corresponding to the target point cloud is obtained;

[0040] Among them, x represents the horizontal position of the target point cloud after projection, y represents the vertical position of the target point cloud after projection, and z represents the channel value of the target point cloud.

[0041] Based on the same inventive concept, a second aspect of an embodiment of the present application provides an auxiliary processing quality assessment device based on three-dimensional point cloud data, the device comprising:

[0042] A point cloud registration module is used to obtain a three-dimensional point cloud corresponding to the workpiece to be detected, and compare the three-dimensional point cloud with a limit model corresponding to the workpiece to be detected to determine an out-of-tolerance area of ​​the workpiece to be detected; the limit model is a simulation design model corresponding to the workpiece to be detected constructed based on a preset tolerance range, and the out-of-tolerance area is composed of point clouds in the three-dimensional point cloud that exceed the range of the limit model;

[0043] An out-of-tolerance detection module is used to determine whether the workpiece to be detected is qualified based on the out-of-tolerance area; an index value determination module is used to determine the roughness value, waviness value and surface profile value corresponding to the target surface of the workpiece to be detected if the workpiece to be detected is qualified;

[0044] An OBB bounding box construction module is used to construct an OBB bounding box of the workpiece to be inspected, and determine a target coordinate system corresponding to the target surface from multiple two-dimensional coordinate systems of the OBB bounding box; a point cloud projection module is used to project a target point cloud in the target surface into the target coordinate system to obtain the two-dimensional coordinates of the projected target point cloud;

[0045] An index selection module, used to determine at least two target evaluation indexes from the roughness value, the waviness value and the surface profile value;

[0046] A quality feature construction module, used to obtain a multidimensional feature tensor corresponding to the target point cloud based on the two-dimensional coordinates of the projected target point cloud and at least two target evaluation indicators corresponding to the target point cloud; the multidimensional feature tensor is used to characterize the quality characteristics of the target point cloud;

[0047] The comprehensive quality assessment module is used to input the multi-dimensional feature tensor corresponding to the target point cloud into a preset quality assessment model to predict the quality level of the target surface.

[0048] Based on the same inventive concept, the third aspect of an embodiment of the present application provides a storage medium, which stores machine executable instructions. When the machine executable instructions are executed by a processor, the auxiliary processing quality assessment method based on three-dimensional point cloud data proposed in the first aspect of the present application is implemented.

[0049] Based on the same inventive concept, the fourth aspect of an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes, the auxiliary processing quality assessment method based on three-dimensional point cloud data proposed in the first aspect of the present application is implemented.

[0050] Compared with the prior art, this application has the following advantages:

[0051] The embodiment of the present application provides an auxiliary processing quality assessment method based on three-dimensional point cloud data. First, the three-dimensional point cloud corresponding to the workpiece to be inspected is obtained, and it is compared with the limit model corresponding to the workpiece to be inspected to determine the out-of-tolerance area of ​​the workpiece to be inspected, which is used for key inspection to determine whether the workpiece to be inspected is qualified. If the workpiece to be inspected is qualified, the OBB directional encirclement method is used to determine the target coordinate system corresponding to the target surface of the workpiece to be inspected, and combined with at least two target evaluation indicators corresponding to the target point cloud in the target surface of the workpiece to be inspected, a multi-dimensional feature tensor corresponding to the target point cloud is established to represent the quality characteristics of the target point cloud through a multi-dimensional data structure. Finally, the multi-dimensional feature tensor corresponding to the target point cloud is input into a preset quality assessment model to predict the quality level of the target surface, thereby assisting in achieving a comprehensive and accurate assessment of the quality of the workpiece machining surface. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0053] Figure 1 is a flow chart of an auxiliary processing quality assessment method based on three-dimensional point cloud data in one embodiment of the present application;

[0054] Figure 2 is a schematic diagram of the structure of a workpiece to be inspected in one embodiment of the present application;

[0055] Figure 3 is a schematic diagram of a three-dimensional point cloud of a workpiece to be inspected in one embodiment of the present application;

[0056] Figure 4 is a schematic diagram of a CAD limit model of a workpiece to be inspected in one embodiment of the present application;

[0057] Figure 5 is a schematic diagram of an OBB bounding box in an embodiment of the present application;

[0058] Figure 6 is a structural diagram of a quality assessment model in one embodiment of the present application;

[0059] Figure 7 It is a logical schematic diagram of an auxiliary processing quality assessment method based on three-dimensional point cloud data in one embodiment of the present application;

[0060] Figure 8 It is a functional module diagram of an auxiliary processing quality assessment device based on three-dimensional point cloud data in one embodiment of the present application;

[0061] Fig. 9 It is a structural schematic diagram of an electronic device in one embodiment of the present application. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0063] Finished product inspection is an important means of evaluating whether the processed workpiece is qualified. However, the current finished product inspection often faces such a dilemma: all inspection indicators meet the inspection requirements, but when assembling, either it cannot be assembled or the assembly gap is too large, which seriously affects the assembly process. The reason for this problem is that the current inspection is the superposition of many local inspections. For example, to detect whether a workpiece is qualified, it is currently mainly tested to see whether the various contour dimensions of the workpiece meet the requirements, and then to see whether the roughness, waviness, coaxiality, etc. of the main surface of the workpiece meet the requirements. When all inspection indicators meet the requirements, the workpiece is judged to be qualified.

[0064] However, since this method cannot predict in advance the key inspection areas that may exceed the tolerance, the inspection efficiency is often low. At the same time, in actual work, it is problematic to judge whether the workpiece is qualified only by superimposing multiple indicators. Specifically, roughness measurement generally measures a local area, so the roughness result obtained is closely related to the selected area. For example, when a multi-coordinate probe is used to measure the shape of a free surface (such as the surface of a blade body), the surface evaluation is generally obtained by measuring points on multiple profile lines. This method cannot allow the probe to traverse all points on the free surface. Therefore, when a point that has not been measured exceeds the tolerance, the detection method cannot detect it, which will cause misjudgment.

[0065] Most of the current surface quality assessment methods for machined parts focus on selecting local indicators based on actual application scenarios, and the evaluation of some indicator values ​​is often based on the local quality of the workpiece surface to represent the overall quality, such as the evaluation of roughness and waviness. For machined parts with complex shapes or deformations, this method is often difficult to fully reflect their overall quality.

[0066] In view of this, the present application proposes an auxiliary processing quality assessment method based on three-dimensional point cloud data to solve the local problems of existing surface quality assessment methods for processed parts, improve the ability to effectively respond to surface deformation of processed parts, and at the same time improve the shortcomings of existing methods in comprehensively evaluating processing quality.

[0067] Please refer to Figure 1 and Figure 7 , Figure 1 is a flow chart of an auxiliary processing quality assessment method based on three-dimensional point cloud data proposed in one embodiment of the present application. Figure 7 FIG. 1 is a logic diagram of an auxiliary processing quality assessment method based on three-dimensional point cloud data in one embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0068] S101: Acquire a three-dimensional point cloud corresponding to the workpiece to be inspected, and compare the three-dimensional point cloud with a limit model corresponding to the workpiece to be inspected to determine an out-of-tolerance area of ​​the workpiece to be inspected.

