CN113961738A - Multi-feature casting three-dimensional model retrieval method and device - Google Patents

Multi-feature casting three-dimensional model retrieval method and device Download PDF

Info

Publication number
CN113961738A
CN113961738A CN202111209575.0A CN202111209575A CN113961738A CN 113961738 A CN113961738 A CN 113961738A CN 202111209575 A CN202111209575 A CN 202111209575A CN 113961738 A CN113961738 A CN 113961738A
Authority
CN
China
Prior art keywords
casting
dimensional model
retrieved
operator
calculating
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.)
Pending
Application number
CN202111209575.0A
Other languages
Chinese (zh)
Inventor
计效园
孙晓龙
周建新
潘徐政
王先飞
李宝辉
殷亚军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huazhong University of Science and Technology
Shanghai Space Precision Machinery Research Institute
Original Assignee
Huazhong University of Science and Technology
Shanghai Space Precision Machinery Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Huazhong University of Science and Technology, Shanghai Space Precision Machinery Research Institute filed Critical Huazhong University of Science and Technology
Priority to CN202111209575.0A priority Critical patent/CN113961738A/en
Publication of CN113961738A publication Critical patent/CN113961738A/en
Pending legal-status Critical Current

Links

Images

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
    • G06F16/58—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
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T19/00—Manipulating 3D models or images for computer graphics
    • G06T19/20—Editing of 3D images, e.g. changing shapes or colours, aligning objects or positioning parts
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2219/00—Indexing scheme for manipulating 3D models or images for computer graphics
    • G06T2219/20—Indexing scheme for editing of 3D models
    • G06T2219/2016—Rotation, translation, scaling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2219/00—Indexing scheme for manipulating 3D models or images for computer graphics
    • G06T2219/20—Indexing scheme for editing of 3D models
    • G06T2219/2021—Shape modification

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Graphics (AREA)
  • Software Systems (AREA)
  • Library & Information Science (AREA)
  • Geometry (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Architecture (AREA)
  • Computer Hardware Design (AREA)
  • Processing Or Creating Images (AREA)

Abstract

The invention provides a method and a device for searching a multi-feature casting three-dimensional model, belonging to the field of casting three-dimensional model search, wherein the method comprises the following steps: acquiring a three-dimensional model of a casting to be retrieved, and performing triangular surface tiling on the three-dimensional model; calculating the shape characteristics and the domain knowledge of the casting to be retrieved; calculating the similarity between the shape characteristics of the casting to be retrieved and each model in the model library, and screening out a model with a certain proportion from the model library; calculating the similarity between the domain knowledge of the casting to be retrieved and the screened model, and screening out a three-dimensional model retrieval result; the domain knowledge comprises a hot spot characteristic, a symmetrical surface characteristic and an envelope size of the casting to be retrieved; the shape characteristics comprise geometric shape characteristics and concave-convex degree characteristics of the casting to be retrieved. The invention realizes the similar three-dimensional model retrieval robustly under the change of translation, rotation and the like of the casting, and simultaneously shortens the period in the aspect of process design compared with the retrieval and construction of a two-dimensional model due to the realization of the three-dimensional model retrieval.

