CN109242955A - A kind of workpiece manufacturing feature automatic identifying method and device based on single image - Google Patents
A kind of workpiece manufacturing feature automatic identifying method and device based on single image Download PDFInfo
- Publication number
- CN109242955A CN109242955A CN201810937987.8A CN201810937987A CN109242955A CN 109242955 A CN109242955 A CN 109242955A CN 201810937987 A CN201810937987 A CN 201810937987A CN 109242955 A CN109242955 A CN 109242955A
- Authority
- CN
- China
- Prior art keywords
- feature
- curved surface
- image
- convex
- shape
- 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.)
- Granted
Links
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a kind of workpiece manufacturing feature automatic identifying method and device based on single image, which comprises obtain a width workpiece image, determine the corresponding reflection map equitation of described image;Converse solved reflection map equitation obtains surface graded and height, rebuilds the three-dimension curved surface of workpiece;The shape feature for obtaining the three-dimension curved surface is split the curved surface according to shape feature;In conjunction with convex and concave feature recognition rule, the convex and concave feature of each segmentation curved surface is obtained.The present invention is independent of CAD 3D model, it only needs on the basis of part single width two dimensional image, the relevant technologies of appliance computer computer aided geometric design can realize the automatic identification to workpiece manufacturing feature, the practical value with higher in terms of robot automatic identification.
Description
Technical field
The invention belongs to based on image feature identification technique field more particularly to a kind of workpiece system based on single image
Make characteristic automatic recognition method and device.
Background technique
Workpiece manufacturing feature identification based on image is one of the typical case of computer vision industrially, the purpose is to
The geometry informations such as shape index and the deflection angle of different classes of workpiece are obtained, workpiece is identified with this, are industry
One of intelligentized development trend of production line.In recent years, with the fast development of software and hardware technology, the manufacture based on image
Feature identification technique has been widely used in the field of machining such as Inverse seeking design, concurrent engineering.
The manufacturing feature recognition methods of comparative maturity at present has based on figure, the method for clue and based on the side of manufacturing recourses
Method and mixed method etc., wherein the method based on figure and clue is referred to as pattern matching method, basic thought is by piece surface
Geometrical property is compared with predefined feature mode, therefrom finds out the region for meeting characteristic boundary mode.It is provided based on manufacture
The recognition methods in source refers to that using mapping building information model, then clustering by surface is manufacturing feature, but such method pair
The recognition effect of complex parts model curved surface features is unsatisfactory.Rule-based and figure hybrid algorithm then requires model surface
Exist at the Curvature varying in region can by the real side of surface segmentation, once but a loaf of bread contain multiple relief regions, then this method
It will be unable to identify these concave-convex manufacturing features.
Features described above recognition methods is required to the CAD 3D model using part as initial input, however in intelligent chemical industry
In factory, robot is the characteristic information and engineering that can not obtain CAD 3D model in the production links such as assembly or secondary operation
Dimension information, at this time in order to from the interpretation part feature of manufacture, it is necessary to a large amount of human assistance and secondary input,
And then the efficiency of enterprise's full-automation production and processing is seriously reduced, increase the production and processing cost of enterprise.In addition, for
Most of Chinese to be engaged in for manufacturing medium-sized and small enterprises, their main business and the source of profit is still processing of investor's raw materials, dress
With with secondary operation and original equipment manufacturer etc., that inevitably will be in face of lacking the CAD model and core design information of part
Difficult situation.Therefore traditional using CAD model is the workpiece shapes characteristic recognition method of foundation for the limitation of such enterprise
Property is with regard to larger.
Therefore, how to get around the conventional thought that manufacturing feature identification is carried out based on CAD design model and carry out feature identification, solution
The certainly manufacturing feature automatic recognition problem in the process of manufacture such as processing of investor's raw materials, secondary assembly, is those skilled in the art's mesh
The technical issues of preceding urgent solution.
Summary of the invention
To overcome above-mentioned the deficiencies in the prior art, the present invention provides a kind of workpiece manufacturing feature based on single image from
Dynamic recognition methods.This method first estimates the light source direction parameter of image using the single image of part as initial input
Meter realizes the three-dimensional reconstruction of workpiece surface accordingly;Then it is special analytical calculation to be carried out to the shape index for rebuilding threedimensional model surface
Cut-off rule is levied, is divided curved surface to obtain corresponding characteristic area using characteristic curve;Finally based on the realization pair of feature recognition rule
Effective identification of workpiece manufacturing feature.This method is independent of CAD 3D model, it is only necessary in part single width two dimensional image
On the basis of, the relevant technologies of appliance computer computer aided geometric design can realize the automatic identification to workpiece manufacturing feature, in machine
Practical value with higher in terms of device people automatic identification.
To achieve the above object, the present invention adopts the following technical scheme:
A kind of workpiece manufacturing feature automatic identifying method based on single image, comprising the following steps:
A width workpiece image is obtained, determines the corresponding reflection map equitation of described image;
Converse solved reflection map equitation obtains surface graded and height, rebuilds the three-dimension curved surface of workpiece;
The shape feature for obtaining the three-dimension curved surface is split the curved surface according to shape feature;
In conjunction with convex and concave feature recognition rule, the convex and concave feature of each segmentation curved surface is obtained.