[0069] Among them, the limit model is a simulation design model corresponding to the workpiece to be tested, which is constructed based on a preset tolerance range. The tolerance range refers to the allowable range of variation in the size, shape, position and other parameters of the workpiece or component during the manufacturing and design process. The limit model can be constructed using CAD (Computer Aided Design) software or Solidworks (a three-dimensional computer-aided design software) software in combination with the tolerance range.

[0070] In this embodiment, if Figure 7 As shown, first, the three-dimensional point cloud data of the workpiece to be inspected can be collected through non-contact measurement methods such as laser scanning or structured light scanning. For example, a laser rangefinder is used to emit a laser to the surface of the workpiece to be inspected, and the distance is calculated based on the time or phase difference when the laser is reflected back, thereby obtaining the point cloud data of the workpiece to be inspected. Alternatively, a structured light projector is used to project a specific pattern onto the surface of the workpiece to be inspected, and the deformation of the pattern on the surface of the workpiece to be inspected is captured to calculate the point cloud data of the workpiece to be inspected. For example, Figure 2 After the workpiece to be inspected is 3D scanned, the distribution of the 3D point cloud is as follows: Figure 3 shown.

[0071] Then, the 3D point cloud is compared with the standard CAD model corresponding to the workpiece to be inspected (i.e. Figure 7 The 3D point cloud is aligned to the standard CAD model through a transformation matrix (including rotation and translation) so that the position and posture of the 3D point cloud and the standard CAD model in space are as consistent as possible. Figure 2 The corresponding CAD limit model of the workpiece to be inspected is shown in Figure 4 shown.

[0072] On the basis of registration, a limit model corresponding to the workpiece to be inspected is constructed based on the preset tolerance range, and the three-dimensional point cloud is compared with the limit model to determine the out-of-tolerance area of ​​the workpiece to be inspected. Among them, the out-of-tolerance area is composed of point clouds in the three-dimensional point cloud that exceed the range of the limit model. Simply put, the limit model defines the maximum and minimum values ​​of each position of the qualified workpiece to be inspected. Only when the three-dimensional point cloud is within the spatial range defined by the limit model, the point cloud at this location is considered to meet the processing requirements. Correspondingly, if the point cloud at a certain location of the workpiece to be inspected exceeds the spatial range defined by the limit model, the point cloud at this location is marked as an out-of-tolerance point, and finally one or more out-of-tolerance areas are formed, which provide a detection basis for the subsequent evaluation of the workpiece to be inspected.

[0073] S102: Determine whether the workpiece to be inspected is qualified based on the out-of-tolerance area.

[0074] It is easy to understand that the out-of-tolerance area reflects the situation where there is a significant difference between the actual shape, size or position of the workpiece to be inspected and the limit model. If the out-of-tolerance area is large, it means that the workpiece to be inspected is likely to be an unqualified product. However, since the three-dimensional point cloud detection itself is error-prone, in order to ensure the accuracy of the evaluation, further testing is required to determine whether the workpiece to be inspected is qualified. In other words, based on the out-of-tolerance area, it can be judged whether it is a key functional area, and further measurements of the key indicators in the area can be made to determine whether the workpiece to be inspected is a qualified workpiece.

[0075] Specifically, during the inspection, the out-of-tolerance area can be taken as the key inspection object, and then the performance of the out-of-tolerance area can be verified through simulation or actual measurement. For example, the service performance of the workpiece to be inspected under actual assembly conditions is simulated and tested through simulation software, and the performance of the out-of-tolerance area is focused on to determine whether the corresponding assembly requirements are met. For example, when a certain surface of the workpiece to be inspected is used as a sealing surface for assembly, if the simulation test finds that there is leakage in the out-of-tolerance area, it is considered that the surface does not meet the assembly requirements, and the workpiece to be inspected is determined to be unqualified. If the simulation test finds that there are no problems with the out-of-tolerance area and other areas, that is, the service performance of the workpiece to be inspected is good, then the workpiece to be inspected is determined to be qualified, and the surface quality of the workpiece to be inspected is further comprehensively evaluated.

[0076] In the related art, the detection of whether a workpiece is qualified is usually based on experience, and each part and corresponding indicator are sequentially detected to see if they meet the corresponding requirements. If they are all satisfied, the workpiece is judged to be qualified. However, this method is blind, and it may often happen that there are no problems in the previous detections, and problems only occur in the last few steps of detection. This seriously wastes detection time and affects detection efficiency. In this embodiment, by performing three-dimensional point cloud registration, the out-of-tolerance area is determined, which provides a direction for product detection. Because the out-of-tolerance area is most likely the part where problems may occur, therefore, by focusing on detecting the out-of-tolerance area first, the quality problems of the workpiece can be detected more quickly, and then the unqualified workpieces can be determined. Compared with the existing workpiece detection method, it can reduce manpower consumption and improve the overall product detection efficiency.

[0077] In addition, for unqualified processed parts detected in the out-of-tolerance area, the process parameter records in the production process can be traced back based on the test result data to find out the possible causes of the out-of-tolerance and solve them in a targeted manner to improve the quality of the processed parts produced subsequently.

[0078] S103: When the workpiece to be inspected is qualified, the roughness value, the waviness value and the surface profile value corresponding to the target surface of the workpiece to be inspected are determined.

[0079] In this embodiment, the workpiece to be inspected has multiple surfaces, and the requirements for the surface vary depending on the application scenario of the workpiece to be inspected. For example, when surface A1 of workpiece A is used as a mating surface for assembly with workpiece B, on the basis that the workpiece is qualified, workpiece A with better surface A1 quality should be selected for assembly as much as possible. Similarly, when surface A2 of workpiece A works as a sealing surface, workpiece A with better surface A2 quality should be selected as much as possible. Therefore, the target surface refers to the surface to be evaluated this time among the multiple surfaces of the workpiece to be inspected, and can be specifically determined according to the application scenario of the workpiece to be inspected this time.

[0080] In order to select the workpiece with the best target surface quality from multiple qualified workpieces, such as Figure 7 As shown, the present application comprehensively evaluates the quality of the target surface by detecting key indicator values ​​such as the roughness value, waviness value and surface profile value of the target surface.

[0081] Among them, the roughness value reflects the slight unevenness of the machined surface. There are peaks and valleys with very small spacing. The distance between two peaks or two valleys (i.e., wave pitch) is very small, usually less than 1mm. The formation of this type of micro-geometric shape error is the result of the combined effect of multiple processing factors. For example, the friction between the tool and the part surface during processing, the plastic deformation of the surface metal at the moment of chip separation, the high-frequency vibration of the process system, and the discharge pits generated by electrical processing will all contribute to the generation of surface roughness. The waviness value reflects the difference in the height of the workpiece surface fluctuations. The wave pitch range it involves is between 1-10mm, which is in a "middle ground" between macro-shape error and micro-surface roughness. It is another type of surface geometric shape error measurement indicator independent of surface roughness. The surface profile value belongs to the category of geometric tolerance, which focuses on controlling the degree of deviation between the actual surface profile and the ideal profile. In actual application scenarios, surface contour is mainly used to control the shape accuracy of the workpiece surface. Of course, in some cases, due to its relationship with surrounding elements, it will also indirectly constrain elements such as the position and direction of the surface. It can be applied to planes, curved surfaces, irregular surfaces, polygons, cylinders, rotating surfaces, cones, coplanar surfaces and other types of workpiece surfaces to ensure processing accuracy and quality in line with design expectations.