Description

Multi-feature casting three-dimensional model retrieval method and device
Technical Field
The invention belongs to the field of casting three-dimensional model retrieval, and particularly relates to a multi-feature casting three-dimensional model retrieval method and device.
Background
The casting is widely applied to various fields of national economy such as aviation, aerospace, rail transit, engineering machinery and the like, wherein typical castings represented by a box body, a steering axle, a casing, an aerospace engine blade and the like have complex process structures such as multi-dimensional distortion, a special-shaped curved surface and the like, so that the problems of high process design difficulty, long period and the like exist, and the realization of 'mature process reuse' is one of keys for solving the problems.
At present, casting enterprises usually store data of mature casting process design schemes of a pouring system, a riser, a chill and the like of historical casting products in casting process cards in the forms of characters, pictures and the like, and the process cards can only be inquired by retrieving information such as product numbers, batch numbers and the like in the retrieving process. When the process design of a new casting product is carried out, the similarity between the product and a certain historical batch and a certain model of product is judged based on the experience of designers, and then the batch number and the product number of the batch of product are used for searching the historical mature process. Obviously, the process of 'new product-manual judgment-batch/product number-historical product mature process' greatly depends on knowledge storage of designers and familiarity of historical products, the problems of long retrieval process, low efficiency, high dependence on manual experience, low intelligent degree and the like exist, the defects of low retrieval result query accuracy, few query results and the like are caused, and the current situations of great difficulty and long period of complex casting process design of boxes, steering axles and the like are difficult to improve through 'mature process reuse'.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide a multi-feature casting three-dimensional model retrieval method and a multi-feature casting three-dimensional model retrieval device, and aims to solve the problems that the existing casting retrieval method is mostly based on two-dimensional retrieval and realizes process reuse, but the result after two-dimensional retrieval needs to be rebuilt into a three-dimensional model, so that the retrieval result obtained by the existing casting retrieval method has higher difficulty and longer period in the aspect of process design.
In order to achieve the above object, in one aspect, the present invention provides a method for retrieving a multi-feature casting three-dimensional model, comprising the following steps:
acquiring a three-dimensional model of a casting to be retrieved, and performing triangular surface tiling and posture normalization processing on the three-dimensional model;
calculating the shape characteristics and the domain knowledge of the casting to be retrieved based on the three-dimensional model of the casting to be retrieved after the triangular surface tiling;
calculating the similarity between the shape characteristics of the casting to be retrieved and each model in the model library, and screening out a model with a certain proportion from the model library;
calculating the similarity between the domain knowledge of the casting to be retrieved and the screened model, and screening out a three-dimensional model retrieval result;
the domain knowledge comprises a hot spot characteristic, a symmetrical surface characteristic and an envelope size of the casting to be retrieved; the shape characteristics comprise geometric shape characteristics and concave-convex degree characteristics of the casting to be retrieved.
Preferably, the shape features include the D2 operator, the N2 operator, and the NaN operator of the three-dimensional model;
the calculation method of the D2 operator comprises the following steps:
randomly selecting a plurality of point pairs on the surface of the casting to be retrieved, calculating the distance between two points of each point pair to obtain an array containing the distances of all the point pairs, and normalizing the array;
setting a statistical interval, and counting the times of the point pair distances in the array after normalization, wherein the formed characteristic vector is a D2 operator of the casting to be retrieved;
the calculation method of the N2 operator comprises the following steps:
randomly selecting a plurality of patch pairs on a three-dimensional model of a casting to be retrieved, calculating cosine values of included angles of normal vectors of two patches of each patch pair, and obtaining an array containing all the cosine values;
setting a statistical interval, counting the number of times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as an N2 operator of the casting to be retrieved;
the calculation method of the NaN operator comprises the following steps:
finding all adjacent patches of each triangular patch by taking patch adjacent points as clues;
randomly selecting a plurality of surface patches on a three-dimensional model of a casting to be retrieved, and calculating a cosine value of an included angle between each surface patch and a normal vector of an adjacent surface patch;
averaging the cosine values to obtain an array containing all the average cosine values;
and setting a statistical interval, counting the times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as a NaN operator of the casting to be retrieved.
Preferably, the domain knowledge comprises the Mod operator, the Sym operator and the Env operator of the three-dimensional model;
the method for acquiring the Env operator comprises the following steps:
determining X, Y and the maximum value and the minimum value of the three-dimensional model on the Z axis under a Cartesian coordinate system by traversing all triangular patch vertexes of the three-dimensional model of the casting to be retrieved, and obtaining the envelope size of the three-dimensional model of the casting to be retrieved along the direction of the coordinate main axis;
calculating the combination of the width-length ratio, the width-height ratio and the height-length ratio of the envelope size to obtain a three-dimensional feature vector, wherein the three-dimensional feature vector is an Env operator;
the method for acquiring the Mod operator comprises the following steps:
taking the sum of the triangular areas of the castings to be retrieved as the heat dissipation area of the three-dimensional models of the castings to be retrieved;
dividing the volume of the three-dimensional model of the casting to be retrieved by the heat dissipation area to obtain a Mod operator;
the method for acquiring the Sym operator comprises the following steps:
comparing the relation between the gravity center of each surface patch of the casting to be retrieved and the coordinate plane, and dividing the three-dimensional model of the casting to be retrieved along the YOZ plane, the XOZ plane and the XOY plane respectively;
and calculating surface area errors of two parts of the three-dimensional model after being divided along each coordinate plane as the symmetry degree of the coordinate plane, and sequentially calculating the symmetry degrees of the three coordinate planes to form a characteristic vector of the three-dimensional model, wherein the characteristic vector is a Sym operator.
Preferably, the method for posture normalization of the three-dimensional model of the casting to be retrieved comprises the following steps:
analyzing and reading a three-dimensional model file of a casting to be retrieved by utilizing three-dimensional modeling software;
acquiring a triangular tiling file of the three-dimensional model of the casting to be retrieved by using a three-dimensional model tiling function in software;
calculating the areas of all triangular patches on the casting to be retrieved according to a Helen formula;
calculating the centers of all triangular patches of the three-dimensional model of the casting to be retrieved according to a central coordinate formula, and further calculating the gravity center of the three-dimensional model of the casting to be retrieved;
calculating a translation matrix of the three-dimensional model of the casting to be retrieved according to the gravity centers of the triangular patches and the gravity center of the three-dimensional model of the casting to be retrieved;
calculating covariance matrixes of vertexes of all triangular surface patches of the three-dimensional model of the casting to be retrieved, and acquiring a rotation matrix of the casting to be retrieved;
and (4) acting the translation matrix and the rotation matrix on each vertex of the three-dimensional model of the casting to be retrieved to finish the posture normalization of the three-dimensional model.
Preferably, the NaN operator obtaining method includes the following steps:
s4.1: calculating the areas of all triangular patches on the three-dimensional model to be retrieved according to a Helen formula to obtain an ordered area array;
s4.2: searching all adjacent patches of each triangular patch under the condition of whether patch adjacent points exist or not, and creating a key-value dictionary according to a patch sequence number-adjacent patch sequence number set;
s4.3: searching a key-value dictionary, finding all adjacent patches of all triangular patches of the three-dimensional model of the casting to be retrieved, calculating cosine values of included angles of normal vectors of all patches and all adjacent patches of each patch, and averaging to obtain a list of cosine values of included angles of normal vectors;
s4.4: generating a plurality of random numbers in the ordered area array by using a random number function, determining subscripts of the random numbers in the area array by using binary search, and determining a plurality of sampling surface patches; inquiring the cosine value list of the normal vector included angle in S4.3 to obtain cosine values of the normal vector included angles of the plurality of sampling surface patches;
s4.5: and determining the number of the statistical intervals so as to determine the range of each statistical interval, counting the occurrence times of cosine values of the normal vector included angle of the surface patch in each interval range to obtain corresponding frequency distribution, and dividing the frequency distribution by the number of samples to be used as a NaN operator of the three-dimensional model.
Preferably, the method for acquiring the Sym operator specifically includes the following steps:
according to a gravity center coordinate formula, recalculating the gravity centers of all triangular patches of the three-dimensional model of the casting to be retrieved after the posture normalization;