Further, the reflection map equitation determination the following steps are included:
Determine reflectogram function, the argument of function includes light source polar angle, light source azimuth angle and surface graded;
Light source polar angle and light source azimuth angle are estimated according to the pixel value on image;
With surface graded for variable, linearization process is carried out to reflectogram function, determines the reflectogram of target entity imaging
Equation.
Further, the Converse solved reflection map equitation, obtaining surface graded and height includes:
Pass through backward finite difference method Approximation Discrete body surface gradient;
Certain point carries out Taylor expansion using reflection map equitation on corresponding given image, obtains height using correlation criteria
The iterative formula of degree;
The height value of entire described image is iteratively solved using the iterative formula.
Further, the shape feature for obtaining the three-dimension curved surface, divides the curved surface according to shape feature
It cuts and includes:
The shape feature of three-dimension curved surface each point according to curvature estimation;
The characteristic curve of segmentation different type curved surface is obtained according to the shape feature;
The curved surface is split based on the characteristic curve.
Further, the shape feature is shape index, and the different type curved surface is by three kinds of different shape indexes
It determines jointly:
Kt=1+3 (1+sgn (Kh, e))+(1-sgn (Kg, e))
Wherein, L=puuN, M=puVN, N=pVVN,F=pu·pv,puWith
pvIt is the first-order derivative of parametric surface, puu、pvvAnd puvIt is the second derivative of parametric surface,
Further, the characteristic curve for obtaining segmentation different type curved surface includes:
Calculate the scalar domain with shape index equivalence;
The parameter curve of the scalar domain and plane Z=0 intersection is mapped back into curved surface, i.e. characteristic curve.
Further, the concave-convex recognition rule are as follows: assuming that siWith sjIndicate two inter-related local surfaces, piWith
pjRespectively represent siWith sjCentral point, niWith njRespectively represent piWith pjPractical normal vector;And it is required that: (1) piWith pjTwo
Person is located in some coordinate system;(2)niWith njIt is all unit vector;(3) normal vector direction is surfaces facing outward;
Remember dij=pi-pj, α=(ni, dij), γ=(nj,dij), if meet the following conditions :-γ >=0 α, then siAnd sjTool
There is localized indentation relationship, forms recessed feature;If eligible :-γ≤0 α, then siAnd sjWith Local Convex relationship, convex spy is formed
Sign;
The Compound Punch feature that three and three or more local surfaces are constituted:
With C0、C1Shape made of successional pit, paddy and planar area set wherein at least includes one
Pit type region is compound recessed feature;
With C0、C1Shape made of successional peak, ridge and planar area set wherein at least includes a peak type
Region is Convex Composite feature;
Wherein, C0Continuity refers to that two regions are connected;C1Continuity refers to that two region first differentials are continuous, either
It is tangent continuous.
Further, the method also includes: the identical segmentation curved surface of convex and concave feature is merged according to connection relationship.
Second purpose according to the present invention the present invention also provides a kind of computing device, including memory, processor and is deposited
The computer program that can be run on a memory and on a processor is stored up, when the processor executes described program described in realization
The workpiece manufacturing feature automatic identifying method based on single image.
Third purpose according to the present invention, the present invention also provides a kind of computer readable storage mediums, are stored thereon with
Computer program realizes the workpiece manufacturing feature automatic identification based on single image when the program is executed by processor
Method.
Beneficial effects of the present invention
1, the present invention is based on single width two dimensional images to carry out feature identification, easy to operate, and efficiency is higher.This method does not depend on
Design a model in CAD 3D, for be engaged in processing of investor's raw materials, secondary operation and original equipment manufacturer type medium-sized and small enterprises have compared with
High practical value.
2, present invention employs the elevations angle and azimuth that the eight neighborhood of pixel point carrys out partial estimation light source, to traditional illumination mould
Type is improved, so that it can apply to the three-dimensional reconstruction for having under complex illumination environment.
3, the present invention does not need the detection device of manual intervention and valuableness, it is only necessary on the basis of single image, application
The relevant technologies of Computer-aided Geometric Design can realize the automatic identification to workpiece manufacturing feature, at low cost.
4, present invention introduces various shapes indexes, the concave-convex type of curved surface are covered comprehensively, in conjunction with the convex and concave feature of setting
Recognition rule can be realized workpiece convex and concave feature and accurately define.
Detailed description of the invention
The accompanying drawings constituting a part of this application is used to provide further understanding of the present application, and the application's shows
Meaning property embodiment and its explanation are not constituted an undue limitation on the present application for explaining the application.