[0082] S104: construct an OBB bounding box of the workpiece to be inspected, and determine a target coordinate system corresponding to the target surface from multiple two-dimensional coordinate systems of the OBB bounding box.

[0083] OBB (Oriented Bounding Box) is a cuboid that can tightly surround an object. It can be rotated according to the direction of the object to better approximate the actual shape of the object. For example, please refer to Figure 5 , Figure 5 It is a schematic diagram of an OBB bounding box in one embodiment of the present application.

[0084] like Figure 5 As shown, this embodiment uses the OBB directional bounding box method to enclose the three-dimensional point cloud of the workpiece to be inspected as tightly as possible. The constructed OBB bounding box has six boundary surfaces, each of which corresponds to the surface of the workpiece to be inspected, and each boundary surface has a different two-dimensional coordinate system.

[0085] In this embodiment, after the target surface to be evaluated of the workpiece to be inspected is determined through the above steps, the two-dimensional coordinate system of the boundary surfaces corresponding to the target surface of the workpiece to be inspected among the six boundary surfaces of the OBB bounding box can be used as the target coordinate system corresponding to the target surface for projection of the three-dimensional point cloud.

[0086] S105: Projecting the target point cloud on the target surface into the target coordinate system to obtain the two-dimensional coordinates of the projected target point cloud.

[0087] In this embodiment, the target point cloud in the target surface is projected into the target coordinate system corresponding to the target surface in the OBB bounding box, and the two-dimensional coordinates (x, y) of each target point cloud are recorded to obtain the position of each target point cloud in the projected boundary surface.

[0088] S106: Determine at least two target evaluation indicators from the roughness value, the waviness value and the surface profile value.

[0089] In this embodiment, at least two of the multiple indicators can be selected as target evaluation indicators according to the on-site inspection requirements to conduct a comprehensive quality assessment of the workpiece to be inspected. For example, the roughness value and the waviness value can be selected from the roughness value, the waviness value, and the surface profile value as the target evaluation indicators required this time, or the roughness value, the waviness value, and the surface profile value can all be determined as target evaluation indicators to adapt to different quality inspection scenarios and requirements.

[0090] S107: Based on the two-dimensional coordinates of the projected target point cloud and at least two target evaluation indicators corresponding to the target point cloud, a multidimensional feature tensor corresponding to the target point cloud is obtained; the multidimensional feature tensor is used to characterize the quality characteristics of the target point cloud.

[0091] Among them, the multi-dimensional feature tensor is used to characterize the quality characteristics of the target point cloud.

[0092] In this embodiment, the two-dimensional target coordinate system corresponding to the target surface is determined by using the OBB directional bounding box method, and after performing point cloud projection to obtain the two-dimensional coordinates of the target point cloud after projection, a multidimensional data structure is established in combination with at least two target evaluation indicators corresponding to the target point cloud to characterize the quality characteristics of the target point cloud in the target surface.

[0093] Specifically, when the target evaluation indicators are roughness value, waviness value and surface profile value, the process includes: determining the roughness value, waviness value and surface profile value corresponding to the target point cloud as the z values ​​of different channels of the target point cloud; based on the two-dimensional coordinates (x, y) and z value after projection of the target point cloud, obtaining the three-dimensional feature tensor (x, y, 3) corresponding to the target point cloud; wherein x represents the position of the target point cloud in the horizontal direction after projection, y represents the position of the target point cloud in the vertical direction after projection, and z represents the channel value of the target point cloud.

[0094] In this embodiment, if Figure 7As shown, considering that the roughness value Sa, waviness value Wsa and surface profile value share the same position coordinate system and the information between the index values ​​is interrelated, the three index values ​​are regarded as the z values ​​of different channels of the same target point cloud, thereby constructing a three-dimensional feature tensor with a size of (x, y, 3) to represent the quality characteristics of the target point cloud in the target surface.

[0095] It can be similarly understood that after projecting the target point cloud in the target surface into the coordinate system of the corresponding boundary surface of the OBB bounding box, a multi-channel image is obtained, and the multi-channel value of each point in the image is the roughness value, waviness value and surface profile value of the target point cloud. Then, by extracting and processing the features of the multi-channel image, the quality evaluation of the target surface is achieved.

[0096] It should be noted that the index values ​​used in this application are not limited to the above-mentioned roughness values, waviness values ​​and surface profile values, but can also be other index values ​​such as surface finish and flatness, and the above-mentioned three-dimensional feature tensor can be constructed based on two index values ​​(in this case, the z value is a dual-channel), or based on three index values ​​(in this case, the z value is a three-channel). That is, based on the two-dimensional coordinates of the projected target point cloud and the selected multiple target evaluation indicators, the three-dimensional feature tensor constructed to characterize the quality characteristics of the target point cloud can be expressed as (x, y, n), where n represents the number of n different feature channels that the target point cloud has in the Z dimension.

[0097] S108: Inputting the multi-dimensional feature tensor corresponding to the target point cloud into a preset quality assessment model to predict the quality level of the target surface.

[0098] In this embodiment, the preset quality assessment model is used to evaluate the quality level of the target surface. The quality level includes three levels: excellent, good and qualified. For details, please refer to Figure 6 , Figure 6 It is a structural diagram of a quality assessment model in one embodiment of the present application.

[0099] like Figure 6As shown, the present application selects the feature extraction network after removing the classification layer as the basic architecture, such as VGGNet (Visual Geometry Group, a deep convolutional network structure), ResNet (residual network), InceptionNet (a deep learning neural network architecture), MobileNet (a lightweight neural network model) or EfficientNet (an efficient convolutional neural network), etc. In this example, taking three channels as an example, after extracting the features of the three-dimensional feature tensor (x, y, 3) corresponding to the input target point cloud, one or more fully connected layers are added to further process the extracted feature vectors, and finally an output layer is connected, which contains nodes equal to the number of quality grade categories (3 grade categories in this embodiment), and the predicted probability of each grade category is calculated through an appropriate activation function (such as Softmax), and the grade category with the largest probability is used as the quality grade evaluation result of the target surface. Thus, the three-dimensional feature tensor of the target surface is feature detected through the quality assessment model, and the surfaces of workpieces with different quality levels are effectively distinguished, so as to realize automated quality assessment.

[0100] Among them, the preset quality assessment model can be trained in the following way: obtain the measured index value of the sample surface of the sample workpiece, and determine the real quality grade of the sample surface based on the measured index value; determine the three-dimensional feature tensor corresponding to each sample point cloud based on the three-dimensional point cloud corresponding to the sample surface; use the initial quality assessment model to extract features from the three-dimensional feature tensor corresponding to the sample point cloud to obtain the predicted quality grade of the sample surface; compare the predicted quality grade with the real quality grade, adjust and optimize the network parameters of the initial quality assessment model, and obtain the preset quality assessment model.