dividing all the surface patches of the gradual three-dimensional model into two parts according to the position relation between the gravity center of each surface patch and the coordinate plane, and respectively calculating the area sum of the two parts of the surface patches;
obtaining surface area errors of two parts of the three-dimensional model according to the area sum of the two parts of the surface patches, and taking the surface area errors as the symmetry of the three-dimensional model relative to the coordinate plane;
combining the symmetries of the three coordinate surfaces to form a three-dimensional characteristic vector as a Sym operator of the three-dimensional model;
wherein the coordinate plane comprises a YOZ plane, an XOZ plane and an XOY plane.
In another aspect, a multi-feature casting three-dimensional model retrieval device comprises:
the three-dimensional model processing module is used for acquiring a three-dimensional model of the casting to be retrieved, and performing triangular surface tiling and posture normalization processing on the three-dimensional model;
the characteristic extraction module is used for calculating the shape characteristic and the domain knowledge of the casting to be retrieved based on the triangular tiled three-dimensional model of the casting to be retrieved; wherein the triangle patch is subjected to attitude normalization processing;
the model screening module is used for calculating the similarity between the shape characteristics of the casting to be retrieved and each model in the model library and screening out a model with a certain proportion in the model library;
the method is used for calculating the similarity between the field knowledge of the casting to be searched and the screened model, and screening out a three-dimensional model searching result;
the domain knowledge comprises a hot spot characteristic, a symmetrical plane characteristic and an envelope size of the casting to be retrieved; the shape characteristics comprise geometric shape characteristics and concave-convex degree characteristics of the casting to be retrieved.
Preferably, the shape features include the D2 operator, the N2 operator, and the NaN operator of the three-dimensional model;
the feature extraction module comprises a shape feature extraction unit and a domain knowledge extraction unit; the shape feature extraction unit comprises a D2 operator calculator, an N2 operator calculator and a NaN operator calculator;
the D2 operator calculator is used for acquiring a D2 operator, and the specific execution process is as follows:
randomly selecting a plurality of point pairs on the surface of the casting to be retrieved, calculating the distance between two points of each point pair to obtain an array containing the distances of all the point pairs, and normalizing the array;
setting a statistical interval, and counting the times of the point pair distances in the array after normalization, wherein the formed characteristic vector is a D2 operator of the casting to be retrieved;
the N2 operator calculator is used for acquiring an N2 operator, and the specific execution process is as follows:
randomly selecting a plurality of patch pairs on a three-dimensional model of a casting to be retrieved, calculating cosine values of included angles of normal vectors of two patches of each patch pair, and obtaining an array containing all the cosine values;
setting a statistical interval, counting the number of times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as an N2 operator of the casting to be retrieved;
the NaN operator calculator is used for acquiring a NaN operator, and the specific execution process comprises the following steps:
finding all adjacent patches of each triangular patch by taking patch adjacent points as clues;
randomly selecting a plurality of surface patches on a three-dimensional model of a casting to be retrieved, and calculating a cosine value of an included angle between each surface patch and a normal vector of an adjacent surface patch;
averaging the cosine values to obtain an array containing all the average cosine values;
and setting a statistical interval, counting the times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as a NaN operator of the casting to be retrieved.
Preferably, the domain knowledge extracting unit includes: the system comprises a Mod operator calculator, a Sym operator calculator and an Env operator calculator;
the Mod operator calculator is used for acquiring a Mod operator, and the specific execution process is as follows:
taking the sum of the triangular areas of the castings to be retrieved as the heat dissipation area of the three-dimensional models of the castings to be retrieved;
dividing the volume of the three-dimensional model of the casting to be retrieved by the heat dissipation area to obtain a Mod operator;
the Sym operator calculator is used for acquiring a Sym operator, and the specific execution process comprises the following steps:
comparing the relation between the gravity center of each surface patch of the casting to be retrieved and the coordinate plane, and dividing the three-dimensional model of the casting to be retrieved along the YOZ plane, the XOZ plane and the XOY plane respectively;
calculating surface area errors of two parts of the three-dimensional model after being divided along each coordinate plane, and taking the surface area errors as the symmetry of the coordinate plane;
sequentially calculating the symmetry degrees of the three coordinate surfaces to form a characteristic vector of the three-dimensional model, wherein the characteristic vector is a Sym operator;
the Env operator calculator is used for acquiring an Env operator, and the specific execution process is as follows:
determining X, Y and the maximum value and the minimum value of the three-dimensional model on the Z axis under a Cartesian coordinate system by traversing all triangular patch vertexes of the three-dimensional model of the casting to be retrieved, and obtaining the envelope size of the three-dimensional model of the casting to be retrieved along the direction of the coordinate main axis;
and calculating the width-length ratio, the width-height ratio and the height-length ratio of the envelope size to obtain a three-dimensional feature vector, wherein the three-dimensional feature vector is an Env operator.
Preferably, the three-dimensional model processing module includes:
the file analyzing unit is used for analyzing and reading a three-dimensional model file of the casting to be retrieved by utilizing three-dimensional modeling software;
the triangular tiling unit is used for acquiring a triangular tiling file of the three-dimensional model of the casting to be retrieved by utilizing the three-dimensional model tiling function in the software;
the area calculation unit is used for calculating the areas of all triangular patches on the casting to be retrieved according to a Helen formula;
the gravity center calculating unit is used for calculating the centers of all triangular patches of the three-dimensional model of the casting to be retrieved according to a central coordinate formula, and further calculating the gravity center of the three-dimensional model of the casting to be retrieved;
the translation matrix calculation unit is used for calculating a translation matrix of the three-dimensional model of the casting to be retrieved according to the gravity centers of the triangular patches and the gravity center of the three-dimensional model of the casting to be retrieved;
the rotation matrix calculation unit is used for calculating covariance matrixes of vertexes of all triangular surface patches of the three-dimensional model of the casting to be retrieved and acquiring a rotation matrix of the casting to be retrieved;
and the attitude normalization unit is used for acting the translation matrix and the rotation matrix on each vertex of the three-dimensional model of the casting to be retrieved to finish the attitude normalization of the three-dimensional model.
Generally, compared with the prior art, the above technical solution conceived by the present invention has the following beneficial effects:
the invention provides a multi-feature casting three-dimensional model retrieval method aiming at the problem of 'historical process reuse' of a complex casting in the typical casting application field; the geometric shape characteristics of the three-dimensional model are extracted through shape distribution operators such as a D2 operator and an N2 operator, and meanwhile, the process characteristics such as casting modulus, casting symmetry plane and the like which are strongly related to the process design in the casting field are innovatively combined (namely, the shape characteristics and the field knowledge of the casting to be retrieved are calculated based on the three-dimensional model). Experimental results show that the method can effectively extract the information of the geometric shape (D2 operator and N2 operator), the concave-convex degree (NaN operator), the thermal node (Mod operator), the symmetric plane (Sym operator), the envelope size (Env operator) and the like of the three-dimensional model. Finally, the similar three-dimensional model is robustly retrieved under the condition that the casting is subjected to changes such as translation, rotation and the like, and meanwhile, as the retrieval of the three-dimensional model is realized, compared with the retrieval and construction of a two-dimensional model, the period is shortened in the aspect of process design.
When the method is used for carrying out characteristic combination on the D2 operator, the N2 operator, the NaN operator, the Mod operator and the like, a 'step-by-step elimination' mode is adopted (similarity of shape characteristics of a retrieved casting and each model in a model library is calculated firstly, a certain proportion of models are screened, domain knowledge is used for screening a three-dimensional model for one time, and a retrieval result is finally obtained).
Drawings
FIG. 1 is a schematic diagram of a multi-feature casting three-dimensional model retrieval method provided by an embodiment of the invention;
FIG. 2 is a three-dimensional model of a casting to be retrieved according to an embodiment of the present invention;
FIG. 3(a) is an initial effect diagram of a three-dimensional model of a casting to be retrieved, provided by an embodiment of the invention;
FIG. 3(b) is a diagram illustrating the effect of normalizing the posture of the three-dimensional model of the casting to be retrieved according to the embodiment of the invention;
FIG. 4 is a schematic diagram of an Env operator extraction flow of a three-dimensional model of a casting to be retrieved, provided by an embodiment of the present invention;
FIG. 5 is a schematic diagram of a Sym operator extraction process of a three-dimensional model of a casting to be retrieved, which is provided by the embodiment of the invention;