Fig. 1 is the flow chart of entire flow of the present invention;
Fig. 2 is system coordinates schematic diagram;
Fig. 3 is that schematic diagram is chosen in direction;
Fig. 4 is that the three-dimensional reconstruction of practical mouse gray level image compares, wherein Fig. 4 (a) is original-gray image, Fig. 4 (b)
The three-dimensional reconstruction effect of traditional SFS algorithm, Fig. 4 (c) improve the three-dimensional reconstruction effect of SFS algorithm;
Fig. 5 is two examples of curved surface segmentation;
Fig. 6 is that convex and concave feature identifies example, wherein Fig. 6 (a) is original two dimensional image, and Fig. 6 (b) is that convex and concave feature is known
Other effect picture;
Fig. 7 is concaveconvex shape tagsort;
Fig. 8 is the CAD design model of ring flange;
Fig. 9 is the formative method of ring flange;
Figure 10 is that the feature of ring flange identifies tree;
Figure 11 is convex and concave feature recognition result of this paper algorithm to ring flange part;
Figure 12 mold convex and concave feature identification, wherein Figure 12 (a) is original-gray image, Figure 12 (b) convex and concave feature identification knot
Fruit;
Figure 13 is that the convex and concave feature of U-type groove part identifies, wherein the original two dimensional image (b) of Figure 13 (a) U-type groove part
U-type groove part convex and concave feature recognition result.
Specific embodiment
It is noted that described further below be all exemplary, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms used herein has logical with the application person of an ordinary skill in the technical field
The identical meanings understood.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular shape
Formula be also intended to include plural form, additionally, it should be understood that, when in the present specification use term "comprising" and/or
When " comprising ", existing characteristics, step, operation, device, component and/or their combination are indicated.
In the absence of conflict, the features in the embodiments and the embodiments of the present application can be combined with each other.
Embodiment one
Present embodiment discloses a kind of workpiece manufacturing feature automatic identifying method based on single image, as shown in Figure 1, tool
Body comprises the steps of:
Step 1: obtaining a width workpiece image, determine the corresponding reflection map equitation of described image.
Step 1.1: a width workpiece image is obtained, using the luminous environment of illumination model simulated scenario;
The three-dimensional rebuilding method selected herein is the method for shape from shading (shape from shading, SFS),
Its main thought is the direction and relative altitude for restoring body surface each point so that the gray value of single image is foundation.
In SFS problem, the reflection equation based on Lambert surface is as follows:
Wherein, E (x, y) is the gray value at pixel (x, y), and I (x, y) is the intensity of incident light, and p is reflectivity, (p,
Q, -1) it is the surface normal indicated by gradient, (ps, qs, -1) be light source incident direction.
Traditional SFS method is to simplify problem generally to do image-forming condition, optical signature etc. following hypothesis: light source is nothing
Limit the point light source of distant place or the directional light of Uniform Illumination;Reflection model is lambert's body surface face reflection model;Imaging geometry closes
System is rectangular projection.But by production scene strong noise, the influence of the factors such as low illumination and equipment vibration, the hypothesis of this harshness
Condition can not have a simulation well to actual conditions, thus cannot accurate, the three-dimensional reconstruction knot with universality
Fruit has seriously affected the precision of subsequent manufacturing feature identification.It therefore is to make this method closer to actual work and experimental ring
Border will improve the hypothesis of light source problem in traditional SFS method herein.
In the research to SFS problem, usually body surface height is expressed as z=f (x, y), formula (1) can be abbreviated as more
The brightness constraint equation of abstract general type, it may be assumed that
E (x, y)=R (p, q) (2)
Wherein, E (x, y) is the gray value at pixel (x, y), and R is reflectogram function, and p and q are that apparent height value z is closed
In x, the partial derivative of y.ρ is reflectivity, and (p, q, -1) is the surface normal indicated by gradient, (ps, qs, -1) and it is entering for light source
Penetrate direction.
Formula (1) is rewritten as following form:
In formula (3), n is surface normal vector;For light source vector direction;And θs
Respectively light source polar angle and azimuth, as shown in Figure 2.In image procossing, ρ and I are generally set to constant.
Light source polar angleIt can estimate to obtain by formula (4):
Wherein: factor alphaiAre as follows: a0=0.5673 a1=0.6230 a2=0.1901 a3=-0.6314 a4=-0.5430
a5=0.8982 a6=0.3534 a7=-0.5001
Light source azimuth angle θsIt can estimate to obtain by formula (5):
Here EX, yO operation is indicated the image mean value that all pixels point calculates thereon after pretreatment;N is the direction number chosen, direction selection side
Method is as shown in figure 3, selecting 8 neighborhood points in this algorithm is increment direction, δ IiFor (δ xi, δ yi) grey scale change value on direction.
So
Estimating illumination parameterAnd θsLater, we can seek the gradient (p, q) and height of body surface.With
Conventional method is compared, and improved method is no longer simply light source direction to be uniformly assumed to be (0,0, -1), but false first
If the distribution of body surface method arrow is consistent in three dimensions, then using the eight neighborhood point of pixel on image come part
Estimate the elevation angle and azimuth of light source, the estimation parameter about light source direction is finally obtained using statistical method.This
Method, which reduces, irrelevantly to be assumed certain priori conditions and limits the limitation of SFS application in conventional method, make this three-dimensional
The method of reconstruction can be more flexible be suitable for different environment.
Step 1.2: linearization process being carried out to reflectogram function using surface graded, determines the reflection of target entity imaging
Figure equation.
Surface graded (p, q) is used as variable, linearization process is carried out to reflectogram function using this variable, assumes item
Part plays a major role for the lower term in reflectogram, therefore directly omits the nonlinear terms in formula (1) about p and q, thus
Obtain following linear shape from shading:
Wherein:
Step 2: Converse solved reflection map equitation obtains surface graded or height, rebuilds the three-dimension curved surface of workpiece.
