CN117370592B - Part similarity recognition method based on machine learning - Google Patents
Part similarity recognition method based on machine learning Download PDFInfo
- Publication number
- CN117370592B CN117370592B CN202311304978.2A CN202311304978A CN117370592B CN 117370592 B CN117370592 B CN 117370592B CN 202311304978 A CN202311304978 A CN 202311304978A CN 117370592 B CN117370592 B CN 117370592B
- Authority
- CN
- China
- Prior art keywords
- data
- point cloud
- points
- point
- cloud data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000010801 machine learning Methods 0.000 title claims abstract description 16
- 239000013598 vector Substances 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 7
- 238000004891 communication Methods 0.000 claims abstract description 4
- 238000012545 processing Methods 0.000 claims description 19
- 238000013075 data extraction Methods 0.000 claims description 14
- 238000011156 evaluation Methods 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000001174 ascending effect Effects 0.000 claims description 5
- 238000000513 principal component analysis Methods 0.000 claims description 5
- 238000013499 data model Methods 0.000 claims description 3
- 238000012356 Product development Methods 0.000 abstract description 7
- 238000011161 development Methods 0.000 abstract description 5
- 238000011960 computer-aided design Methods 0.000 description 7
- 238000004519 manufacturing process Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000007405 data analysis Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000008676 import Effects 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012827 research and development Methods 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/53—Querying
- G06F16/532—Query formulation, e.g. graphical querying
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5854—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using shape and object relationship
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Library & Information Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Multimedia (AREA)
- Complex Calculations (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a part similarity identification method based on machine learning, which comprises the following steps: analyzing the data file through the on-off source data exchange library, and acquiring the local geometric feature key point data; forming the local geometric feature key point data communication graph and acquiring a space coordinate value of a node; converting the node space coordinate values into point cloud data for data training; obtaining a feature vector according to the point cloud data; searching the most similar model in a database according to the feature vector; the method can rapidly select the most similar model on the premise of not using commercial software, meet the rapid searching requirement for parts in the daily product development process, integrate the existing data of enterprises, rapidly develop new products by using the existing product data, and shorten the development period.
Description
Technical Field
The invention relates to the field of industrial product development, in particular to a part similarity identification method based on machine learning.
Background
Part matching identification technology is a technique for determining the degree of similarity or matching between two or more parts (typically parts or components in engineering or manufacturing); the application fields of this technology include industrial product development, engineering design, CAD (computer aided design), manufacturing, part replacement, fitting matching, and the like.
The existing technology for part matching identification mainly relies on commercial software for model matching, and the efficiency is poor due to single-dimension evaluation according to indexes such as part volume, hole number and surface number.
Not only does the cost of the software increase for the user, but it is also necessary to import the model into the business software for storage in a format recognizable by the software. The efficiency is low when a large number of parts are processed, and the working requirements cannot be met.
In addition, the existing CAD data of the enterprise cannot be effectively utilized for data integration in the prior art, which causes trouble in rapid development of new products and shortening of the development period. Therefore, a new technical solution is needed to improve the efficiency and precision of matching parts, and meanwhile, the existing data resources can be effectively utilized to quickly find the data most suitable for the current product from the previous data.
Disclosure of Invention
The invention provides a part similarity identification method based on machine learning, which is used for solving the technical defects that commercial software is relied on, single-dimensional evaluation, data import and format conversion are difficult and data integration is not facilitated in the prior art, and adopts the following technical scheme:
a part similarity identification method based on machine learning comprises the following steps:
s1, analyzing a data file through an open source data exchange library, and acquiring local geometric feature key point data;
S2, forming the local geometric feature key point data communication graph and acquiring a space coordinate value of a node;
S3, converting the node space coordinate values into point cloud data for data training;
S4, obtaining a feature vector according to the point cloud data;
And S5, finding the most similar model in the database according to the feature vector and different technical indexes.
Preferably, the data file is a part three-dimensional data file, typically an STP format data file.