FIG. 6 is a multi-domain multi-class casting model test set provided by an embodiment of the present invention;
FIG. 7 is a test set of casting models in the rail transit field provided by an embodiment of the present invention;
FIG. 8 is a test set of casting models in the field of engineering machinery provided by an embodiment of the present invention;
FIG. 9 is a test set of fluid field casting models provided by embodiments of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1, in one aspect, the present invention provides a method for searching a multi-feature casting three-dimensional model, including the following steps:
acquiring a three-dimensional model of a casting to be retrieved, and performing triangular surface tiling and posture normalization processing on the three-dimensional model;
calculating the shape characteristics and the domain knowledge of the casting to be retrieved based on the three-dimensional model of the casting to be retrieved after the triangular surface tiling;
calculating the similarity between the shape characteristics of the casting to be retrieved and each model in the model library, and screening out a model with a certain proportion from the model library;
calculating the similarity between the domain knowledge of the casting to be retrieved and the screened model, and screening out a three-dimensional model retrieval result;
the domain knowledge comprises a hot spot characteristic, a symmetrical surface characteristic and an envelope size of the casting to be retrieved; the shape characteristics comprise geometric shape characteristics and concave-convex degree characteristics of the casting to be retrieved.
Preferably, the shape features include the D2 operator, the N2 operator, and the NaN operator of the three-dimensional model;
the calculation method of the D2 operator comprises the following steps:
randomly selecting a plurality of point pairs on the surface of the casting to be retrieved, calculating the distance between two points of each point pair to obtain an array containing the distances of all the point pairs, and normalizing the array;
setting a statistical interval, and counting the times of the point pair distances in the array after normalization, wherein the formed characteristic vector is a D2 operator of the casting to be retrieved;
the calculation method of the N2 operator comprises the following steps:
randomly selecting a plurality of patch pairs on a three-dimensional model of a casting to be retrieved, calculating cosine values of included angles of normal vectors of two patches of each patch pair, and obtaining an array containing all the cosine values;
setting a statistical interval, counting the number of times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as an N2 operator of the casting to be retrieved;
the calculation method of the NaN operator comprises the following steps:
finding all adjacent patches of each triangular patch by taking patch adjacent points as clues;
randomly selecting a plurality of surface patches on a three-dimensional model of a casting to be retrieved, and calculating a cosine value of an included angle between each surface patch and a normal vector of an adjacent surface patch;
averaging the cosine values to obtain an array containing all the average cosine values;
and setting a statistical interval, counting the times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as a NaN operator of the casting to be retrieved.
Preferably, the domain knowledge comprises the Mod operator, the Sym operator and the Env operator of the three-dimensional model;
the method for acquiring the Env operator comprises the following steps:
determining X, Y and the maximum value and the minimum value of the three-dimensional model on the Z axis under a Cartesian coordinate system by traversing all triangular patch vertexes of the three-dimensional model of the casting to be retrieved, and obtaining the envelope size of the three-dimensional model of the casting to be retrieved along the direction of the coordinate main axis;
calculating the combination of the width-length ratio, the width-height ratio and the height-length ratio of the envelope size to obtain a three-dimensional feature vector, wherein the three-dimensional feature vector is an Env operator;
the method for acquiring the Mod operator comprises the following steps:
taking the sum of the triangular areas of the castings to be retrieved as the heat dissipation area of the three-dimensional models of the castings to be retrieved;
dividing the volume of the three-dimensional model of the casting to be retrieved by the heat dissipation area to obtain a Mod operator;
the method for acquiring the Sym operator comprises the following steps:
comparing the relation between the gravity center of each surface patch of the casting to be retrieved and the coordinate plane, and dividing the three-dimensional model of the casting to be retrieved along the YOZ plane, the XOZ plane and the XOY plane respectively;
and calculating surface area errors of two parts of the three-dimensional model after being divided along each coordinate plane as the symmetry degree of the coordinate plane, and sequentially calculating the symmetry degrees of the three coordinate planes to form a characteristic vector of the three-dimensional model, wherein the characteristic vector is a Sym operator.
Preferably, the method for posture normalization of the three-dimensional model of the casting to be retrieved comprises the following steps:
analyzing and reading a three-dimensional model file of a casting to be retrieved by utilizing three-dimensional modeling software;
acquiring a triangular tiling file of the three-dimensional model of the casting to be retrieved by using a three-dimensional model tiling function in software;
calculating the areas of all triangular patches on the casting to be retrieved according to a Helen formula;
calculating the centers of all triangular patches of the three-dimensional model of the casting to be retrieved according to a central coordinate formula, and further calculating the gravity center of the three-dimensional model of the casting to be retrieved;
calculating a translation matrix of the three-dimensional model of the casting to be retrieved according to the gravity centers of the triangular patches and the gravity center of the three-dimensional model of the casting to be retrieved;
calculating covariance matrixes of vertexes of all triangular surface patches of the three-dimensional model of the casting to be retrieved, and acquiring a rotation matrix of the casting to be retrieved;
and (4) acting the translation matrix and the rotation matrix on each vertex of the three-dimensional model of the casting to be retrieved to finish the posture normalization of the three-dimensional model.
Preferably, the NaN operator obtaining method includes the following steps:
s4.1: calculating the areas of all triangular patches on the three-dimensional model to be retrieved according to a Helen formula to obtain an ordered area array;
s4.2: searching all adjacent patches of each triangular patch under the condition of whether patch adjacent points exist or not, and creating a key-value dictionary according to a patch sequence number-adjacent patch sequence number set;
s4.3: searching a key-value dictionary, finding all adjacent patches of all triangular patches of the three-dimensional model of the casting to be retrieved, calculating cosine values of included angles of normal vectors of all patches and all adjacent patches of each patch, and averaging to obtain a list of cosine values of included angles of normal vectors;
s4.4: generating a plurality of random numbers in the ordered area array by using a random number function, determining subscripts of the random numbers in the area array by using binary search, and determining a plurality of sampling surface patches; inquiring the cosine value list of the normal vector included angle in S4.3 to obtain cosine values of the normal vector included angles of the plurality of sampling surface patches;
s4.5: and determining the number of the statistical intervals so as to determine the range of each statistical interval, counting the occurrence times of cosine values of the normal vector included angle of the surface patch in each interval range to obtain corresponding frequency distribution, and dividing the frequency distribution by the number of samples to be used as a NaN operator of the three-dimensional model.
Preferably, the method for acquiring the Sym operator specifically includes the following steps:
according to a gravity center coordinate formula, recalculating the gravity centers of all triangular patches of the three-dimensional model of the casting to be retrieved after the posture normalization;
dividing all the surface patches of the gradual three-dimensional model into two parts according to the position relation between the gravity center of each surface patch and the coordinate plane, and respectively calculating the area sum of the two parts of the surface patches;
obtaining surface area errors of two parts of the three-dimensional model according to the area sum of the two parts of the surface patches, and taking the surface area errors as the symmetry of the three-dimensional model relative to the coordinate plane;
combining the symmetries of the three coordinate surfaces to form a three-dimensional characteristic vector as a Sym operator of the three-dimensional model;
wherein the coordinate plane comprises a YOZ plane, an XOZ plane and an XOY plane.
In another aspect, a multi-feature casting three-dimensional model retrieval device comprises:
the three-dimensional model processing module is used for acquiring a three-dimensional model of the casting to be retrieved, and performing triangular surface tiling and posture normalization processing on the three-dimensional model;
the characteristic extraction module is used for calculating the shape characteristic and the domain knowledge of the casting to be retrieved based on the triangular tiled three-dimensional model of the casting to be retrieved; wherein the triangle patch is subjected to attitude normalization processing;
the model screening module is used for calculating the similarity between the shape characteristics of the casting to be retrieved and each model in the model library and screening out a model with a certain proportion in the model library;
the method is used for calculating the similarity between the field knowledge of the casting to be searched and the screened model, and screening out a three-dimensional model searching result;
the domain knowledge comprises a hot spot characteristic, a symmetrical plane characteristic and an envelope size of the casting to be retrieved; the shape characteristics comprise geometric shape characteristics and concave-convex degree characteristics of the casting to be retrieved.
Preferably, the shape features include the D2 operator, the N2 operator, and the NaN operator of the three-dimensional model;