Step 2.1: passing through backward finite difference method Approximation Discrete body surface gradient (p, q), that is, makeM, N are respectively the row of discrete picture
Several and columns.Above-mentioned difference, which is substituted into formula (6), and carries out appropriate deformation can be obtained following formula:
Wherein:
(8) h ≠ 0 in, that is, ps≠qs.For formula (7), if it is known that boundary condition zI, 0, z0, j, it can it solves
The height value of surface every bit.
Step 2.2: the reflection map equitation of image being deformed, certain point utilizes reflection on corresponding given gray level image
Figure equation carries out Taylor expansion, finally obtains the iterative formula (9) of height using correlation criteria:
Wherein ω is relaxation factor,For zI, jNth iteration as a result, and have:
For the convergence for guaranteeing relaxative iteration format, the value range of relaxation factor ω are as follows: 0 < ω < 2.For iteration
Initial value generally assume that for
Step 2.3: the height value of entire 2-D gray image can be iteratively solved out using formula (9), realized to part
Three-dimension curved surface reconstruct.
Fig. 4 gives the three-dimensional reconstruction effect picture of practical mouse gray level image, can be seen by the comparison of Fig. 4 (b) and (c)
The 3 d effect graph that the innovatory algorithm of this paper reconstructs out has good raising in continuity and slickness, and surface profile is clear
Spend higher, the more true to nature three-dimensional appearance for having reappeared mouse.Table 1 shows that improved SFS algorithm and tradition SFS algorithm exist
Error on mouse images three-dimensional reconstruction and the time-related comparison needed for algorithm.Although improved SFS algorithm is being located
It manages on the time slightly slowly, but is greatly reduced the error of reconstruct, so that average error value reduces.
Table 1 improves SFS algorithm compared with tradition SFS algorithm is to the performance of mouse three-dimensional reconstruction
Step 3: obtaining the shape feature of the three-dimension curved surface, the curved surface is split according to shape feature;
The feature of machine components also has certain general character with function-specific property simultaneously, if these general character parts are designed
For Share Model, data information redundancy can be greatly reduced, played a multiplier effect.And convex-concave machining feature is machining
The basic machining feature of part is the bridge connected between low level geometric description and feature towards specific area, is
The preferred features of Share Model are developed, therefore are identified herein mainly for the convex and concave feature in geometric characteristic.
It needs to identify manufacturing feature on the basis of the threedimensional model of reconstruct below.Any point on space curved surface
Curvature be to describe the important attribute of threedimensional model shape, a concave-convex degree for curved surface where it is reflected has invariable rotary
Property and translation invariance.Gaussian curvature, average curvature contain the shape information of curved surface, are calculated based on the two curvature
Different relief region, so that it may the basic research method as Three-dimension object recognition.Certainly if it is considered that part material direction
Influence, can be improved really convex and concave feature identification details and precision.However, there remains manual intervention and detection device,
Efficiency is too low, and for robot automatic identification, does not need to identify so subtle feature.Therefore herein according to
The shape index of flexometer calculating curved surface, so that it may easily identify the concaveconvex shape structure of curved surface.
Step 3.1: the shape index of curved surface is calculated according to flexometer;
In three dimensions, a discrete parameter curved surface can be typically expressed as:
P=p (u, v)=[u, v, f (u, v)]T, u=1 ..., m;V=1 ..., n. (11)
The shape index of the present embodiment can choose KgOr Kh:
Wherein, L=puuN, M=puVN, N=pVVN,F=pu·pv,puWith
pvIt is the first-order derivative of parametric surface, puu、pvvAnd puvIt is the second derivative of parametric surface.
KgAnd KhFor what is defined by the two of curved surface kind citation form.
To make this method that there is more pinpoint accuracy and stability, another shape index K may be incorporated intot, calculate public
Formula are as follows:
Kt=1+3 (1+sgn (Kh, e))+(1-sgn (Kg, e)) (13)
Wherein:
Step 3.2: the characteristic curve of segmentation different zones need to be obtained using shape index;
According to shape index KgAnd KhPositive and negative property, generally there are the expressions of the region of 8 seed types, as shown in table 2.