[0101] In this embodiment, first, the measured index values ​​of the sample surface of the sample workpiece can be obtained by a stylus profilometer and a three-dimensional coordinate measuring machine, including the actual roughness value, waviness value and surface profile value. Based on these data, and in combination with the tolerance range of each index, the true quality level of the sample surface is determined, including three levels: excellent, good and qualified. Next, the processing quality data of the sample surface and its classification labels are integrated to form training samples for training and testing the initial quality assessment model. At the same time, the multi-dimensional feature tensor of the sample surface is determined through the above steps S103-S107, and it is used as the input of the initial quality assessment model to process the predicted quality level of the sample surface. During the training process, based on the difference between the predicted quality level and the true quality level, the network parameters of the initial quality assessment model are reversely adjusted and optimized, and finally a preset quality assessment model is obtained.

[0102] Convolutional neural networks can effectively capture the local features and spatial hierarchical structure of the workpiece surface. Convolutional neural networks automatically learn and extract high-level features from raw data, which can accurately reflect the quality of the workpiece surface. In this way, the model can accurately classify and evaluate the quality of the machined surface. Compared with traditional measurement methods, this method not only improves efficiency, but is also more convenient, providing the manufacturing industry with a smarter and faster solution for the surface quality assessment of workpieces.

[0103] In addition, when constructing a multi-dimensional feature tensor of the sample surface based on only two target evaluation indicators, for example, when the current scene only requires roughness and waviness values ​​as quality evaluation indicators, the input of the model can be changed to two channels by freezing part of the network structure based on the above-mentioned trained three-channel quality assessment model, and then retraining. Freezing certain network layers means that the parameters of these layers will not be updated during the training process, so that some common features that the model has learned can be retained. For example, in a deep neural network, the previous layers may have learned low-level features of the data, which may still be valuable for the new two-channel data. Freezing these layers can prevent the learned common knowledge from being destroyed. At the same time, retraining the unfrozen layers allows the model to adjust its parameters to adapt to the input method of only two-channel data, so that the model can better perform quality assessment based on the data of these two channels.

[0104] The above embodiment describes the process of performing quality assessment on the entire target surface of the workpiece to be inspected. In addition, quality assessment may be performed only on a partial area of ​​interest (e.g., a partial area selected manually) on the target surface, so as to perform local quality inspection and assessment according to the actual application scenario.

[0105] In another embodiment of the present application, the process of performing auxiliary processing quality assessment based on the three-dimensional point cloud data of the workpiece to be inspected mainly includes:

[0106] S401: Acquire a three-dimensional point cloud corresponding to the workpiece to be inspected, and determine a plurality of continuous smooth regions and discontinuous smooth regions in the workpiece to be inspected according to the curvature or normal change rate of the three-dimensional point cloud.

[0107] In this embodiment, if Figure 7As shown in the figure, after obtaining the 3D point cloud corresponding to the workpiece to be inspected, the 3D point cloud data is first preprocessed, including: 1. Data filtering and denoising, by setting a threshold, outliers and noise in the point cloud data are removed. Exemplarily, the main filtering methods include bilateral filtering, Gaussian filtering, conditional filtering, and straight-through filtering. 2. Downsampling, by reducing the resolution of the point cloud data to reduce the amount of data, thereby speeding up the processing speed. Exemplarily, the downsampling methods include random downsampling, grid downsampling, etc. 3. Data format conversion and standardization, according to actual needs, the point cloud data is converted from one format to another. Common point cloud data formats include PCD (Point Cloud Data), PLY (Polygon File Format, a file format for storing and exchanging 3D geometric models and point cloud data), OBJ (Object File Format, a file format for 3D models), etc. Libraries such as Open3D (an open source point cloud processing library) and PCL (Point Cloud Library, a large cross-platform open source C++ programming library) can be used for format conversion. Then, the point cloud data is standardized, such as unifying the coordinate system and units, to ensure the accuracy and consistency of subsequent processing. 4. Normal and curvature calculation, calculate the normal and curvature of each point in the point cloud. These features are very important for subsequent point cloud segmentation, classification and reconstruction processes.

[0108] Then, according to the curvature or normal change rate of the three-dimensional point cloud obtained after preprocessing, multiple continuous smooth areas and discontinuous smooth areas (i.e. Figure 7 The continuous smooth area is used to reflect the quality characteristics of the roughness and waviness of the surface of the workpiece to be inspected, and the deformation information in the discontinuous smooth area is used to reflect the deformation error characteristics of the surface of the workpiece to be inspected, and is used to provide the inspection direction for determining the out-of-tolerance area.

[0109] This embodiment divides the three-dimensional point cloud data into continuous smooth areas and discontinuous smooth areas according to local features (such as curvature, normal changes, etc.). The continuous smooth areas have lower curvature and can better reflect the quality characteristics of surface roughness and waviness. The discontinuous smooth areas (such as edges, corners, etc.) are the main source of deformation errors and are also the key areas of focus for point cloud registration.

[0110] S402: Compare the three-dimensional point cloud with the limit model corresponding to the workpiece to be inspected to determine the out-of-tolerance area of ​​the workpiece to be inspected.

[0111] In this embodiment, before comparing with the limit model, it is necessary to first align the 3D point cloud with the standard CAD model corresponding to the workpiece to be inspected. Point cloud alignment mainly includes two stages: coarse alignment and fine alignment. Specifically, the process mainly includes:

[0112] S402-1: Roughly align the three-dimensional point cloud with the standard CAD model corresponding to the workpiece to be inspected.

[0113] The main purpose of coarse registration is to quickly estimate a rough point cloud registration matrix when the initial conditions are unknown, so as to roughly align the 3D point cloud and the standard CAD model and provide a better transformation initial value for fine registration.

[0114] Specifically, the principal component analysis (PCA) method can be used, which mainly uses the principal axis direction of the point cloud data for registration. First, the covariance matrix of the 3D point cloud and the standard CAD model is calculated, and the main characteristic components, that is, the principal axis direction of the point cloud data, are calculated based on the covariance matrix. Then, the rotation matrix is ​​calculated through the principal axis direction, and the translation vector is directly calculated by calculating the displacement of the center coordinates of the 3D point cloud and the standard CAD model.

[0115] S402-2: Optimize and adjust the result of the coarse registration to obtain the position of the three-dimensional point cloud after fine registration.

[0116] Fine registration is based on coarse registration. Based on the initial value of the rotation and translation matrix of coarse registration, a more accurate rotation and translation matrix is ​​further calculated, so that the position and posture of the 3D point cloud and the standard CAD model in space are as consistent as possible, achieving accurate matching of the disordered point cloud.

[0117] Specifically, the iterative closest point algorithm (ICP algorithm) can be used to reduce the distance error between point clouds through iterative updates. Each iteration includes two steps: finding matching point pairs and solving the optimal transformation. Finding matching points is done through nearest neighbor search, while solving the optimal transformation is usually achieved using the singular value decomposition (SVD) method.