Preferably, the local geometric feature key data comprises geometric key discrete points;
Forming a connected graph by combining the geometric key characteristic discrete points through a triangular mesh algorithm and performing triangle subdivision;
The space coordinate value of the node is obtained by a space triangulation base method;
the triangular mesh algorithm is a Delaunay algorithm, and the goal is to divide a given point set into non-overlapping triangles, so that the circumscribed circle of each triangle does not contain any other points;
The spatial triangulation library is specifically "scipy.spatial.Delaunay", which is a module in the Python scientific calculation library SciPy for performing Delaunay triangulation operation; delaunay triangulation is a geometric operation that divides a set of points into non-overlapping triangles, wherein the vertices of each triangle are points in the original set of points and satisfy the Delaunay property that no other points are contained within the circumscribed circle of each triangle;
the scipy.spatial.Delaunay module allows the coordinates of a set of points to be entered and generates Delaunay triangulation associated with those points.
Preferably, the point cloud data is converted by a point cloud object in a point cloud library;
The point cloud library is specifically a 'pyntcloud' library, pyntCloud is a Python library for processing point cloud data; a point cloud is a three-dimensional dataset made up of a large number of points, typically used to represent a surface or scene of an object in the real world.
Through the point cloud object of pyntcloud libraries, the unit node data are converted into point cloud data in PCD format for data training, wherein the PCD format is a point cloud data format and is commonly used for storing and processing 3D point cloud data.
Preferably, the point cloud data is subjected to dimension reduction through a principal component analysis algorithm, noise and unnecessary features are removed, and feature vectors are obtained;
The principal component analysis algorithm is a PCA algorithm, and specifically is a data dimension reduction algorithm: the main idea of PCA is to map n-dimensional features onto k-dimensions, which are completely new orthogonal features, also called principal components, and are k-dimensional features reconstructed on the basis of original n-dimensional features; PCA works by sequentially finding a group of mutually orthogonal coordinate axes from the original space, and the selection of the new coordinate axes is closely related to the data; the first new coordinate axis is selected to be the direction with the maximum variance in the original data, the second new coordinate axis is selected to be the plane orthogonal to the first coordinate axis so as to make the variance maximum, and the third axis is selected to be the plane orthogonal to the first axis and the second axis so as to make the variance maximum; by analogy, n such coordinate axes can be obtained; obtaining new coordinate axes in this way, most of the variances are contained in the previous k coordinate axes, and the variances contained in the latter coordinate axes are almost 0; thus, the remaining coordinate axes can be ignored, and only the previous k coordinate axes containing most variance are reserved; in fact, this amounts to retaining only dimensional features containing a substantial portion of variance, while ignoring feature dimensions containing variances of almost 0, achieving dimension reduction of the data features.
Preferably, the feature vector is compared with a model of a database through a point cloud data extraction algorithm, euclidean distance, minimum enveloping volume and evaluation indexes of average distances of key points matched between objects to obtain a similar model;
Preferably, the point cloud data extraction algorithm specifically includes:
importing an open source library for three-dimensional data processing;
Loading the point cloud data;
determining each point and other points in the neighborhood around the point;
calculating the geometric relationship between the neighborhood points and the center point;
summarizing the geometric relationships and creating a histogram;
For each point, extracting a characteristic of the point cloud data as the histogram, wherein the histogram describes local geometric characteristics around the point;
The point cloud data extraction algorithm is specifically 'FPFH', is a feature descriptor for 3D point cloud matching, and can calculate geometric information and normal vector information around each point so as to describe the features of the points.
Preferably, the algorithm of the minimum envelope volume specifically comprises: importing an open source library for three-dimensional data processing; regarding points in the point cloud data as a set in three-dimensional space, calculating a minimum convex hull capable of surrounding the points;
the convex hull is a polygon with convexity, and the boundary of the convex hull comprises all points in the point cloud;
Preferably, ascending order is carried out on the result values returned by the point cloud data extraction algorithm, euclidean distance, minimum enveloping volume and average distance calculation of the matched key points between objects, and when the minimum values are the same model, the data model is represented to be most similar to the input model; and when the minimum values are not uniform, returning the model with the minimum Euclidean distance as the most similar part.