the feature extraction module comprises a shape feature extraction unit and a domain knowledge extraction unit; the shape feature extraction unit comprises a D2 operator calculator, an N2 operator calculator and a NaN operator calculator;
the D2 operator calculator is used for acquiring a D2 operator, and the specific execution process is as follows:
randomly selecting a plurality of point pairs on the surface of the casting to be retrieved, calculating the distance between two points of each point pair to obtain an array containing the distances of all the point pairs, and normalizing the array;
setting a statistical interval, and counting the times of the point pair distances in the array after normalization, wherein the formed characteristic vector is a D2 operator of the casting to be retrieved;
the N2 operator calculator is used for acquiring an N2 operator, and the specific execution process is as follows:
randomly selecting a plurality of patch pairs on a three-dimensional model of a casting to be retrieved, calculating cosine values of included angles of normal vectors of two patches of each patch pair, and obtaining an array containing all the cosine values;
setting a statistical interval, counting the number of times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as an N2 operator of the casting to be retrieved;
the NaN operator calculator is used for acquiring a NaN operator, and the specific execution process comprises the following steps:
finding all adjacent patches of each triangular patch by taking patch adjacent points as clues;
randomly selecting a plurality of surface patches on a three-dimensional model of a casting to be retrieved, and calculating a cosine value of an included angle between each surface patch and a normal vector of an adjacent surface patch;
averaging the cosine values to obtain an array containing all the average cosine values;
and setting a statistical interval, counting the times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as a NaN operator of the casting to be retrieved.
Preferably, the domain knowledge extracting unit includes: the system comprises a Mod operator calculator, a Sym operator calculator and an Env operator calculator;
the Mod operator calculator is used for acquiring a Mod operator, and the specific execution process is as follows:
taking the sum of the triangular areas of the castings to be retrieved as the heat dissipation area of the three-dimensional models of the castings to be retrieved;
dividing the volume of the three-dimensional model of the casting to be retrieved by the heat dissipation area to obtain a Mod operator;
the Sym operator calculator is used for acquiring a Sym operator, and the specific execution process comprises the following steps:
comparing the relation between the gravity center of each surface patch of the casting to be retrieved and the coordinate plane, and dividing the three-dimensional model of the casting to be retrieved along the YOZ plane, the XOZ plane and the XOY plane respectively;
calculating surface area errors of two parts of the three-dimensional model after being divided along each coordinate plane, and taking the surface area errors as the symmetry of the coordinate plane;
sequentially calculating the symmetry degrees of the three coordinate surfaces to form a characteristic vector of the three-dimensional model, wherein the characteristic vector is a Sym operator;
the Env operator calculator is used for acquiring an Env operator, and the specific execution process is as follows:
determining X, Y and the maximum value and the minimum value of the three-dimensional model on the Z axis under a Cartesian coordinate system by traversing all triangular patch vertexes of the three-dimensional model of the casting to be retrieved, and obtaining the envelope size of the three-dimensional model of the casting to be retrieved along the direction of the coordinate main axis;
and calculating the width-length ratio, the width-height ratio and the height-length ratio of the envelope size to obtain a three-dimensional feature vector, wherein the three-dimensional feature vector is an Env operator.
Preferably, the three-dimensional model processing module includes:
the file analyzing unit is used for analyzing and reading a three-dimensional model file of the casting to be retrieved by utilizing three-dimensional modeling software;
the triangular tiling unit is used for acquiring a triangular tiling file of the three-dimensional model of the casting to be retrieved by utilizing the three-dimensional model tiling function in the software;
the area calculation unit is used for calculating the areas of all triangular patches on the casting to be retrieved according to a Helen formula;
the gravity center calculating unit is used for calculating the centers of all triangular patches of the three-dimensional model of the casting to be retrieved according to a central coordinate formula, and further calculating the gravity center of the three-dimensional model of the casting to be retrieved;
the translation matrix calculation unit is used for calculating a translation matrix of the three-dimensional model of the casting to be retrieved according to the gravity centers of the triangular patches and the gravity center of the three-dimensional model of the casting to be retrieved;
the rotation matrix calculation unit is used for calculating covariance matrixes of vertexes of all triangular surface patches of the three-dimensional model of the casting to be retrieved and acquiring a rotation matrix of the casting to be retrieved;
and the attitude normalization unit is used for acting the translation matrix and the rotation matrix on each vertex of the three-dimensional model of the casting to be retrieved to finish the attitude normalization of the three-dimensional model.
Examples
The embodiment provides a three-dimensional model retrieval method for a multi-feature casting, wherein the casting three-dimensional model shown in fig. 2 is used as a casting to be retrieved for description, and the number of the statistical intervals of the D2 operators is 128; the number of the N2 operator and NaN operator statistical intervals is 180; the number of the statistical intervals of the D2 operator is 127, and the number of sampling points (sampling patches) is 100000; the method comprises the following steps:
s1: obtaining three-dimensional models of various castings, and performing triangular surface tiling on the three-dimensional models;
carrying out attitude normalization processing on the model, which mainly comprises the following steps: (1) translating the origin of the coordinate system to the gravity center of the three-dimensional model to realize the translation invariance of the three-dimensional model; (2) determining a main shaft of a coordinate system to realize the rotation invariance of the three-dimensional model; the method comprises the following specific steps:
s1.1: analyzing and reading the three-dimensional model file of the casting by using three-dimensional modeling software;
s1.2: obtaining a triangular tiling file of the three-dimensional model of the casting by using a three-dimensional model tiling function in software, and storing the triangular tiling file by using an STL text file;
specifically, after the three-dimensional model of the casting is analyzed and read and the triangular patches are formed by using three-dimensional modeling software, three-dimensional vertex coordinates and three-dimensional normal vector coordinates of the model patches are respectively stored by using a list data structure, and subscripts of a list are used for corresponding to serial numbers of the triangular patches;
s1.3: according to the Helen formula, the areas of all triangular patches on the three-dimensional model of the casting are calculated and stored in a list S [ S ]1,S2,S3…Sn]The Helen formula is as follows:
p＝(a+b+c)/2
Figure BDA0003308346010000151
s1.4: according to a central coordinate formula, calculating the gravity centers of all triangular patches of the three-dimensional model of the casting, and storing a list Gm[G1[x1，y1，z1]，G2[x2，y2，z2]，……，Gn[xn，yn，zn]]And then calculate the center of gravity of the model, denoted as GM[Gxm，Gym，Gzm]If the translation invariance of the three-dimensional model is not changed, the translation matrix is recorded as MT(ii) a The calculation formula is as follows:
Figure BDA0003308346010000152
Figure BDA0003308346010000153
Figure BDA0003308346010000154
Figure BDA0003308346010000155
s1.5: recording the set of all triangular patch vertexes of the three-dimensional model as [ [ x ]1 o，y1 o，z1 o]，[x2 o，y2 o，z2 o]，……，[xp o，yp o，zp o]]Wherein p is the total number of vertexes of the three-dimensional model; the covariance matrix of the vertex is M, and the rotation matrix M of the rotation invariance of the three-dimensional model is realizedRFrom three unitized eigenvectors [ a ] of the covariance matrix Mi，bi，ci]The method comprises the following steps of specifically arranging three eigenvectors in a rotation matrix according to an ascending order of eigenvalues, and calculating the formula as follows:
Figure BDA0003308346010000161
Figure BDA0003308346010000162
s1.6: the translation matrix M isTRotation matrix MRActing on each Vertex of the three-dimensional model to obtain a Vertex coordinate set Vertex of the three-dimensional model after the posture normalization, wherein a calculation formula is as follows:
Figure BDA0003308346010000163
specifically, for the three-dimensional model of the casting to be retrieved as shown in fig. 2, fig. 3(a) is an effect diagram of the casting to be retrieved in an initial coordinate system; after the action of the translation matrix and the rotation matrix, the effect of posture normalization is shown in fig. 3 (b);
s2: randomly selecting a plurality of point pairs on the surface of the three-dimensional model of the casting, calculating the distance between two points of each point pair to obtain an array containing the distances of all the point pairs, normalizing the array, setting a statistical interval, counting the times of the point pair distances in the array appearing in each interval to form a characteristic vector, namely a D2 operator of the three-dimensional model of the casting, and extracting the overall geometric distance distribution information of all vertexes of the three-dimensional model of the casting;
s3: randomly selecting a plurality of patch pairs on a three-dimensional model of the casting, calculating cosine values of normal vector included angles of two patches of each patch pair, obtaining an array containing all the cosine values, setting a statistical interval, counting the times of the cosine values of the included angles in the array appearing in each interval to form a characteristic vector, namely an N2 operator of the three-dimensional model of the casting, and extracting the integral geometric included angle distribution information of all patches of the three-dimensional model of the casting;