8 kinds of surface types that table 2 is determined by shape index
Curved surface area type | Kg | Kh | Kt |
Pit | + | + | 7 |
Peak/top surface | + | - | 1 |
Mountain valley/paddy face | 0 | + | 8 |
Ridge/crestal surface | 0 | - | 2 |
Saddle paddy | - | + | 9 |
Saddle ridge | - | - | 3 |
It is minimum/isotropic | - | 0 | 6 |
Plane | 0 | 0 | 5 |
Affiliated area can be judged according to the positive and negative property that each region up-samples dot shape index after the completion of curved surface segmentation
Type.Such as: any K on certain region after segmentationgValue is timing, then the region may be pit or peak type region.This
When, it is also necessary to judge point KhIt is positive and negative, if KhValue be the canonical region be pit type;It otherwise, is peak type.Assuming that p (u, v)
It is a C2Continuous regular parameter curved surface, then meet K on curved surfacegThe trajectory line of=0 point is referred to as parabolic curve, it will
Curved surface is divided into shape index KgTwo parts region greater than 0 and less than 0.Wushu (13) substitutes into formula (12) and obtains
Since p=p (u, v) is regular surface, then
Therefore, formula (14) can be further simplified according to formula (15), it is desirable that shape index Kg=0, it is equivalent to calculate scalar
Domain:
Ψ=(puu·(pu×pv))(pvv·(pu×pv))-(puv·(pu×pv))2=0 (16)
With KgIt is similar, pass through abbreviation shape index Kh, available and KhThere is the scalar domain Φ of identical null solution collection:
Φ=(puu·(pu×pv))|pv|2-2(puv·(pu×pv))pu·pv+(pvv·(pu×pv))|pu|2 (17)
The parameter curve of scalar domain Ψ and Φ and plane Z=0 intersection is mapped back on curved surface p=p (u, v) now, can be obtained
The indicatrix that former curved surface is divided into different zones is obtained, segmentation example is as shown in Figure 5.
Step 3.3: curved surface being split by these characteristic curves.
Only region segmentation table corresponding with master mould is obtained by these different zones and its mutual topological relation
Show.
Step 4: in conjunction with convex and concave feature recognition rule, defining the convex and concave feature of each segmentation curved surface.
The identification of curved surface convex and concave feature is based on defined below:
Assuming that: siWith sjIndicate two inter-related local surfaces, piWith pjRespectively represent siWith sjCentral point, niWith
njRespectively represent piWith pjPractical normal vector.And it is required that: (1) piWith pjThe two is located in some coordinate system;(2)niWith
njIt is all unit vector;(3) normal vector is surfaces facing outward.
Define 1 localized indentation relationship, siAnd sjThe two can be respectively in connection with piWith pjAnd its niWith njThe corresponding description of expansion, note
dij=pi-pj, α=(ni, dij), γ=(nj, dij), if meet the following conditions :-γ >=0 α, then siAnd sjIt is closed with localized indentation
System, forms recessed feature, corresponding shape index Kg>=0, Kh> 0.
2 Local Convex relationships are defined, and localized indentation relationship initial description having the same, if eligible :-γ≤0 α,
Then siAnd sjWith Local Convex relationship, convex feature, corresponding shape index K are formedg>=0, Kh< 0.
Convex and concave feature recognition rule defined above is based on two surfaces, but if there are three and three or more
The Compound Punch feature that surface is constituted then is needed using following recognition rule:
C in following definition0Continuity refer to two regions be connected or position be it is continuous, it is this continuously to merely ensure that
Between curved surface without gap but completely attach to;C1Continuity refers to that two region first differentials are continuous or tangent continuous
's.
Defining 3 compound recessed features is a kind of with C0、C1Made of successional pit, paddy and planar area set
Shape wherein at least includes a pit type region, corresponding shape index Kh> 0.
It defines 4 Convex Composites and is characterized in that one kind has C0、C1Shape made of successional peak, ridge and planar area set,
It wherein at least include a peak type region, corresponding shape index Kh< 0.
It defines 5 Interims and belongs to shape-aided feature, 3 kinds of fin fillet, concave rib corners and fillets situations can be divided into,
Primarily serve the effect of separation.It is composed of following 2 groups of curvature domains:
(1) comprising the convex transition combination at peak, saddle or ridge curvature domain;
(2) comprising the recessed transition combination of pit, saddle or paddy type curvature domain.
Step 5: the identical segmentation curved surface of convex and concave feature is merged according to connection relationship.
The curved surface of same type can form area region not of uniform size after polymerization according to connection relationship, obtain mould
The local surface feature of type, it is also necessary to these regions be connected according to certain rules and merged.Here region growing is used
Method from region indicate model in identify manufacturing feature.Specific step is as follows:
(1) using any one not visited patch as starting point, its adjacent curved surface piece is searched for, if the curved surface of access
Piece and starting patch possess identical recessed (convex) characterizing definition, then are fused in initiation region, which is arranged
To have accessed.
(2) if all patch have all been accessed, search terminates, and otherwise turns to (1) step.Until never again
Until qualified patch.
Embodiment two
The purpose of the present embodiment is to provide a kind of computing device.
A kind of workpiece manufacturing feature automatic identification equipment based on single image, including memory, processor and be stored in
On memory and the computer program that can run on a processor, the processor realize following steps when executing described program,
Include:
A width workpiece image is obtained, determines the corresponding reflection map equitation of described image;
Converse solved reflection map equitation obtains surface graded and height, rebuilds the three-dimension curved surface of workpiece;
The shape feature for obtaining the three-dimension curved surface is split the curved surface according to shape feature;
In conjunction with convex and concave feature recognition rule, the convex and concave feature of each segmentation curved surface is obtained.
Embodiment three
The purpose of the present embodiment is to provide a kind of computer readable storage medium.
A kind of computer readable storage medium, is stored thereon with computer program, execution when which is executed by processor
Following steps:
A width workpiece image is obtained, determines the corresponding reflection map equitation of described image;
Converse solved reflection map equitation obtains surface graded and height, rebuilds the three-dimension curved surface of workpiece;
The shape feature for obtaining the three-dimension curved surface is split the curved surface according to shape feature;
In conjunction with convex and concave feature recognition rule, the convex and concave feature of each segmentation curved surface is obtained.