The invention has the following beneficial effects:
Data analysis and feature extraction improvement: the three-dimensional data file (usually in STP format) of the part is analyzed by using an open source data exchange library, and the local geometric feature key point data is extracted by using a Delaunay algorithm and other methods, so that the data processing is more efficient and accurate.
And D, converting point cloud data: converting the spatial coordinate values of the nodes into a point cloud data format facilitates better processing and analysis of the part data, making the data easier to train and match.
Feature vector generation: the PCA algorithm is used for reducing the dimension and removing the noise, and representative feature vectors are generated, so that the matching precision is improved, and the influence of the noise is reduced.
Multidimensional matching evaluation: by introducing indexes such as Euclidean distance, minimum envelope volume, average distance of matched key points among objects and the like, and a point cloud data extraction algorithm and FPFH (field programmable gate array) characteristics, multidimensional matching evaluation is realized, and matching precision is improved, so that the similarity among parts is better reflected.
Reducing reliance on business software: the method uses an open source library and an algorithm, reduces the dependence on expensive commercial software, reduces the cost and improves the flexibility.
Using existing CAD data: the improved method can more effectively utilize the existing CAD data of enterprises, and quicken the new product development and research and development period, as the past design and model data can be better integrated and utilized.
In general, the improved part similarity recognition method improves matching efficiency and accuracy, reduces cost, better utilizes existing data resources, and is helpful for finding data most suitable for current products more quickly. This has potentially important application value for part matching and identification in the manufacturing and engineering fields.
The method can rapidly select the corresponding model on the premise of not using commercial software, meet the rapid matching requirement of parts in the daily product development process, integrate the existing data of enterprises, rapidly develop new products by using the existing product data, and shorten the development period.
Drawings
FIG. 1is a schematic flow diagram of a machine learning-based part similarity recognition method;
FIG. 2 is a schematic illustration of the geometry of a stringer of example 2 of a machine learning based method of part similarity identification;
FIG. 3 is a triangle unit subdivision effect diagram of a stringer of example 2 of a machine learning based part similarity recognition method;
FIG. 4 is a point cloud visualization effect diagram of embodiment 2 of a machine learning-based part similarity identification method;
Fig. 5 is a comparison chart of sampled point cloud data of embodiment 2 of a machine learning-based part similarity recognition method.
Detailed Description
The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.
1. Example 1
A part similarity identification method based on machine learning comprises the following steps:
s1, analyzing a data file through an open source data exchange library, and acquiring local geometric feature key point data;
S2, forming the local geometric feature key point data communication graph and acquiring a space coordinate value of a node;
S3, converting the node space coordinate values into point cloud data for data training;
S4, obtaining a feature vector according to the point cloud data;
And S5, finding the most similar model in the database according to the feature vector and different technical indexes.
Specifically, the data file is a three-dimensional data file of a part, typically an STP format data file.
Specifically, the local geometric feature key data comprises geometric key discrete points;
Forming a connected graph by combining the geometric key characteristic discrete points through a triangular mesh algorithm and performing triangle subdivision;
The space coordinate value of the node is obtained by a space triangulation base method;
the triangular mesh algorithm is a Delaunay algorithm, and the goal is to divide a given point set into non-overlapping triangles, so that the circumscribed circle of each triangle does not contain any other points;
The spatial triangulation library is specifically "scipy.spatial.Delaunay", which is a module in the Python scientific calculation library SciPy for performing Delaunay triangulation operation; delaunay triangulation is a geometric operation that divides a set of points into non-overlapping triangles, wherein the vertices of each triangle are points in the original set of points and satisfy the Delaunay property that no other points are contained within the circumscribed circle of each triangle;
the scipy.spatial.Delaunay module allows the coordinates of a set of points to be entered and generates Delaunay triangulation associated with those points.
Specifically, the point cloud data is converted by a point cloud object in a point cloud library;
The point cloud library is specifically a 'pyntcloud' library, pyntCloud is a Python library for processing point cloud data; a point cloud is a three-dimensional dataset made up of a large number of points, typically used to represent a surface or scene of an object in the real world.