[0118] S402-3: Based on the preset tolerance range and the standard CAD model, determine the maximum simulation model and the minimum simulation model corresponding to the workpiece to be inspected.

[0119] S402-4: Determine the point cloud in the three-dimensional point cloud that exceeds the maximum simulation model or the minimum simulation model as an out-of-tolerance point cloud, and determine the out-of-tolerance area of ​​the workpiece to be inspected based on the out-of-tolerance point cloud.

[0120] In this embodiment, the maximum simulation model and the minimum simulation model define the range of qualified workpieces to be inspected in space. Only when the three-dimensional point cloud is within the spatial range defined by the maximum simulation model and the minimum simulation model, the point cloud at that location is considered to meet the processing requirements. Correspondingly, if the point cloud at a certain location of the workpiece to be inspected exceeds the spatial range defined by the maximum simulation model and the minimum simulation model, the point cloud at that location is marked as an out-of-tolerance point, and ultimately one or more out-of-tolerance areas are formed, providing a detection basis for the subsequent evaluation of the workpiece to be inspected.

[0121] S403: Determine whether the workpiece to be inspected is qualified based on the out-of-tolerance area.

[0122] Specifically, the process mainly includes: obtaining test data of the out-of-tolerance area in a simulation test or a physical test scenario; and determining whether the workpiece to be inspected is qualified based on the test data.

[0123] In this embodiment, after determining the out-of-tolerance area, it is possible to verify whether there is a problem with the out-of-tolerance part through simulation or actual measurement. For example, the roughness, waviness, and surface profile of the out-of-tolerance area can be measured by a stylus profilometer and a three-coordinate measuring instrument, and then the out-of-tolerance area can be judged according to the preset tolerance range to see whether the processing quality requirements are met, thereby determining whether the workpiece to be tested is qualified. In addition, simulation testing can also be performed through simulation software to collect test data of the out-of-tolerance area during service, including assembly data of the mating surface, sealing performance data of the sealing surface, etc., thereby determining whether the workpiece to be tested is qualified. For example, if the assembly data obtained by simulating the out-of-tolerance area of ​​the mating surface indicates that the workpiece is difficult to assemble or the operation is unstable, then the workpiece to be tested is considered unqualified, or if the sealing performance data obtained by simulating the out-of-tolerance area of ​​the sealing surface indicates that there is leakage on the sealing surface, then the workpiece to be tested is considered unqualified.

[0124] Furthermore, when it is determined that the workpiece to be inspected is unqualified, the above method further includes:

[0125] Based on the test data, the process parameters of the workpiece to be tested in the production process are traced back; based on the process parameters, the cause of the deviation is determined, and based on the cause of the deviation, an alarm message is generated; the alarm message is used to remind the staff to optimize and adjust the process parameters.

[0126] In this embodiment, for unqualified workpieces to be inspected, the possible causes of the deviation can be found out based on their test data by tracing back the process parameters in the production process, and then reported to the staff, so that the staff can optimize the production process in a targeted manner and improve the quality of the workpieces produced subsequently.

[0127] For example, in the manufacturing process of aircraft engine blades, if the surface processing is improper, excessive surface roughness, waviness or poor surface profile may be generated, which will cause stress concentration and other problems. Therefore, after determining the out-of-tolerance area, the test data of the out-of-tolerance area can be collected by simulation or actual measurement to verify whether there is a problem in the out-of-tolerance area. These problems are often caused by incorrect process parameter settings, such as insufficient or interrupted coolant supply, which will cause poor heat dissipation of the tool, thereby affecting the processing quality; it may also be because the debris generated during the processing is not cleaned up in time, hindering the normal operation of the tool. In addition, when tracing back the process parameters, the process parameters can be optimized and adjusted accordingly by detecting whether the actual flow rate of the coolant meets the requirements, whether there are abnormal fluctuations in the tool shaft torque monitoring value, and whether the key processing parameters such as cutting depth and width exceed the preset range.

[0128] S404: When the workpiece to be inspected is qualified, determine the roughness value and waviness value corresponding to the target surface of the workpiece to be inspected.

[0129] In this embodiment, after a plurality of continuous smooth regions of the target surface are determined in step S401, the roughness value and waviness value corresponding to the target surface of the workpiece to be inspected may be determined in the following manner:

[0130] S404-1: For any continuous smooth region in the target surface, perform multi-scale filtering on the continuous smooth region to obtain a high-frequency feature point cloud and a low-frequency feature point cloud corresponding to the continuous smooth region.

[0131] In this embodiment, there are multiple continuous smooth areas on the target surface, and each continuous smooth area reflects the quality characteristics of the roughness and waviness of the surface of the workpiece to be inspected. Figure 7 As shown in FIG. 1 , the high-frequency feature point cloud obtained after multi-scale filtering reflects the roughness of the target surface, and the low-frequency feature point cloud reflects the waviness of the target surface.

[0132] All continuous smooth areas in the target surface reflect the roughness and waviness values ​​of the target surface together. Specifically, for any continuous smooth area in the target surface, a high-pass filter or edge detection algorithm, such as a variant of bilateral filtering or Gaussian filtering (adjusting parameters to emphasize high-frequency components), can be used to filter the three-dimensional point cloud data. The point cloud data after high-pass filtering will mainly contain high-frequency components, that is, the tiny undulations and details of the surface, which reflect the roughness of the continuous smooth area. At the same time, the three-dimensional point cloud data is filtered using a low-pass filter, such as a mean filter, which can smooth out high-frequency noise and retain low-frequency components. The point cloud data after low-pass filtering will mainly contain low-frequency components, that is, the macroscopic undulations and shapes of the surface, which reflect the waviness of the continuous smooth area.

[0133] S404-2: Divide the high-frequency feature point cloud into regions based on the first minimum adjacent distance to obtain multiple first sub-regions, and fit the multiple high-frequency feature point clouds in the first sub-regions to obtain a first reference surface corresponding to each first sub-region.

[0134] S404-3: Determine the roughness value corresponding to each first sub-region based on the high-frequency feature point cloud and the first reference surface; wherein, in the first sub-region, the roughness values ​​of all high-frequency feature point clouds are the same.

[0135] In this embodiment, the first minimum adjacent distance refers to a preset search radius corresponding to the high-frequency feature point cloud, that is, for the high-frequency feature point cloud, by defining a search radius, the high-frequency feature point cloud is divided into regions to obtain multiple first sub-regions. According to the first sub-region, it can be determined how many points can be used to fit the first reference surface, and further according to the distance from the high-frequency feature point cloud in the first sub-region to the first reference surface, the roughness value corresponding to the first sub-region in the continuous smooth region is determined. In other words, a continuous smooth region includes multiple first sub-regions, and for a first sub-region, the roughness values ​​of all point clouds therein are the same. That is, different first sub-regions correspond to different roughness values.

[0136] Specifically, for any continuous smooth area, when determining the roughness value corresponding to the continuous smooth area, the vertical distance from each high-frequency feature point to its corresponding first reference surface is calculated through the point cloud data after high-frequency feature extraction, and these distances are statistically analyzed, such as calculating the standard deviation, average value, etc., to quantify the roughness of the continuous smooth area. The larger the roughness value, the more uneven the target surface is, that is, the higher the roughness.