Each step involved in above embodiments two and three is corresponding with embodiment of the method one, and specific embodiment can be found in
The related description part of embodiment one.Term " computer readable storage medium " is construed as including one or more instructions
The single medium or multiple media of collection;It should also be understood as including any medium, any medium can be stored, be encoded
Or it carries instruction set for being executed by processor and processor is made either to execute in the present invention method.
Verification result:
By taking the model of a machine components as an example, as shown in fig. 6, Fig. 6 (a) is original two dimensional image, Fig. 6 (b) is bumps
Feature recognition effect figure.Thus figure can be seen that the relief region that this machine components model is successfully identified using this paper algorithm, altogether
In respect of 3 concave regions (region shown in alphabetical a) and 4 convex regions (region shown in alphabetical b), the wherein face a1 central point
Shape index Kg=1.0845e-04, Kh=0.0206, Kt=8, according to table 2, this face is mountain valley concave surface, and remaining surface is then
Non- characteristic area.
Common concaveconvex shape feature is specifically classified as shown in Figure 7.
A ring flange model (such as Fig. 8) is designed in CAD, using traditional characteristic recognition methods and this paper recognition methods into
Row compares.In former CAD model, the formative method used includes (such as Fig. 9) such as stretching, excision, mirror image and fillets, this
A little information are all the operations towards Geometric Modeling, are not the manufacturing features of part, therefore are needed in identification process artificial secondary
The feature and dimensional parameters information of part are inputted, using traditional manufacturing feature recognition methods, system reflects these design features
Penetrate into the processing and manufacturing such as corresponding slot, hole, step, chamfering feature (such as Figure 10), the precision identified under conditions of human assistance
It is higher.Figure 11 is the recognition result using this paper algorithm to this ring flange part, and identification process does not need manual intervention, it is only necessary to
Want the single width two dimensional image of part that can identify that (wherein the shape at four circular holes refers to for the concave-convex manufacturing feature of this ring flange
Number Kh is all larger than zero as recessed feature, and middle is a Convex Composite feature), efficiency is higher, but the precision identified is to be improved.
For the effect for further verifying methods described herein, the knowledge of convex and concave feature has been carried out to two part images respectively
Not.Wherein Figure 12 is the original-gray image and convex and concave feature recognition result of a mold, the white wheel as shown in Figure 12 (b)
It is the compound recessed feature as made of pit, paddy and planar area set in profile.
Figure 13 gives the partial results during a U-type groove Parts Recognition, and table 3 gives each label in Figure 13 (b)
The principal shape characteristic value of sampled point.It may recognize that the part of each sampled point is concave-convex special in conjunction with table 3 and convex and concave feature recognition rule
Property, 1 position of sampled point is recessed feature;2 position of sampled point is Convex Composite feature, but close to plane;3 position of sampled point is multiple
Convex feature is closed, 4 position of sampled point is notch feature.It can thus be concluded that based on three-dimensional reconstruction feature recognition result and bumps in kind
Structure matches very much, also sufficiently demonstrates the validity of this paper algorithm.
The shape index feature list of each label point in 3 Figure 13 (b) of table
Beneficial effects of the present invention
1, the present invention is based on single width two dimensional images to carry out feature identification, easy to operate, and efficiency is higher.This method does not depend on
Design a model in CAD 3D, for be engaged in processing of investor's raw materials, secondary operation and original equipment manufacturer type medium-sized and small enterprises have compared with
High practical value.
2, present invention employs the elevations angle and azimuth that the eight neighborhood of pixel point carrys out partial estimation light source, to traditional illumination mould
Type is improved, so that it can apply to the three-dimensional reconstruction for having under complex illumination environment.
3, the present invention does not need the detection device of manual intervention and valuableness, it is only necessary on the basis of single image, application
The relevant technologies of Computer-aided Geometric Design can realize the automatic identification to workpiece manufacturing feature, at low cost.
4, present invention introduces various shapes indexes, the concave-convex type of curved surface are covered comprehensively, in conjunction with the convex and concave feature of setting
Recognition rule can be realized workpiece convex and concave feature and accurately define.
It will be understood by those skilled in the art that each module or each step of aforementioned present invention can be filled with general computer
It sets to realize, optionally, they can be realized with the program code that computing device can perform, it is thus possible to which they are stored
Be performed by computing device in the storage device, perhaps they are fabricated to each integrated circuit modules or by they
In multiple modules or step be fabricated to single integrated circuit module to realize.The present invention is not limited to any specific hardware
With the combination of software.
The foregoing is merely preferred embodiment of the present application, are not intended to limit this application, for the skill of this field
For art personnel, various changes and changes are possible in this application.Within the spirit and principles of this application, made any
Modification, equivalent replacement, improvement etc., should be included within the scope of protection of this application.
Above-mentioned, although the foregoing specific embodiments of the present invention is described with reference to the accompanying drawings, not protects model to the present invention
The limitation enclosed, those skilled in the art should understand that, based on the technical solutions of the present invention, those skilled in the art
The various modifications or changes that can be made are not needed to make the creative labor still within protection scope of the present invention.