Through the point cloud object of pyntcloud libraries, the unit node data are converted into point cloud data in PCD format for data training, wherein the PCD format is a point cloud data format and is commonly used for storing and processing 3D point cloud data.
Specifically, the dimension of the point cloud data is reduced through a principal component analysis algorithm, noise and unnecessary features are removed, and feature vectors are obtained;
the principal component analysis algorithm is a PCA algorithm, and specifically is a data dimension reduction algorithm: the main idea of PCA is to map n-dimensional features onto k-dimensions, which are completely new orthogonal features, also called principal components, and are k-dimensional features reconstructed on the basis of original n-dimensional features; PCA works by sequentially finding a group of mutually orthogonal coordinate axes from the original space, and the selection of the new coordinate axes is closely related to the data; the first new coordinate axis is selected to be the direction with the maximum variance in the original data, the second new coordinate axis is selected to be the plane orthogonal to the first coordinate axis so as to make the variance maximum, and the third axis is selected to be the plane orthogonal to the first axis and the second axis so as to make the variance maximum; by analogy, n such coordinate axes can be obtained; the new axes obtained in this way, most of the variance is contained in the first k axes, and the latter axes contain variances of almost 0; thus, the remaining coordinate axes can be ignored, and only the previous k coordinate axes containing most variance are reserved; this is equivalent to only preserving dimension features containing a substantial portion of variance, and ignoring feature dimensions containing variances of nearly 0, achieving dimension reduction processing of the data features.
Specifically, comparing the feature vector with a model of a database through a point cloud data extraction algorithm, euclidean distance, minimum enveloping volume and evaluation indexes of average distances of key points matched between objects to obtain a similar model;
The Euclidean distance can be found by the following formula:
Euclidean distance, also known as Euclidean distance, is a commonly used distance metric for measuring the absolute distance between two points in a multidimensional space. In an n-dimensional european space, the space is a set of points, each point in the space can be represented as (X (1), X (2), X (n)), where X (i) (i=1, 2, n) is the i-th coordinate of point X. The expression of point a is (a (1), a (2), a (i), a (n)), (i=1, 2, n); the expression of point B is (B (1), B (2), B (i), B (n)), (i=1, 2, n).
The average distance is determined by the following equation:
Assuming that there are two parts, part a, part B, each with n points, then:
The formula shows that the average distance between two parts is calculated, firstly, one key point of the first part is sequentially selected, euclidean distances between the key point and all key points of the second part are calculated (the calculation formula refers to the Euclidean distance formula), and the minimum value is taken; the average of these minima is then calculated and recorded as the average distance.
Specifically, the point cloud data extraction algorithm specifically includes:
importing an open source library for three-dimensional data processing;
Loading the point cloud data;
Determining other points in the neighborhood around each point;
calculating the geometric relationship between the neighborhood points and the center point;
summarizing the geometric relationships and creating a histogram;
For each point, extracting a characteristic of the point cloud data as the histogram, wherein the histogram describes local geometric characteristics around the point;
The point cloud data extraction algorithm is specifically 'FPFH', is a feature descriptor for 3D point cloud matching, and can calculate geometric information and normal vector information around each point so as to describe the features of the points.
Specifically, the algorithm for minimum envelope volume specifically includes: importing an open source library for three-dimensional data processing; regarding points in the point cloud data as a set in three-dimensional space, calculating a minimum convex hull capable of surrounding the points;
the convex hull is a polygon with convexity, and the boundary of the convex hull comprises all points in the point cloud;
The minimum envelope volume is calculated using the following formula:
λ,V＝eig(H)；
OBB＝{C,V}；
Size＝max(Pi·V)-min(Pi·V)；
Where C is the centroid, n is the number of points, and P i is the position of each point; h is a covariance matrix, T represents a transpose of the matrix; lambda is the eigenvalue and V is the eigenvector; size is the size of the minimum envelope volume, representing the dot product.
Specifically, ascending order is carried out on the result values returned by the point cloud data extraction algorithm, euclidean distance, minimum enveloping volume and average distance calculation of the matched key points between objects, and when the minimum values are the same model, the data model is the most similar to the input model; and when the minimum values are not uniform, returning the model with the minimum Euclidean distance as the most similar part.