s4: finding out all adjacent patches of each triangular patch by taking patch adjacent points as clues; randomly selecting a plurality of surface patches on the three-dimensional model of the casting, calculating cosine values of included angles between each surface patch and normal vectors of adjacent surface patches of the surface patches and averaging to obtain an array containing all average cosine values, setting a statistical interval, counting the times of the average cosine values in the array appearing in each interval to form a characteristic vector, namely a NaN operator of the three-dimensional model of the casting, and extracting local concave-convex degree distribution information of all the surface patches of the three-dimensional model of the casting; the method specifically comprises the following steps:
s4.1: according to the Helen formula, the areas of all triangular patches on the three-dimensional model of the casting are calculated and stored in a list S [ S ]1,S2,S3…Sn]Then, an ordered area array T [ T ] is obtained according to the list S1，T2，T3，……，Tn]，x＝1，2，3，…，n；
S4.2: searching all adjacent patches of each triangular patch under the condition of whether patch adjacent points exist or not, and creating a key-value dictionary according to a patch sequence number-adjacent patch sequence number set;
s4.3: searching the key-value dictionary created in the step S4.2, finding all adjacent patches of all triangular patches of the three-dimensional model, calculating the cosine values of included angles of normal vectors of all patches and all adjacent patches of each patch, and averaging to obtain a cosine value list of the included angles of normal vectors;
s4.4: using random number functions, from 0 to TnGenerating a plurality of random numbers, determining subscripts of the random numbers in the area array by utilizing binary search, and determining a plurality of sampling surface patches; inquiring the cosine value list of the normal vector included angle in S4.3 to obtain cosine values of the normal vector included angles of the plurality of sampling surface patches;
s4.5: determining the number of statistical intervals so as to determine the range of each statistical interval, counting the number of times of appearance of cosine values of a normal vector included angle of a surface patch in each interval range to obtain corresponding frequency distribution, and dividing the frequency distribution by the number of samples to be used as a NaN operator of the three-dimensional model;
s5: calculating the volume V (cm) of the three-dimensional model of the casting3) Taking the sum of all triangular areas as the heat dissipation area A (cm) of the three-dimensional model2) Calculating the integral modulus of the three-dimensional model of the casting by using the following formula, taking the integral modulus as a Mod operator of the three-dimensional model, and extracting the wall thickness distribution information, namely the heat node information, of the integral three-dimensional model of the casting;
Figure BDA0003308346010000171
s6: determining the maximum value and the minimum value of the model vertex on X, Y and Z axes by traversing all triangular patch vertexes of the three-dimensional model after the attitude normalization, and finally obtaining the envelope size of the three-dimensional model along the main axis direction of a coordinate system; calculating the width-length ratio, the width-height ratio and the height-length ratio of the envelope size to obtain a three-dimensional characteristic vector, namely an Env operator of the three-dimensional casting model, and extracting the integral envelope size distribution information of the three-dimensional casting model; the extraction process of the method is shown in figure 4; the method specifically comprises the following steps:
s6.1: traversing all vertex coordinates of the casting three-dimensional model, respectively finding the maximum value and the minimum value of the vertex in X, Y and Z directions, obtaining the range distribution of the vertex of the three-dimensional model in a main shaft of a coordinate system, and taking the distribution size of the X-axis direction as the width (width) of the casting three-dimensional model, the distribution size of the Y-axis direction as the height (height) of the casting three-dimensional model and the distribution size of the Z-axis direction as the length (length) of the casting three-dimensional model;
s6.2: in order to eliminate the influence caused by different sizes of different three-dimensional models, three characteristic values of width-length ratio (width/length), width-height ratio (width/height) and height-length ratio (height/length) are respectively calculated and combined into a three-dimensional characteristic vector to be used as an Env operator of the three-dimensional model;
s7: comparing the position relationship of the gravity center and the coordinate plane of each surface patch of the three-dimensional model, dividing the three-dimensional model along a YOZ plane, an XOZ plane and an XOY plane respectively, calculating the surface area error of two parts of the divided model as the symmetry degree of the coordinate plane, sequentially calculating the model symmetry degrees of the three coordinate planes to form a three-dimensional characteristic vector, namely a Sym operator of the three-dimensional model of the casting, extracting the integral symmetry plane information of the three-dimensional model of the casting, wherein the extraction flow of the method is shown in FIG. 5; the method comprises the following specific steps:
s7.1: according to a barycentric coordinate formula, recalculating barycenter G of all triangular patches of the three-dimensional casting model after posture normalizationm a[G1 a[x1，y1，z1]，G2 a[x2，y2，z2]，……，Gn a[xn，yn，zn]]；
S7.2: firstly, dividing all patches of a gradual three-dimensional model into two parts according to the position relation between the gravity center of each patch and an XOY coordinate plane, namely, considering that the patch is positioned on the coordinate plane, wherein a coordinate value on a Z axis of the gravity center of the patch is more than or equal to 0; when the coordinate value is less than 0, the patch is considered to be positioned below the coordinate plane, and the segmentation is completed through the method; then, the areas of the two parts of the patches, AreaSumA and AreaSumB are respectively calculated; then, the surface area error of the two parts of the three-dimensional model can be obtained through the following formula and is used as the symmetry of the three-dimensional model about the XOY plane;
Figure BDA0003308346010000181
s7.3: simulating the process of S7.2, and sequentially calculating the symmetry of the three-dimensional model about the YOZ plane and the XOZ plane;
s7.4: combining the three symmetries of the three-dimensional model obtained in the S7.2 and the S7.3 about an XOY plane, an XOZ plane and a YOZ plane to form a three-dimensional characteristic vector which is used as a Sym operator of the three-dimensional model;
s8: firstly, sequentially calculating the similarity between a model to be retrieved under a D2 operator and each model in a model library, and selecting the first 25-30 models with higher similarity for the next round of feature screening; then, sequentially calculating the similarity between the model to be retrieved under the N2 operator and each screening model in the previous round, and selecting the first 20-25 models with larger similarity; then, eliminating a certain proportion of models through operators such as NaN, Mod, Sym and the like in sequence; finally, 5-8 models which are reserved after the Env operator is subjected to sixth screening are retrieval models for providing process references; the method comprises the following specific steps:
s8.1: sequentially calculating the similarity between the three-dimensional model to be retrieved and all models in the model library under the D2 operator, retrieving 25-30 three-dimensional models with large similarity according to the similarity, and performing one-round model characteristic elimination; the operator similarity measure is as follows:
Figure BDA0003308346010000191
wherein, D is a feature vector D ═ D1, D2, D3, …, di of the model to be searched obtained by a certain operator]I represents the dimension of D; dnFeature vectors Dn ═ d1, d2, d3, …, di obtained for models in the model library from the same operator]nN is the nth model;
s8.2: sequentially calculating the similarity between the model to be retrieved and the 25-30 three-dimensional models obtained in S8.1 under the N2 operator, retrieving and removing 20-25 three-dimensional models with high similarity according to the similarity, and performing next round of model feature elimination;
s8.3: sequentially calculating the similarity between the model to be retrieved and the 20-25 three-dimensional models obtained in S8.2 under the NaN operator, retrieving and removing 15-20 three-dimensional models with large similarity according to the similarity, and performing next round of model feature elimination;
s8.4: sequentially calculating the similarity between the model to be retrieved and the 15-20 three-dimensional models obtained in S8.3 under the Mod operator, retrieving 10-15 three-dimensional models with high similarity according to the similarity, and performing next round of model feature elimination;
s8.5: sequentially calculating the similarity between the model to be retrieved and 10-15 three-dimensional models obtained in S8.4 under the Sym operator, retrieving 5-8 three-dimensional models with high similarity according to the similarity, and performing next round of model feature elimination;
s8.6: and sequentially calculating the similarity between the model to be retrieved and 8-12 three-dimensional models obtained in S8.5 under the Env operator, and retrieving 5-8 three-dimensional models with high similarity according to the similarity, namely the final three-dimensional model retrieval result.
In order to better embody the classification capability of the method of the invention on common categories of steel castings, table 1 shows the retrieval effect of the method of the invention in the casting field casting model test set such as fig. 6, fig. 7, fig. 8 and fig. 9;
TABLE 1
Figure BDA0003308346010000201
The invention provides a multi-feature casting three-dimensional model retrieval method combining shape features and field knowledge and a step-by-step elimination mechanism, and provides a method for extracting information such as geometric shape, concave-convex degree, thermal junction, symmetrical plane, envelope size and the like of a three-dimensional model from two dimensions of the shape features and the casting field knowledge on the basis of the 'surface tiling' of the model so as to realize the similar retrieval of the three-dimensional model. The method has good retrieval effect on typical casting application fields such as aviation, aerospace, rail transit, engineering machinery and fluid machinery.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims (10)