Claims (10)
1. a kind of workpiece manufacturing feature automatic identifying method based on single image, which comprises the following steps:
A width workpiece image is obtained, determines the corresponding reflection map equitation of described image;
Converse solved reflection map equitation obtains surface graded and height, rebuilds the three-dimension curved surface of workpiece;
The shape feature for obtaining the three-dimension curved surface is split the curved surface according to shape feature;
In conjunction with convex and concave feature recognition rule, the convex and concave feature of each segmentation curved surface is obtained.
2. a kind of workpiece manufacturing feature automatic identifying method based on single image as described in claim 1, which is characterized in that
The determination of the reflection map equitation the following steps are included:
Determine reflectogram function, the argument of function includes light source polar angle, light source azimuth angle and surface graded;
Light source polar angle and light source azimuth angle are estimated according to the pixel value on image;
With surface graded for variable, linearization process is carried out to reflectogram function, determines the reflection map equitation of target entity imaging.
3. a kind of workpiece manufacturing feature automatic identifying method based on single image as described in claim 1, which is characterized in that
The Converse solved reflection map equitation, obtaining surface graded and height includes:
Pass through backward finite difference method Approximation Discrete body surface gradient;
Certain point carries out Taylor expansion using reflection map equitation on corresponding given image, obtains height using correlation criteria
Iterative formula;
The height value of entire described image is iteratively solved using the iterative formula.
4. a kind of workpiece manufacturing feature automatic identifying method based on single image as described in claim 1, which is characterized in that
The shape feature for obtaining the three-dimension curved surface, is split the curved surface according to shape feature and includes:
The shape feature of three-dimension curved surface each point according to curvature estimation;
The characteristic curve of segmentation different type curved surface is obtained according to the shape feature;
The curved surface is split based on the characteristic curve.
5. a kind of workpiece manufacturing feature automatic identifying method based on single image as claimed in claim 4, which is characterized in that
The shape feature is shape index, and the different type curved surface is determined jointly by three kinds of different shape indexes:
Wherein, L=puuN, M=puVN, N=pVVN,F=pu·pv,puAnd pvIt is
The first-order derivative of parametric surface, puu、pvvAnd puvIt is the second derivative of parametric surface,
6. a kind of workpiece manufacturing feature automatic identifying method based on single image as claimed in claim 5, which is characterized in that
It is described obtain segmentation different type curved surface characteristic curve include:
Calculate the scalar domain with shape index equivalence;
The parameter curve of the scalar domain and plane Z=0 intersection is mapped back into curved surface, i.e. characteristic curve.
7. a kind of workpiece manufacturing feature automatic identifying method based on single image as claimed in claim 5, which is characterized in that
The bumps recognition rule are as follows: assuming that siWith sjIndicate two inter-related local surfaces, piWith pjRespectively represent siWith sj's
Central point, niWith njRespectively represent piWith pjPractical normal vector;And it is required that: (1) piWith pjThe two is located at some coordinate system
In the middle;(2)niWith njIt is all unit vector;(3) normal vector direction is surfaces facing outward;
Remember dij=pi-pj, α=(ni, dij), γ=(nj, dij), if meet the following conditions :-γ >=0 α, then siAnd sjWith office
The recessed relationship in portion, forms recessed feature;If eligible :-γ≤0 α, then siAnd sjWith Local Convex relationship, convex feature is formed;
The type of affiliated area is judged according to the positive and negative property of pixel shape index on each region.Such as: the segmentation area Hou Mou
The K put on domaingValue is timing, then the region may be pit or peak type region.At this time, it is also necessary to judge K a littlehIt is positive and negative, if Kh
Value be the canonical region be pit type;It otherwise, is peak type.
The Compound Punch feature that three and three or more local surfaces are constituted:
With C0、C1Shape made of successional pit, paddy and planar area set wherein at least includes a pit type
Region is compound recessed feature;
With C0、C1Shape made of successional peak, ridge and planar area set wherein at least includes a peak type region,
For Convex Composite feature;
Wherein, C0Continuity refers to that two regions are connected;C1Continuity refers to that two region first differentials are continuous or tangent
Continuously.
8. a kind of workpiece manufacturing feature automatic identifying method based on single image as described in claim 1, which is characterized in that
The method also includes: the identical segmentation curved surface of convex and concave feature is merged according to connection relationship.
9. a kind of computing device including memory, processor and stores the calculating that can be run on a memory and on a processor
Machine program, which is characterized in that the processor is realized when executing described program as claim 1-8 is described in any item based on single
The workpiece manufacturing feature automatic identifying method of width image.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
Such as claim 1-8 described in any item workpiece manufacturing feature automatic identifying methods based on single image are realized when execution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810937987.8A CN109242955B (en) | 2018-08-17 | 2018-08-17 | Workpiece manufacturing characteristic automatic identification method and device based on single image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810937987.8A CN109242955B (en) | 2018-08-17 | 2018-08-17 | Workpiece manufacturing characteristic automatic identification method and device based on single image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109242955A true CN109242955A (en) | 2019-01-18 |
CN109242955B CN109242955B (en) | 2023-03-24 |
Family
ID=65069702
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810937987.8A Active CN109242955B (en) | 2018-08-17 | 2018-08-17 | Workpiece manufacturing characteristic automatic identification method and device based on single image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109242955B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110659652A (en) * | 2019-09-10 | 2020-01-07 | 东华大学 | Characteristic matching detection system of Creo model of fan device |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012071688A1 (en) * | 2010-12-03 | 2012-06-07 | 中国科学院自动化研究所 | Method for analyzing 3d model shape based on perceptual information |
WO2015154601A1 (en) * | 2014-04-08 | 2015-10-15 | 中山大学 | Non-feature extraction-based dense sfm three-dimensional reconstruction method |
CN105354353A (en) * | 2015-09-28 | 2016-02-24 | 武汉开目信息技术有限责任公司 | MBD model based processing feature identification and modeling method |
CN108154066A (en) * | 2016-12-02 | 2018-06-12 | 中国科学院沈阳自动化研究所 | A kind of Three-dimensional target recognition method based on curvature feature recurrent neural network |
-
2018
- 2018-08-17 CN CN201810937987.8A patent/CN109242955B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2012071688A1 (en) * | 2010-12-03 | 2012-06-07 | 中国科学院自动化研究所 | Method for analyzing 3d model shape based on perceptual information |
US20140125663A1 (en) * | 2010-12-03 | 2014-05-08 | Institute of Automation, Chinese Academy of Scienc | 3d model shape analysis method based on perception information |
WO2015154601A1 (en) * | 2014-04-08 | 2015-10-15 | 中山大学 | Non-feature extraction-based dense sfm three-dimensional reconstruction method |
CN105354353A (en) * | 2015-09-28 | 2016-02-24 | 武汉开目信息技术有限责任公司 | MBD model based processing feature identification and modeling method |
CN108154066A (en) * | 2016-12-02 | 2018-06-12 | 中国科学院沈阳自动化研究所 | A kind of Three-dimensional target recognition method based on curvature feature recurrent neural network |
Non-Patent Citations (3)
Title |
---|
吴伟仁等: "基于月面单幅图像的软着陆障碍识别与安全区选取方法", 《深空探测学报》 * |
张宁等: "基于Phong模型的多幅图像SFS算法研究", 《计算机技术与发展》 * |
胡彭勇: "基于球面调和函数的人脸识别", 《合肥学院学报(自然科学版)》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110659652A (en) * | 2019-09-10 | 2020-01-07 | 东华大学 | Characteristic matching detection system of Creo model of fan device |
CN110659652B (en) * | 2019-09-10 | 2023-08-18 | 东华大学 | Feature matching detection system of fan device Creo model |
Also Published As
Publication number | Publication date |
---|---|
CN109242955B (en) | 2023-03-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Hf-neus: Improved surface reconstruction using high-frequency details | |
Ji et al. | A novel simplification method for 3D geometric point cloud based on the importance of point | |
CN108305289B (en) | Three-dimensional model symmetry characteristic detection method and system based on least square method | |
CN111738265B (en) | Semantic segmentation method, system, medium, and electronic device for RGB-D image | |
CN106530247B (en) | A kind of multi-scale image restorative procedure based on structural information | |
Nie | Extracting feature lines from point clouds based on smooth shrink and iterative thinning | |
CN110208769B (en) | Ray tracing method and system based on NURBS curved surface | |
CN114494570A (en) | Rendering method and device of three-dimensional model, storage medium and computer equipment | |
CN108230452B (en) | Model hole filling method based on texture synthesis | |
Hu et al. | Geometric feature enhanced line segment extraction from large-scale point clouds with hierarchical topological optimization | |
CN113963138A (en) | Complete and accurate extraction method of three-dimensional laser point cloud characteristic point line | |
Liu et al. | A novel rock-mass point cloud registration method based on feature line extraction and feature point matching | |
Wang et al. | Window normalization: enhancing point cloud understanding by unifying inconsistent point densities | |
CN109816767B (en) | Fuzzy matching method for layered household-division diagrams of three-dimensional building model houses based on tangent space | |
CN109242955A (en) | A kind of workpiece manufacturing feature automatic identifying method and device based on single image | |
Takashima et al. | Recognition of free-form features for finite element meshing using deep learning | |
CN116721230A (en) | Method, device, equipment and storage medium for constructing three-dimensional live-action model | |
Ji et al. | Point cloud segmentation for complex microsurfaces based on feature line fitting | |
Zhang et al. | Building façade element extraction based on multidimensional virtual semantic feature map ensemble learning and hierarchical clustering | |
Frías et al. | Exploiting BIM objects for synthetic data generation toward indoor point cloud classification using deep learning | |
CN114186603B (en) | Multi-source model single fusion method and system, storage medium and terminal | |
Xiao et al. | Debsdf: Delving into the details and bias of neural indoor scene reconstruction | |
CN115147471A (en) | Laser point cloud automatic registration method based on curvature density characteristics | |
CN110533781B (en) | Automatic labeling method for multi-class three-dimensional model components | |
Fu et al. | Feature line extraction from point clouds based on geometric structure of point space |
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 |