The beneficial effects of this embodiment are:
Data analysis and feature extraction improvement: the three-dimensional data file (usually in STP format) of the part is analyzed by using an open source data exchange library, and the local geometric feature key point data is extracted by using a Delaunay algorithm and other methods, so that the data processing is more efficient and accurate.
And D, converting point cloud data: converting the spatial coordinate values of the nodes into a point cloud data format facilitates better processing and analysis of the part data, making the data easier to train and match.
Feature vector generation: the PCA algorithm is used for reducing the dimension and removing the noise, and representative feature vectors are generated, so that the matching precision is improved, and the influence of the noise is reduced.
Multidimensional matching evaluation: by introducing indexes such as Euclidean distance, minimum envelope volume, average distance of matched key points among objects and the like, and a point cloud data extraction algorithm and FPFH (field programmable gate array) characteristics, multidimensional matching evaluation is realized, and matching precision is improved, so that similarity among parts is better reflected.
Reducing reliance on business software: the method uses an open source library and an algorithm, reduces the dependence on expensive commercial software, reduces the cost and improves the flexibility.
Using existing CAD data: the improved method can more effectively utilize the existing CAD data of enterprises, and quicken the new product development and research and development period, as the past design and model data can be better integrated and utilized at present.
In general, the improved part similarity recognition method improves matching efficiency and accuracy, reduces cost, better utilizes existing data resources, and is helpful for finding data most suitable for current products more quickly. This has potentially important application value for part matching and identification in the manufacturing and engineering fields;
the method can rapidly select the corresponding model on the premise of not using commercial software, meet the rapid matching requirement of parts in the daily product development process, integrate the existing data of enterprises, rapidly develop new products by using the existing product data, and shorten the development period.
2. Example two
The following is a specific implementation case of the system:
taking a front floor side member of an automobile as an example, the structure of the parts is illustrated in fig. 2.
Analyzing the STP format data file of the longitudinal beam by the OCC.extension.Dataexchange module, extracting the entity objects and the attribute, geometry and topological structure thereof contained in the STEP file and the relation information among the entity objects (the STEP file content is shown in the figure I), carrying out model comparison on key feature points which only need to embody geometric features, extracting return information, and storing the extracted key point coordinate information in a list A(-1828.455;509.134;187.326,-1848.291;366.417;189.067,-1000.434;350.413;386.512,-998.120;451.410,451.410;………).
And (3) connecting the feature points of the list A obtained in the step (A) into a connected graph according to a Delaunay algorithm under the SciPy module. And according to the connected graph, triangle unit subdivision is carried out, and then the space coordinate values of the nodes of the generated triangle units are obtained through a scipy.spatial.Delaunay method and stored in a list B, wherein triangle unit subdivision effects of (-1583.083;398.204;171.385,-1807.343;479.814;188.634,-1817.997;398.486;188.911,-1524.188460.114;169.919). longitudinal beams are shown in a figure 3.
And converting the triangle unit nodes obtained in the step two into point cloud data in PCD format for data training through a PointCloud algorithm under a pyntcloud module. The point cloud visualization effect is shown in fig. 4.
And according to the generated point cloud data, performing dimension reduction on the data through PCA component analysis, removing noise and unimportant features, and acquiring local geometric feature key point data to obtain a representative feature vector so as to achieve the condition of reducing the calculated amount.
According to the obtained sampled point cloud data, the obtained point cloud data is compared with the existing data of the database, as shown in fig. 5, according to evaluation indexes such as FPFH, euclidean distance, object average distance, minimum envelope volume and the like, and a model of the database, a result value list is obtained C,{Model1：(FPFH：-0.574011,EuclideanDistance：0.651828,AverageDistance：4.925642,OBB：10.465350);Model2：(FPFH：-0.66923,EuclideanDistance：0.785394,AverageDistance：5.124459,OBB：8.798400)Model3：(……);Model4：
(… …) … …), And then ascending the target values, choosing the smallest value Model1 as the most similar part.