1. A multi-feature casting three-dimensional model retrieval method is characterized by comprising the following steps:
acquiring a three-dimensional model of a casting to be retrieved, and performing triangular surface tiling and posture normalization processing on the three-dimensional model;
calculating the shape characteristics and the domain knowledge of the casting to be retrieved based on the triangular tiled three-dimensional model of the casting to be retrieved; wherein the triangle patch is subjected to attitude normalization processing;
calculating the similarity between the shape characteristics of the casting to be retrieved and each model in the model library, and screening out a model with a certain proportion in the model library;
calculating the similarity between the domain knowledge of the casting to be retrieved and the screened model, and screening out a three-dimensional model retrieval result;
the domain knowledge comprises a hot spot characteristic, a symmetrical plane characteristic and an envelope size of the casting to be retrieved; the shape characteristics comprise geometric shape characteristics and concave-convex degree characteristics of the casting to be retrieved.
2. The method for retrieving the multi-feature casting three-dimensional model according to claim 1, wherein the shape features comprise a D2 operator, an N2 operator, and a NaN operator of the three-dimensional model;
the calculation method of the D2 operator comprises the following steps:
randomly selecting a plurality of point pairs on the surface of the casting to be retrieved, calculating the distance between two points of each point pair to obtain an array containing the distances of all the point pairs, and normalizing the array;
setting a statistical interval, and counting the times of the point pair distances in the array after normalization, wherein the formed characteristic vector is a D2 operator of the casting to be retrieved;
the calculation method of the N2 operator comprises the following steps:
randomly selecting a plurality of patch pairs on a three-dimensional model of a casting to be retrieved, calculating cosine values of included angles of normal vectors of two patches of each patch pair, and obtaining an array containing all the cosine values;
setting a statistical interval, counting the number of times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as an N2 operator of the casting to be retrieved;
the calculation method of the NaN operator comprises the following steps:
finding all adjacent patches of each triangular patch by taking patch adjacent points as clues;
randomly selecting a plurality of surface patches on a three-dimensional model of a casting to be retrieved, and calculating a cosine value of an included angle between each surface patch and a normal vector of an adjacent surface patch;
averaging the cosine values to obtain an array containing all the average cosine values;
and setting a statistical interval, counting the times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as a NaN operator of the casting to be retrieved.
3. The method for retrieving the multi-feature casting three-dimensional model according to claim 2, wherein the domain knowledge comprises a Mod operator, a Sym operator and an Env operator of the three-dimensional model;
the method for acquiring the Env operator comprises the following steps:
determining X, Y and the maximum value and the minimum value of the three-dimensional model on the Z axis under a Cartesian coordinate system by traversing all triangular patch vertexes of the three-dimensional model of the casting to be retrieved, and obtaining the envelope size of the three-dimensional model of the casting to be retrieved along the direction of the coordinate main axis;
calculating the combination of the width-length ratio, the width-height ratio and the height-length ratio of the envelope size to obtain a three-dimensional feature vector, wherein the three-dimensional feature vector is an Env operator;
the method for acquiring the Mod operator comprises the following steps:
taking the sum of the triangular areas of the castings to be retrieved as the heat dissipation area of the three-dimensional models of the castings to be retrieved;
dividing the volume of the three-dimensional model of the casting to be retrieved by the heat dissipation area to obtain a Mod operator;
the method for acquiring the Sym operator comprises the following steps:
comparing the relation between the gravity center of each surface patch of the casting to be retrieved and the coordinate plane, and dividing the three-dimensional model of the casting to be retrieved along the YOZ plane, the XOZ plane and the XOY plane respectively;
calculating surface area errors of two parts of the three-dimensional model after being divided along each coordinate plane, and taking the surface area errors as the symmetry of the coordinate plane;
and sequentially calculating the symmetry degrees of the three coordinate surfaces to form a characteristic vector of the three-dimensional model, wherein the characteristic vector is a Sym operator.
4. The method for retrieving the multi-feature casting three-dimensional model as claimed in any one of claims 1 to 3, wherein the method for posture normalization of the three-dimensional model of the casting to be retrieved comprises the following steps:
analyzing and reading a three-dimensional model file of a casting to be retrieved by utilizing three-dimensional modeling software;
acquiring a triangular tiling file of the three-dimensional model of the casting to be retrieved by using a three-dimensional model tiling function in software;
calculating the areas of all triangular patches on the casting to be retrieved according to a Helen formula;
calculating the centers of all triangular patches of the three-dimensional model of the casting to be retrieved according to a central coordinate formula, and further calculating the gravity center of the three-dimensional model of the casting to be retrieved;
calculating a translation matrix of the three-dimensional model of the casting to be retrieved according to the gravity centers of the triangular patches and the gravity center of the three-dimensional model of the casting to be retrieved;
calculating covariance matrixes of vertexes of all triangular surface patches of the three-dimensional model of the casting to be retrieved, and acquiring a rotation matrix of the casting to be retrieved;
and (4) acting the translation matrix and the rotation matrix on each vertex of the three-dimensional model of the casting to be retrieved to finish the posture normalization of the three-dimensional model.
5. The retrieval method for the multi-feature casting three-dimensional model is characterized in that the acquisition method for the NaN operator comprises the following steps:
s4.1: calculating the areas of all triangular patches on the three-dimensional model to be retrieved according to a Helen formula to obtain an ordered area array;
s4.2: searching all adjacent patches of each triangular patch under the condition of whether patch adjacent points exist or not, and creating a key-value dictionary according to a patch sequence number-adjacent patch sequence number set;
s4.3: searching a key-value dictionary, finding all adjacent patches of all triangular patches of the three-dimensional model of the casting to be retrieved, calculating cosine values of included angles of normal vectors of all patches and all adjacent patches of each patch, and averaging to obtain a list of cosine values of included angles of normal vectors;
s4.4: generating a plurality of random numbers in the ordered area array by using a random number function, determining subscripts of the random numbers in the area array by using binary search, and determining a plurality of sampling surface patches; inquiring the cosine value list of the normal vector included angle in S4.3 to obtain cosine values of the normal vector included angles of the plurality of sampling surface patches;
s4.5: and determining the number of the statistical intervals so as to determine the range of each statistical interval, counting the occurrence times of cosine values of the normal vector included angle of the surface patch in each interval range to obtain corresponding frequency distribution, and dividing the frequency distribution by the number of samples to be used as a NaN operator of the three-dimensional model.
6. The method for retrieving the multi-feature casting three-dimensional model according to claim 3, wherein the method for obtaining the Sym operator comprises the following steps:
according to a gravity center coordinate formula, recalculating the gravity centers of all triangular patches of the three-dimensional model of the casting to be retrieved after the posture normalization;
dividing all the surface patches of the gradual three-dimensional model into two parts according to the position relation between the gravity center of each surface patch and the coordinate plane, and respectively calculating the area sum of the two parts of the surface patches;
obtaining surface area errors of two parts of the three-dimensional model according to the area sum of the two parts of the surface patches, and taking the surface area errors as the symmetry of the three-dimensional model relative to the coordinate plane;
combining the symmetries of the three coordinate surfaces to form a three-dimensional characteristic vector as a Sym operator of the three-dimensional model;
wherein the coordinate plane comprises a YOZ plane, an XOZ plane and an XOY plane.
7. A multi-feature casting three-dimensional model retrieval device is characterized by comprising:
the three-dimensional model processing module is used for acquiring a three-dimensional model of the casting to be retrieved, and performing triangular surface tiling and posture normalization processing on the three-dimensional model;
the characteristic extraction module is used for calculating the shape characteristic and the domain knowledge of the casting to be retrieved based on the triangular tiled three-dimensional model of the casting to be retrieved; wherein the triangle patch is subjected to attitude normalization processing;
the model screening module is used for calculating the similarity between the shape characteristics of the casting to be retrieved and each model in the model library and screening out a model with a certain proportion in the model library;
the method is used for calculating the similarity between the field knowledge of the casting to be searched and the screened model, and screening out a three-dimensional model searching result;
the domain knowledge comprises a hot spot characteristic, a symmetrical plane characteristic and an envelope size of the casting to be retrieved; the shape characteristics comprise geometric shape characteristics and concave-convex degree characteristics of the casting to be retrieved.
8. The multi-feature casting three-dimensional model retrieval device of claim 1, wherein the shape features comprise a D2 operator, an N2 operator, and a NaN operator of the three-dimensional model;
the feature extraction module comprises a shape feature extraction unit and a domain knowledge extraction unit; the shape feature extraction unit comprises a D2 operator calculator, an N2 operator calculator and a NaN operator calculator;
the D2 operator calculator is used for acquiring a D2 operator, and the specific execution process is as follows:
randomly selecting a plurality of point pairs on the surface of the casting to be retrieved, calculating the distance between two points of each point pair to obtain an array containing the distances of all the point pairs, and normalizing the array;
setting a statistical interval, and counting the times of the point pair distances in the array after normalization, wherein the formed characteristic vector is a D2 operator of the casting to be retrieved;
the N2 operator calculator is used for acquiring an N2 operator, and the specific execution process is as follows:
randomly selecting a plurality of patch pairs on a three-dimensional model of a casting to be retrieved, calculating cosine values of included angles of normal vectors of two patches of each patch pair, and obtaining an array containing all the cosine values;
setting a statistical interval, counting the number of times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as an N2 operator of the casting to be retrieved;
the NaN operator calculator is used for acquiring a NaN operator, and the specific execution process comprises the following steps:
finding all adjacent patches of each triangular patch by taking patch adjacent points as clues;
randomly selecting a plurality of surface patches on a three-dimensional model of a casting to be retrieved, and calculating a cosine value of an included angle between each surface patch and a normal vector of an adjacent surface patch;
averaging the cosine values to obtain an array containing all the average cosine values;
and setting a statistical interval, counting the times of cosine values of included angles in the group appearing in each interval, and taking the formed characteristic vector as a NaN operator of the casting to be retrieved.
9. The multi-feature casting three-dimensional model retrieval device according to claim 8, wherein the domain knowledge extraction unit includes: the system comprises a Mod operator calculator, a Sym operator calculator and an Env operator calculator;
the Mod operator calculator is used for acquiring a Mod operator, and the specific execution process is as follows:
taking the sum of the triangular areas of the castings to be retrieved as the heat dissipation area of the three-dimensional models of the castings to be retrieved;
dividing the volume of the three-dimensional model of the casting to be retrieved by the heat dissipation area to obtain a Mod operator;
the Sym operator calculator is used for acquiring a Sym operator, and the specific execution process comprises the following steps:
comparing the relation between the gravity center of each surface patch of the casting to be retrieved and the coordinate plane, and dividing the three-dimensional model of the casting to be retrieved along the YOZ plane, the XOZ plane and the XOY plane respectively;
calculating surface area errors of two parts of the three-dimensional model after being divided along each coordinate plane, and taking the surface area errors as the symmetry of the coordinate plane;
sequentially calculating the symmetry degrees of the three coordinate surfaces to form a characteristic vector of the three-dimensional model, wherein the characteristic vector is a Sym operator;
the Env operator calculator is used for acquiring an Env operator, and the specific execution process is as follows:
determining X, Y and the maximum value and the minimum value of the three-dimensional model on the Z axis under a Cartesian coordinate system by traversing all triangular patch vertexes of the three-dimensional model of the casting to be retrieved, and obtaining the envelope size of the three-dimensional model of the casting to be retrieved along the direction of the coordinate main axis;
and calculating the width-length ratio, the width-height ratio and the height-length ratio of the envelope size to obtain a three-dimensional feature vector, wherein the three-dimensional feature vector is an Env operator.
10. The multi-feature casting three-dimensional model retrieval device according to any one of claims 7 to 9, wherein the three-dimensional model processing module comprises:
the file analyzing unit is used for analyzing and reading a three-dimensional model file of the casting to be retrieved by utilizing three-dimensional modeling software;
the triangular tiling unit is used for acquiring a triangular tiling file of the three-dimensional model of the casting to be retrieved by utilizing the three-dimensional model tiling function in the software;
the area calculation unit is used for calculating the areas of all triangular patches on the casting to be retrieved according to a Helen formula;
the gravity center calculating unit is used for calculating the centers of all triangular patches of the three-dimensional model of the casting to be retrieved according to a central coordinate formula, and further calculating the gravity center of the three-dimensional model of the casting to be retrieved;
the translation matrix calculation unit is used for calculating a translation matrix of the three-dimensional model of the casting to be retrieved according to the gravity centers of the triangular patches and the gravity center of the three-dimensional model of the casting to be retrieved;
the rotation matrix calculation unit is used for calculating covariance matrixes of vertexes of all triangular surface patches of the three-dimensional model of the casting to be retrieved and acquiring a rotation matrix of the casting to be retrieved;
and the attitude normalization unit is used for acting the translation matrix and the rotation matrix on each vertex of the three-dimensional model of the casting to be retrieved to finish the attitude normalization of the three-dimensional model.
CN202111209575.0A 2021-10-18 2021-10-18 Multi-feature casting three-dimensional model retrieval method and device Pending CN113961738A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111209575.0A CN113961738A (en) 2021-10-18 2021-10-18 Multi-feature casting three-dimensional model retrieval method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111209575.0A CN113961738A (en) 2021-10-18 2021-10-18 Multi-feature casting three-dimensional model retrieval method and device