[0137] S404-4: Divide the low-frequency feature point cloud into regions based on the second minimum adjacent distance to obtain multiple second sub-regions, and fit the multiple low-frequency feature point clouds in the second sub-regions to obtain a second reference surface corresponding to each second sub-region. The first minimum adjacent distance is smaller than the second minimum adjacent distance.

[0138] S404-5: Based on the low-frequency feature point cloud and the second reference surface, determine the waviness value corresponding to each second sub-region; wherein, in the second sub-region, the waviness values ​​of all the low-frequency feature point clouds are the same.

[0139] In this embodiment, the second minimum adjacent distance refers to a preset search radius corresponding to the low-frequency feature point cloud, that is, for the low-frequency feature point cloud, by defining a search radius, the low-frequency feature point cloud is divided into regions to obtain multiple second sub-regions. According to the second sub-region, it can be determined how many points can be used to fit the second reference surface, and further according to the distance from the low-frequency feature point cloud in the second sub-region to the second reference surface, the corrugation value corresponding to the second sub-region in the continuous smooth region is determined. In other words, a continuous smooth region includes multiple second sub-regions, and for a second sub-region, the corrugation values ​​of all point clouds therein are the same. That is, different second sub-regions correspond to different corrugation values.

[0140] Specifically, the waviness value is determined in a similar manner to the roughness value. The larger the waviness value, the more obvious the macroscopic fluctuations of the target surface, that is, the higher the waviness.

[0141] In addition, since the roughness is reflected by the high-frequency feature point cloud and the waviness is reflected by the low-frequency feature point cloud, and the distance between points in the low-frequency feature point cloud is relatively far, a larger search radius needs to be designed for the low-frequency feature point cloud. Therefore, the first minimum adjacent distance needs to be smaller than the second minimum adjacent distance.

[0142] S405: When the workpiece to be inspected is qualified, a surface profile value corresponding to a target surface of the workpiece to be inspected is determined.

[0143] Specifically, the process mainly includes:

[0144] S405-1: Based on the result of precise registration of the three-dimensional point cloud with the standard CAD model corresponding to the workpiece to be inspected, determine the Euclidean distance between any point cloud and the corresponding point in the standard CAD model.

[0145] In this embodiment, the fitting surface formed by the three-dimensional point cloud is compared with the corresponding surface in the standard CAD model. For each point cloud, the least square method is used to calculate the Euclidean distance between the corresponding position of the point cloud and the standard CAD model, so as to provide a data basis for the subsequent calculation of the surface profile value.

[0146] S405-2: Divide the continuous smooth area into regions based on the second minimum adjacent distance to obtain a plurality of third sub-regions, wherein any third sub-region includes a plurality of high-frequency feature point clouds and / or low-frequency feature point clouds.

[0147] In this embodiment, there is no need to perform multi-scale filtering on the three-dimensional point cloud, but the second minimum adjacent distance is used to directly divide the three-dimensional point cloud into regions to obtain multiple third sub-regions. In other words, any third sub-region may include multiple high-frequency feature point clouds and low-frequency feature point clouds, or may only include high-frequency feature point clouds or low-frequency feature point clouds. Therefore, the surface profile of the entire continuous smooth area (and the entire target surface) can be reflected and understood through the surface profile values ​​of multiple third sub-regions.

[0148] S405-3: For any third sub-region, determine the mean of the Euclidean distances of all point clouds in the third sub-region, and use the mean as the surface contour value of the third sub-region; wherein, in the third sub-region, the surface contour values ​​of all point clouds are the same.

[0149] In this embodiment, the surface contour values ​​of different third sub-regions are different, while the surface contour values ​​of all point clouds in the same third sub-region are the same.

[0150] S406: Projecting the target point cloud on the target surface into the target coordinate system to obtain the two-dimensional coordinates of the projected target point cloud.

[0151] S407: Based on the two-dimensional coordinates of the projected target point cloud and the roughness value, waviness value and surface profile value corresponding to the target point cloud, a three-dimensional feature tensor corresponding to the target point cloud is obtained; the three-dimensional feature tensor is used to characterize the quality characteristics of the target point cloud.

[0152] In this embodiment, the target surface has multiple continuous smooth areas. The prior art determines the roughness value, waviness value and surface contour value of the target surface by taking the average values ​​of these areas, while the present application directly uses the OBB bounding box method to determine the position range (x, y) of the target surface mapped to the corresponding plane of the OBB bounding box, and then uses the roughness value, waviness value and surface contour value within the range as z values, which are stacked up to form a 3-channel three-dimensional feature tensor, similar to an RGB (Red, Green, Blue) image (any point cloud within the range has corresponding roughness value, waviness value and surface contour value), so as to better represent the quality characteristics of the target surface.

[0153] S408: Inputting the three-dimensional feature tensor corresponding to the target point cloud into a preset quality assessment model to predict the quality level of the target surface.

[0154] Among them, steps S406-S408 please refer to the above steps S105-S107, which will not be repeated here.

[0155] This application uses point cloud preprocessing, feature extraction, and point cloud segmentation to divide continuous smooth areas and discontinuous smooth areas in the global view. Then, a limit model is reconstructed based on the standard CAD model, and the three-dimensional point cloud is aligned with the standard CAD model, and then compared with the limit model to find out the out-of-tolerance area for key detection to determine whether the workpiece is qualified and improve the detection efficiency. In addition, the processing parameters can be traced back according to the service performance of the out-of-tolerance area to find out the causes that may cause the out-of-tolerance and solve them in a targeted manner. Finally, based on the roughness value, waviness value, and surface profile value of the target surface of the workpiece to be detected, the OBB directional bounding box method and convolutional neural network are used to construct a comprehensive processing quality assessment model, thereby assisting in the realization of a comprehensive and accurate assessment of the surface quality of the workpiece, solving the local problems of the existing workpiece surface quality assessment methods, and improving the ability to effectively respond to the deformation of the workpiece processing surface.

[0156] Please refer to Figure 8 Based on the same inventive concept, the second aspect of the embodiment of the present application provides an auxiliary processing quality assessment device based on three-dimensional point cloud data. The auxiliary processing quality assessment device based on three-dimensional point cloud data 800 includes:

[0157] The point cloud registration module 801 is used to obtain the three-dimensional point cloud corresponding to the workpiece to be inspected, and compare the three-dimensional point cloud with the limit model corresponding to the workpiece to be inspected to determine the out-of-tolerance area of ​​the workpiece to be inspected; the limit model is a simulation design model corresponding to the workpiece to be inspected constructed based on a preset tolerance range, and the out-of-tolerance area is composed of point clouds in the three-dimensional point cloud that exceed the range of the limit model;

[0158] The out-of-tolerance detection module 802 is used to determine whether the workpiece to be inspected is qualified based on the out-of-tolerance area;

[0159] The index value determination module 803 is used to determine the roughness value, waviness value and surface profile value corresponding to the target surface of the workpiece to be inspected if the workpiece to be inspected is qualified;

[0160] An OBB bounding box construction module 804 is used to construct an OBB bounding box of the workpiece to be inspected, and determine a target coordinate system corresponding to the target surface from multiple two-dimensional coordinate systems of the OBB bounding box;

[0161] The point cloud projection module 805 is used to project the target point cloud in the target surface into the target coordinate system to obtain the two-dimensional coordinates of the projected target point cloud;

[0162] An index selection module 806, used to determine at least two target evaluation indexes from the roughness value, the waviness value and the surface profile value;

[0163] The quality feature construction module 807 is used to obtain a multidimensional feature tensor corresponding to the target point cloud based on the two-dimensional coordinates of the projected target point cloud and at least two target evaluation indicators corresponding to the target point cloud; the multidimensional feature tensor is used to characterize the quality characteristics of the target point cloud;

[0164] The comprehensive quality assessment module 808 is used to input the multi-dimensional feature tensor corresponding to the target point cloud into a preset quality assessment model to predict the quality level of the target surface.

[0165] Optionally, the point cloud registration module 801 includes:

[0166] A coarse registration submodule is used to coarsely register the 3D point cloud with the standard CAD model corresponding to the workpiece to be inspected;

[0167] The fine registration submodule is used to optimize and adjust the result of the rough registration to obtain the position of the three-dimensional point cloud after fine registration;

[0168] The limit model determination submodule is used to determine the maximum simulation model and the minimum simulation model corresponding to the workpiece to be inspected based on the preset tolerance range and the standard CAD model;

[0169] The registration submodule is used to determine the point cloud in the three-dimensional point cloud that exceeds the maximum simulation model or the minimum simulation model as an out-of-tolerance point cloud, and determine the out-of-tolerance area of ​​the workpiece to be inspected based on the out-of-tolerance point cloud.

[0170] Optionally, the out-of-tolerance detection module 802 includes:

[0171] The test submodule is used to obtain the test data of the out-of-tolerance area in the simulation test or physical test scenario;

[0172] The judgment submodule is used to determine whether the workpiece to be inspected is qualified based on the test data.

[0173] Optionally, when it is determined that the workpiece to be inspected is unqualified, the device further comprises:

[0174] The backtracking module is used to trace back the process parameters of the workpiece to be tested during production based on the test data;

[0175] The analysis module is used to determine the cause of the deviation based on the process parameters and generate alarm information based on the cause of the deviation; the alarm information is used to remind the staff to optimize and adjust the process parameters.

[0176] Optionally, after obtaining the three-dimensional point cloud corresponding to the workpiece to be inspected, the device further includes:

[0177] A region division module is used to determine a plurality of continuous smooth regions and discontinuous smooth regions in the workpiece to be inspected according to the curvature or normal change rate of the three-dimensional point cloud;

[0178] Among them, the continuous smooth area is used to reflect the quality characteristics of the roughness and waviness of the surface of the workpiece to be inspected, and the deformation information in the discontinuous smooth area is used to reflect the deformation error characteristics of the surface of the workpiece to be inspected, and is used to provide a detection direction for determining the out-of-tolerance area.

[0179] Optionally, the indicator value determination module 803 includes:

[0180] The filtering submodule is used to perform multi-scale filtering on any continuous smooth area on the target surface to obtain a high-frequency feature point cloud and a low-frequency feature point cloud corresponding to the continuous smooth area;

[0181] A first division submodule is used to divide the high-frequency feature point cloud into regions based on a first minimum adjacent distance to obtain a plurality of first subregions, and to fit the plurality of high-frequency feature point clouds in the first subregions to obtain a first reference surface corresponding to each first subregion;

[0182] A roughness determination submodule, for determining a roughness value corresponding to each first sub-region based on the high-frequency feature point cloud and the first reference surface; wherein in the first sub-region, the roughness values ​​of all high-frequency feature point clouds are the same;

[0183] A second division submodule is used to divide the low-frequency feature point cloud into regions based on the second minimum adjacent distance to obtain a plurality of second sub-regions, and to fit the plurality of low-frequency feature point clouds in the second sub-regions to obtain a second reference surface corresponding to each second sub-region; wherein the first minimum adjacent distance is smaller than the second minimum adjacent distance;

[0184] The waviness determination submodule is used to determine the waviness value corresponding to each second sub-region based on the low-frequency feature point cloud and the second reference surface; wherein, in the second sub-region, the waviness values ​​of all low-frequency feature point clouds are the same.

[0185] Optionally, the indicator value determination module 803 further includes:

[0186] The distance calculation submodule is used to determine the Euclidean distance between any point cloud and the corresponding point in the standard CAD model based on the result of the precise registration of the three-dimensional point cloud with the standard CAD model corresponding to the workpiece to be inspected;

[0187] A region division submodule, used for dividing the continuous smooth region into regions based on the second minimum adjacent distance to obtain a plurality of third subregions, wherein any third subregion includes a plurality of high-frequency feature point clouds and / or low-frequency feature point clouds;

[0188] The surface contour calculation submodule is used to determine the mean of the Euclidean distances of all point clouds in any third subregion, and use the mean as the surface contour value of the third subregion; wherein, in the third subregion, the surface contour values ​​of all point clouds are the same.

[0189] Optionally, when the target evaluation index is a roughness value, a waviness value, and a surface profile value, the quality feature construction module 807 includes:

[0190] The index mapping submodule is used to determine the roughness value, waviness value and surface profile value corresponding to the target point cloud as the z value of different channels of the target point cloud;

[0191] The three-dimensional feature tensor construction submodule is used to obtain the three-dimensional feature tensor (x, y, 3) corresponding to the target point cloud based on the two-dimensional coordinates (x, y) and z value after projection of the target point cloud;

[0192] Among them, x represents the horizontal position of the target point cloud after projection, y represents the vertical position of the target point cloud after projection, and z represents the channel value of the target point cloud.

[0193] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0194] On the third aspect, based on the same inventive concept, an embodiment of the present application provides a storage medium, which stores machine executable instructions. When the machine executable instructions are executed by a processor, the auxiliary processing quality assessment method based on three-dimensional point cloud data proposed in the first aspect of the present application is implemented.

[0195] It should be noted that the specific implementation of the storage medium of the embodiment of the present application refers to the specific implementation of the auxiliary processing quality assessment method based on three-dimensional point cloud data proposed in the first aspect of the embodiment of the present application, and will not be repeated here.

[0196] In the fourth aspect, based on the same inventive concept, refer to Fig. 9 An embodiment of the present application provides an electronic device 900, including a processor 901 and a memory 902; the memory 902 stores machine executable instructions that can be executed by the processor 901, and the processor 901 is used to execute the machine executable instructions to implement the auxiliary processing quality assessment method based on three-dimensional point cloud data as proposed in the first aspect of the present application.

[0197] It should be noted that the specific implementation of the electronic device 900 in the embodiment of the present application refers to the specific implementation of the auxiliary processing quality assessment method based on three-dimensional point cloud data proposed in the first aspect of the embodiment of the present application, and will not be repeated here.