The beneficial effects of this embodiment are:
accurate data analysis, namely analyzing a longitudinal beam STP format data file by using an OCC.extension.Dataexchange module, and accurately extracting entity objects and attribute, geometric and topological structures thereof and relationship information among the entity objects in the file; such parsing not only provides detailed data, but also ensures the integrity and accuracy of the data.
The characteristic points of the list A can be extracted efficiently by using a Delaunay algorithm under SciPy modules, and the space coordinate values of the triangle units can be further generated; this method provides a structured way to capture and represent the geometric features of the part.
The flexible conversion of the data format is that triangle unit nodes are converted into point cloud data in PCD format through PointCloud algorithm under pyntcloud module, thus providing great convenience for subsequent data analysis and processing; this ensures that the data can exist in the most appropriate format at different processing stages.
Optimized data processing, namely, performing dimension reduction on the data by utilizing PCA component analysis so as to remove noise and unimportant characteristics; the optimization method not only improves the efficiency of data processing, but also ensures the accuracy of the result.
The comprehensive matching evaluation is carried out by combining various evaluation indexes such as FPFH, european distance, object average distance and minimum envelope volume, the parts can be evaluated from multiple dimensions, and the matching accuracy is improved; such multi-dimensional evaluation provides a solid basis for finding the best matching part.
And the automatic matching selection is that by carrying out ascending order sequencing on the target values, the system can automatically select the most similar parts, thereby greatly improving the working efficiency and reducing the error rate of manual selection.
Cost-effectiveness with such a system, enterprises can reduce human intervention and errors, improving production efficiency, and thus saving significant costs over a long period of time.
The system has good expansibility because of using an open source library and a module; with the advancement of technology or the advent of new needs, upgrades or adjustments can be easily made on an existing basis.
Claims (4)
1. A machine learning-based part similarity recognition method, comprising:
s1, analyzing a data file through an open source data exchange library, and acquiring local geometric feature key point data;
S2, forming the local geometric feature key point data communication graph and acquiring a space coordinate value of a node;
S3, converting the node space coordinate values into point cloud data for data training;
S4, obtaining a feature vector according to the point cloud data;
s5, finding the most similar model in a database according to the feature vector and different technical indexes;
The data file is a part three-dimensional data file;
the local geometric feature key point data comprises key geometric feature discrete points;
forming a connected graph by combining the key geometric feature discrete points through a triangular mesh algorithm and performing triangle subdivision;
The space coordinate value of the node is obtained by a space triangulation base method;
the point cloud data are converted by point cloud objects in a point cloud library;
Performing dimension reduction on the point cloud data through a principal component analysis algorithm, and removing noise and unnecessary features to obtain feature vectors;
And comparing the feature vector with a model of a database through a point cloud data extraction algorithm, euclidean distance, minimum enveloping volume and average distance evaluation index of matched key points among objects to obtain a similar model.
2. The machine learning-based part similarity recognition method of claim 1, wherein the point cloud data extraction algorithm specifically comprises:
importing an open source library for three-dimensional data processing;
Loading the point cloud data;
determining each point and other points in the neighborhood around the point;
calculating the geometric relationship between the neighborhood points and the center point;
summarizing the geometric relationships and creating a histogram, wherein the histogram is FPTH characteristic histogram;
For each point, the point cloud data extraction feature is the histogram, which characterizes local geometric features around the point.
3. The machine learning based part similarity recognition method of claim 2, wherein the algorithm for minimum envelope volume comprises:
importing an open source library for three-dimensional data processing;
Regarding points in the point cloud data as a set in three-dimensional space, calculating a minimum convex hull capable of surrounding the points;
the convex hull is a polygon with convexity, and the boundary of the convex hull comprises all points in the point cloud;
and calculating the convex hull volume.