Publications (1)

Publication Number Publication Date
CN113961738A true CN113961738A (en) 2022-01-21

Family

ID=79465027

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111209575.0A Pending CN113961738A (en) 2021-10-18 2021-10-18 Multi-feature casting three-dimensional model retrieval method and device

Country Status (1)

Country Link
CN (1) CN113961738A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115423947A (en) * 2022-11-03 2022-12-02 成都飞机工业（集团）有限责任公司 Three-dimensional model retrieval method, device, equipment and medium
CN115496865A (en) * 2022-11-18 2022-12-20 南京智程信息科技有限公司 Similar model searching method and device and storage medium
CN116402988A (en) * 2023-05-11 2023-07-07 北京冰河起源科技有限公司 Three-dimensional model processing method, device and storage medium

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115423947A (en) * 2022-11-03 2022-12-02 成都飞机工业（集团）有限责任公司 Three-dimensional model retrieval method, device, equipment and medium
CN115423947B (en) * 2022-11-03 2023-03-24 成都飞机工业（集团）有限责任公司 Three-dimensional model retrieval method, device, equipment and medium
CN115496865A (en) * 2022-11-18 2022-12-20 南京智程信息科技有限公司 Similar model searching method and device and storage medium
CN115496865B (en) * 2022-11-18 2023-03-14 南京智程信息科技有限公司 Similar model searching method, device and storage medium
CN116402988A (en) * 2023-05-11 2023-07-07 北京冰河起源科技有限公司 Three-dimensional model processing method, device and storage medium
CN116402988B (en) * 2023-05-11 2023-12-19 北京冰河起源科技有限公司 Three-dimensional model processing method, device and storage medium

Similar Documents

Publication Publication Date Title
CN113961738A (en) Multi-feature casting three-dimensional model retrieval method and device
CN109887015B (en) Point cloud automatic registration method based on local curved surface feature histogram
Kazmi et al. A survey of 2D and 3D shape descriptors
Caelli et al. An eigenspace projection clustering method for inexact graph matching
Zhong Intrinsic shape signatures: A shape descriptor for 3D object recognition
Papadakis et al. PANORAMA: A 3D shape descriptor based on panoramic views for unsupervised 3D object retrieval
US8015125B2 (en) Multi-scale segmentation and partial matching 3D models
CN111080684B (en) Point cloud registration method for point neighborhood scale difference description
Assfalg et al. Content-based retrieval of 3-D objects using spin image signatures
CN111914112B (en) Part CAD model reuse method based on point cloud classification network
CN114677418B (en) Registration method based on point cloud feature point extraction
Cybenko et al. Pattern recognition of 3 d cad objects: Towards an electronic yellow pages of mechanical parts
CN111553409A (en) Point cloud identification method based on voxel shape descriptor
CN108388902B (en) Composite 3D descriptor construction method combining global framework point and local SHOT characteristics
CN105654483A (en) Three-dimensional point cloud full-automatic registration method
CN113516695B (en) Point cloud registration strategy in laser profiler flatness measurement
CN109299301B (en) Three-dimensional model retrieval method based on shape distribution and curvature
Kim et al. Shape distribution-based retrieval of 3D CAD models at different levels of detail
CN116008671A (en) Lightning positioning method based on time difference and clustering
CN115082716A (en) Multi-source point cloud rough matching algorithm for road fine reconstruction
CN112651408B (en) Point-to-point transformation characteristic-based three-dimensional local surface description method and system
CN111597367B (en) Three-dimensional model retrieval method based on view and hash algorithm
Vajramushti et al. Efficient 3D object retrieval using depth images
CN112509018B (en) Quaternion space optimized three-dimensional image registration method
CN115147471A (en) Laser point cloud automatic registration method based on curvature density characteristics

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