[0198] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0199] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, devices, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0200] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0201] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0202] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0203] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0204] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0205] The above is a detailed introduction to the auxiliary processing quality assessment method, medium and equipment based on three-dimensional point cloud data provided by the present application. This article uses specific examples to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for evaluating auxiliary machining quality based on three-dimensional point cloud data, characterized in that: The method comprises: Acquire a three-dimensional point cloud corresponding to the workpiece to be detected, and compare the three-dimensional point cloud with a limit model corresponding to the workpiece to be detected to determine an out-of-tolerance area of ​​the workpiece to be detected; the limit model is a simulation design model corresponding to the workpiece to be detected constructed based on a preset tolerance range, and the out-of-tolerance area is composed of point clouds in the three-dimensional point cloud that exceed the range of the limit model; Based on the out-of-tolerance area, determining whether the workpiece to be inspected is qualified; In the case that the workpiece to be inspected is qualified, the roughness value, waviness value and surface profile value corresponding to the target surface of the workpiece to be inspected are determined; wherein the roughness value reflects the slight unevenness of the target surface of the workpiece to be inspected, the waviness value reflects the difference in the height of the target surface fluctuations, and the surface profile value reflects the shape accuracy of the target surface; Constructing an OBB bounding box of the workpiece to be inspected, and determining a target coordinate system corresponding to the target surface from a plurality of two-dimensional coordinate systems of the OBB bounding box; Projecting the target point cloud on the target surface into the target coordinate system to obtain the two-dimensional coordinates of the projected target point cloud; Determine at least two target evaluation indicators from the roughness value, the waviness value and the surface profile value; Based on the two-dimensional coordinates of the projected target point cloud and at least two target evaluation indicators corresponding to the target point cloud, a multidimensional feature tensor corresponding to the target point cloud is obtained; the multidimensional feature tensor is used to characterize the quality characteristics of the target point cloud; Inputting the multidimensional feature tensor corresponding to the target point cloud into a preset quality assessment model to predict the quality level of the target surface; In the case where the target evaluation index is a roughness value, a waviness value, and a surface profile value, a multidimensional feature tensor corresponding to the target point cloud is obtained based on the two-dimensional coordinates after the projection of the target point cloud and at least two of the target evaluation indexes corresponding to the target point cloud, including: Determine the roughness value, the waviness value, and the surface profile value corresponding to the target point cloud as z values ​​of different channels of the target point cloud; Based on the projected two-dimensional coordinates (x, y) of the target point cloud and the z value, a three-dimensional feature tensor (x, y, 3) corresponding to the target point cloud is obtained; Among them, x represents the horizontal position of the target point cloud after projection, y represents the vertical position of the target point cloud after projection, and z represents the channel value of the target point cloud.

2. The method according to claim 1, characterized in that: Comparing the three-dimensional point cloud with the limit model corresponding to the workpiece to be inspected to determine the out-of-tolerance area of ​​the workpiece to be inspected, including: Roughly aligning the three-dimensional point cloud with a standard CAD model corresponding to the workpiece to be inspected; Optimizing and adjusting the result of the rough registration to obtain the position of the three-dimensional point cloud after fine registration; Based on the preset tolerance range and the standard CAD model, determining a maximum simulation model and a minimum simulation model corresponding to the workpiece to be inspected; The point cloud in the three-dimensional point cloud that exceeds the maximum simulation model or the minimum simulation model is determined as an out-of-tolerance point cloud, and an out-of-tolerance area of ​​the workpiece to be inspected is determined based on the out-of-tolerance point cloud.

3. The method according to claim 1, characterized in that Based on the out-of-tolerance area, determining whether the workpiece to be inspected is qualified includes: Acquire test data of the out-of-tolerance area in a simulation test or physical test scenario; Based on the test data, determine whether the workpiece to be inspected is qualified.

4. The method according to claim 3, characterized in that: In the case where it is determined that the workpiece to be inspected is unqualified, the method further includes: Based on the test data, tracing back the process parameters in the production process of the workpiece to be tested; Based on the process parameters, the reasons for the deviation are determined, and based on the reasons for the deviation, alarm information is generated; the alarm information is used to remind the staff to optimize and adjust the process parameters.

5. The method according to claim 1, characterized in that After obtaining the three-dimensional point cloud corresponding to the workpiece to be inspected, the method further includes: Determining a plurality of continuous smooth regions and discontinuous smooth regions in the workpiece to be inspected according to the curvature or normal change rate of the three-dimensional point cloud; Among them, the continuous smooth area is used to reflect the quality characteristics of the roughness and waviness of the surface of the workpiece to be inspected, and the deformation information in the discontinuous smooth area is used to reflect the deformation error characteristics of the surface of the workpiece to be inspected, and is used to provide a detection direction for determining the out-of-tolerance area.

6. The method according to claim 5, characterized in that In the case that the workpiece to be inspected is qualified, determining the roughness value and the waviness value corresponding to the target surface of the workpiece to be inspected includes: For any of the continuous smooth regions in the target surface, multi-scale filtering is performed on the continuous smooth region to obtain a high-frequency feature point cloud and a low-frequency feature point cloud corresponding to the continuous smooth region; Dividing the high-frequency feature point cloud into regions based on a first minimum adjacent distance to obtain a plurality of first sub-regions, and fitting the plurality of high-frequency feature point clouds in the first sub-regions to obtain a first reference surface corresponding to each of the first sub-regions; Based on the high-frequency feature point cloud and the first reference surface, determining a roughness value corresponding to each of the first sub-regions; wherein in the first sub-regions, the roughness values ​​of all high-frequency feature point clouds are the same; Dividing the low-frequency feature point cloud into regions based on a second minimum adjacent distance to obtain a plurality of second sub-regions, and fitting the plurality of low-frequency feature point clouds in the second sub-regions to obtain a second reference surface corresponding to each of the second sub-regions; wherein the first minimum adjacent distance is smaller than the second minimum adjacent distance; Based on the low-frequency feature point cloud and the second reference surface, a waviness value corresponding to each of the second sub-regions is determined; wherein, in the second sub-regions, the waviness values ​​of all the low-frequency feature point clouds are the same.

7. The method according to claim 6, characterized in that In the case that the workpiece to be inspected is qualified, determining a surface profile value corresponding to a target surface of the workpiece to be inspected comprises: Based on the result of precise registration of the three-dimensional point cloud with the standard CAD model corresponding to the workpiece to be inspected, determining the Euclidean distance between any of the point clouds and the corresponding point in the standard CAD model; Dividing the continuous smooth area into regions based on the second minimum adjacent distance to obtain a plurality of third sub-regions, any of the third sub-regions including a plurality of the high-frequency feature point clouds and / or the low-frequency feature point clouds; For any of the third sub-regions, determine the mean of the Euclidean distances of all point clouds in the third sub-region, and use the mean as the surface contour value of the third sub-region; wherein, in the third sub-region, the surface contour values ​​of all point clouds are the same.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the auxiliary processing quality assessment method based on three-dimensional point cloud data as described in any one of claims 1 to 7 is implemented.

9. An electronic device 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, it implements the auxiliary processing quality assessment method based on three-dimensional point cloud data as described in any one of claims 1 to 7.

Images