4. The machine learning-based part similarity recognition method according to claim 3, wherein the result values returned by the point cloud data extraction algorithm, the euclidean distance, the minimum envelope volume and the average distance calculation of the matched key points between objects are sorted in ascending order, and when the minimum values are the same model, the data model is the most similar to the input model;
and when the minimum values are not uniform, returning the model with the minimum Euclidean distance as the most similar part.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311304978.2A CN117370592B (en) | 2023-10-10 | 2023-10-10 | Part similarity recognition method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311304978.2A CN117370592B (en) | 2023-10-10 | 2023-10-10 | Part similarity recognition method based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117370592A CN117370592A (en) | 2024-01-09 |
| CN117370592B true CN117370592B (en) | 2024-04-26 |
Family
ID=89393976
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311304978.2A Active CN117370592B (en) | 2023-10-10 | 2023-10-10 | Part similarity recognition method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117370592B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111914112A (en) * | 2020-07-07 | 2020-11-10 | 西安交通大学 | Part CAD model reusing method based on point cloud classification network |
| CN113901616A (en) * | 2021-10-27 | 2022-01-07 | 上海设序科技有限公司 | Part searching method and system |
| CN116309847A (en) * | 2023-04-28 | 2023-06-23 | 江苏大学 | Stacked workpiece pose estimation method based on combination of two-dimensional image and three-dimensional point cloud |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210183097A1 (en) * | 2017-11-13 | 2021-06-17 | Siemens Aktiengesellschaft | Spare Part Identification Using a Locally Learned 3D Landmark Database |
-
2023
- 2023-10-10 CN CN202311304978.2A patent/CN117370592B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111914112A (en) * | 2020-07-07 | 2020-11-10 | 西安交通大学 | Part CAD model reusing method based on point cloud classification network |
| CN113901616A (en) * | 2021-10-27 | 2022-01-07 | 上海设序科技有限公司 | Part searching method and system |
| CN116309847A (en) * | 2023-04-28 | 2023-06-23 | 江苏大学 | Stacked workpiece pose estimation method based on combination of two-dimensional image and three-dimensional point cloud |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117370592A (en) | 2024-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110866864B (en) | Face pose estimation/three-dimensional face reconstruction method and device and electronic equipment | |
| US8429174B2 (en) | Methods, systems, and data structures for performing searches on three dimensional objects | |
| CN110188228B (en) | Cross-modal retrieval method based on sketch retrieval three-dimensional model | |
| CN101567051B (en) | Image matching method based on characteristic points | |
| CN110599506B (en) | Point cloud segmentation method for three-dimensional measurement of complex special-shaped curved surface robot | |
| CN112651944B (en) | 3C component high-precision six-dimensional pose estimation method and system based on CAD model | |
| US20220189070A1 (en) | Machine-learning for 3d object detection | |
| CN106844620B (en) | View-based feature matching three-dimensional model retrieval method | |
| CN116596755B (en) | Method, device, equipment and storage medium for splicing point cloud data | |
| Kim et al. | Shape distribution-based retrieval of 3D CAD models at different levels of detail | |
| CN113111741A (en) | Assembly state identification method based on three-dimensional feature points | |
| CN117132630A (en) | Point cloud registration method based on second-order spatial compatibility measurement | |
| CN115600268A (en) | High-dimensional data set visualization method and device, electronic equipment and storage medium | |
| CN111597367B (en) | Three-dimensional model retrieval method based on view and hash algorithm | |
| CN112651408B (en) | Point-to-point transformation characteristic-based three-dimensional local surface description method and system | |
| CN117370592B (en) | Part similarity recognition method based on machine learning | |
| CN115588178B (en) | Automatic extraction method for high-precision map elements | |
| Dong et al. | Pose estimation of components in 3c products based on point cloud registration | |
| CN115147471A (en) | Laser point cloud automatic registration method based on curvature density characteristics | |
| CN115661218A (en) | Laser point cloud registration method and system based on virtual super point | |
| CN111862328B (en) | Three-dimensional grid segmentation result labeling method based on small samples | |
| Wuhrer et al. | Posture invariant correspondence of incomplete triangular manifolds | |
| CN114202662A (en) | Cutter characteristic point identification method and equipment combining transverse geometric characteristics of adjacent tool paths | |
| JP7054336B2 (en) | Modeling equipment, methods and programs | |
| Bespalov et al. | Local feature extraction using scale-space decomposition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |