CN109472853A - A kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity - Google Patents
A kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity Download PDFInfo
- Publication number
- CN109472853A CN109472853A CN201811368655.9A CN201811368655A CN109472853A CN 109472853 A CN109472853 A CN 109472853A CN 201811368655 A CN201811368655 A CN 201811368655A CN 109472853 A CN109472853 A CN 109472853A
- Authority
- CN
- China
- Prior art keywords
- image
- variance
- window
- normal
- combination
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/04—Texture mapping
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/005—General purpose rendering architectures
Landscapes
- Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
A kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity.Fixed suitable coaxial source of parallel light direction, acquisition one are reconstructed the digitized video on surface;Acquisition image is filtered using average curvature filtering;The filter window for creating n × n is enumerated all half windows by filter window central element and is combined, calculates all combined variances, and selects the smallest half window combination of variance, i.e. brightness of image changes more gentle half window;The smallest combination of variance is selected, all micro- plane direction of normal that may be present are enumerated according to the relationship of brightness of image and reflectogram, all possible direction of normal forms a circular cone around coaxial source of parallel light direction;Variance minimum combination is selected, the method is selected to swear the direction of normal as filter window central element;Sequentially find the corresponding direction of normal of all pixels point;According to Green's function, gauging surface texture relative altitude selects different path integrals and finally calculates its average value.
Description
Technical field
The present invention relates to machine vision body surface reconfiguration techniques, more particularly, to a kind of based on the bright of image irradiation intensity
The reconstructing method of primary body microcosmic surface.
Background technique
It carries out three-dimensionalreconstruction for single image to have practical significance engineer application, shape is stayed in wherein shadow cause
(shape from shading, SFS) is a kind of commonly by body surface brightness caused by the effect of coaxial source of parallel light
Method ([1] Sun W, Chen B, Yao B, et the al.Complex wavelet of its three-dimensional space position of Information recovering
enhanced shape from shading transform for estimating surface roughness of
milled mechanical components[J].Journal of Mechanical Science&Technology,
2017,31(2):823-833;[2] He Yuchao, Sun Weifang, Chen Binqiang wait the cause of complex wavelet domain shadow to stay shape and metal milling surface
Quality evaluation [J] foreign countries electronic measurement technique, 2017 (5)).Shadow cause stay shape method SFS be image objects mechanism mathematically
Inverting.However due to uncontrollable factors such as imaging noises, this method is difficult to carry out correction early period.Curvature filtering is a kind of novel
Image filtering method has obtained more and more concerns ([3] Gong Y, Sbalzarini I F.Curvature Filters
Efficiently Reduce Certain Variational Energies[J].IEEE Transactions on Image
Processing,2017,26(4):1786-1798;[4]Gong Y.Bernstein filter:A new solver for
mean curvature regularized models[C]//IEEE International Conference on
Acoustics,Speech and Signal Processing.IEEE,2016).Obtain image information after post-processing mostly according to
Rely the priori knowledge in operator, therefore it is widely applied and is exerted a certain influence.Compared with conventional method, energy of the present invention
The problems such as mutation of enough effective solution Non-smooth surface curved surfaces.
Summary of the invention
It is a kind of lambert's body microcosmic surface reconstruct side based on image irradiation intensity that the present invention, which needs the technical problem solved,
Method.
The present invention the following steps are included:
1) fixed suitable coaxial source of parallel light direction, acquisition one are reconstructed the digitized video on surface;
2) acquisition image is filtered using average curvature filtering;
3) filter window of n × n is created, wherein n is odd number, enumerates all half-windows by filter window central element
Mouth combination, calculates all combined variances, and selects the smallest half window combination of variance, i.e. brightness of image variation is more gentle
Half window;
In step 3), the specific algorithm of the smallest half window combination of selection variance can are as follows:
Wherein, loc is that the smallest half window of variance combines position, varmFor the variance of m-th of half window combination, N is half
The included number of pixels of window, XnFor half window nth pixel value, μ is half window average value.
4) the smallest combination of variance is selected, is enumerated according to the relationship of brightness of image and reflectogram all that may be present micro- flat
Face direction of normal, all possible direction of normal form a circular cone around coaxial source of parallel light direction;
In step 4), all possible direction of normal forms the tool of a circular cone around coaxial source of parallel light direction
Body algorithm can are as follows:
Wherein, θn=arctan (In/ max (I)) it is pixel InWith the angle between Image Acquisition direction,For with x
The angle of axis positive direction.
5) variance minimum combination is selected, the method is selected to swear the direction of normal as filter window central element, it is specific to calculate
Method are as follows:
6) the corresponding direction of normal of all pixels point is sequentially found;
7) according to Green's function, gauging surface texture relative altitude, selects different path integrals and finally to calculate it flat
Mean value, specific algorithm are as follows:
Wherein, L1、L2、LnIt is any curve that point-to-point (x, y) is fixed from certain, c is integral constant.
The present invention suitable coaxial source of parallel light direction of illumination fixed first, by digital image acquisition apparatus to tested table
Face is shot;Then acquisition image is filtered by average curvature filtering, is retaining the same of picture edge characteristic
The filtering of Shi Jinhang curvature;And then create the filter window (n is odd number) of n × n, it enumerates all by filter window central element
Half window combination calculates all combined pixel brightness variances and the smallest variance is selected to combine;According to brightness of image and instead
The relationship for penetrating figure enumerates all micro- plane direction of normal that may be present;Calculate all direction of normal that may be present and selecting party
Poor minimum combination selects the method to swear the direction of normal as filter window central element;Sequentially find the method arrow of all pixels point
Direction;Finally according to Green's function, each pixel relative altitude information is calculated, selects different paths to be integrated and finally counts
Calculate its average value.If demarcating before carrying out image collection to camera brightness, its absolute altitude information can be further obtained, and
It can carry out relevant surfaces parameter evaluation.The reconstructing method hardware cost is lower, and the reconstruct on surface can be completed, and is appropriate for microcosmic
The reconstruct and measurement on surface.
Detailed description of the invention
Fig. 1 is image-pickup method schematic diagram.It is each to mark in Fig. 1 are as follows: 1, the image obtained;2, industrial camera;3, half
Lens;4, lambert's body surface to be measured face;5, source of parallel light.
Fig. 2 is the digitized video of acquisition.
Fig. 3 is image after curvature filtering.
Fig. 4 is that micro- planar process swears direction schematic diagram.It is each to mark in Fig. 4 are as follows: 1, lambert's body microcosmic surface any light source
Incident direction;2, lambert's body microcosmic surface;3, the circular conical surface of the point vector composition in a certain angle with light source incidence direction.
Half window combination diagram when Fig. 5 is n=3.
Fig. 6 is surface reconstruction result three-dimensional figure.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings.
In order to carry out validation verification to lambert's body microcosmic surface reconstructing method based on image irradiation intensity, metal is carried out
Slabbing experiment.The present invention carries out on VMC650E/850E machining center, and test is using carbide end mill to aviation
Aluminium alloy 7075 carries out end mill processing.Workpiece material is aerolite 7075, and milling mode is climb cutting, and cutter is diameter 60mm
Indexable face milling cutter.Square stock is having a size of 145mm × 92mm × 25mm.Milling parameter: speed of mainshaft 1200r/min, feeding speed
200mm/min is spent, carries out Milling Process along material length direction.Test carries out on VMC650E/850E machining center, feeds
Range is 650mm × 400mm × 500mm, and axial repetitive positioning accuracy is ± 0.003mm.Assuming that the milling surface is one bright
Primary body surface face, then its detailed process is as follows:
Image capturing is carried out to surface is reconstructed using coaxial source of parallel light, structural schematic diagram is as shown in Figure 1.Wherein by
Reconstruct the aerolite 7075 that object is above-mentioned processing.The digitized video of acquisition as shown in Fig. 2, the photo of acquisition contained compared with
Good surface texture information.
For the filtered image of curvature as shown in figure 3, selection average curvature is filtered, the number of iterations is 50 times, bent
Rate filtering has effectively filtered out the noises such as hot-tempered point, while the Edge texture information of image has obtained retaining well.
For lambert's body surface face, when incident light direction is determined with intensity, the irradiation intensity of body surface point and the point
Method arrow is related with the angle of incident light, it is possible to which the brightness based on each pixel of filtered image described previously solves and each picture
The angle of the method arrow and source of parallel light incident direction of lambert's body surface face each point of vegetarian refreshments correspondence.
It can not only determine on lambert's body surface face the direction of normal of any completely by method arrow and the angle in light source incidence direction, such as
Shown in Fig. 4, the incident light vector of any is in angle theta with incident light vector using the point as starting point as indicated with 1 on lambert's body surface face 2
Vector constitute a circular conical surface, as shown in 3 in Fig. 4.Determine the direction of normal of the point on lambert's body surface face, it is also necessary to really
Determine the brightness step direction of the point on image.
Microcosmic surface is reconstructed, it is necessary to consider the subtle texture on surface, then for small around certain point in image
Region, the irradiation intensity variable gradient from this to all directions is variant, thus there are a biggish sides of irradiation intensity gradient
To.
For digitized image, tiny area is sized to 3 × 3 pixels, only considers that half irradiation intensity is higher
The lower Great possibility of half irradiation intensity, including 8 kinds of possible situations as shown in Figure 5, circle indicates pixel, white in figure
Circle indicates that the irradiation intensity of pixel is higher, and black indicates that pixel irradiation intensity is lower, perpendicular to black circles and white circle
Boundary line be direction that irradiation intensity changes greatly.
As it was noted above, establishing 8 kinds of Filtering Templates, as shown in Figure 5.For a point to be processed in image, will filter
The center of template is overlapped with the point, calculates the side of overlapping region image each pixel irradiation intensity value and each corresponding points of Filtering Template
Difference traverses all Filtering Templates and selects the smallest half window combination of variance, and then traverses whole picture and solve whole picture
Edge data, specific algorithm can are as follows:
2 to picturedeep -1 of for i from
2 to picturewide -1 of for j from
SubArea { 1 }=[im (i-1, j-1), im (i-1, j), im (i, j-1), im (i+1, j-1), im (i+1, j), im
(i,j)];
SubArea { 2 }=[im (i-1, j-1), im (i-1, j), im (i-1, j+1), im (i, j-1), im (i, j+1), im
(i,j)];
SubArea { 3 }=[im (i-1, j), im (i-1, j+1), im (i, j+1), im (i+1, j), im (i+1, j+1), im
(i,j)];
SubArea { 4 }=[im (i, j-1), im (i, j+1), im (i+1, j-1), im (i+1, j), im (i+1, j+1), im
(i,j)];
SubArea { 5 }=[im (i-1, j-1), im (i-1, j), im (i-1, j+1), im (i, j-1), im (i+1, j-1),
im(i,j)];
SubArea { 6 }=[im (i-1, j-1), im (i-1, j), im (i-1, j+1), im (i, j+1), im (i+1, j+1),
im(i,j)];
SubArea { 7 }=[im (i+1, j+1), im (i, j+1), im (i+1, j-1), im (i+1, j), im (i+1, j+1),
im(i,j)];
SubArea { 8 }=[im (i-1, j-1), im (i, j-1), im (i+1, j-1), im (i+1, j), im (i+1, j+1),
im(i,j)];
SubArea { 9 }=[im (i-1, j-1), im (i, j-1), im (i+1, j-1), im (i-1, j), im (i, j), im (i
+1,j),…
im(i-1,j+1),im(i,j+1),im(i+1,j+1)];
VarRange=[var (subArea { 1 }), var (subArea { 2 }), var (subArea { 3 }), var
(subArea{4}),...
var(subArea{5}),var(subArea{6}),var(subArea{7}),var(subArea{8},var(…
subarea{9})];
[~, minLoc]=min (varRange);
ThetaRange (i, j)=acos (subArea { minLoc } (1:end-1)/1);
end
end
5) the smallest combination minLoc of variance is selected, enumerating according to the relationship of brightness of image and reflectogram there may be
Micro- plane direction of normal.Its algorithm are as follows:
PhiRange=linspace (0,2pi* ((phiNum-1)/phiNum, phiNum);Wherein phiNum is segmentation
Number
NxRange (1 :)=sin (thetaRange (1) * cos (phiRange));
NyRange (1 :)=sin (thetaRange (1) * sin (phiRange));
NzRange (1 :)=cos (thetaRange (1));
NxRange (2 :)=sin (thetaRange (2) * cos (phiRange));
NyRange (2 :)=sin (thetaRange (2) * sin (phiRange));
NzRange (2 :)=cos (thetaRange (2));
NxRange (3 :)=sin (thetaRange (3) * cos (phiRange));
NyRange (3 :)=sin (thetaRange (3) * sin (phiRange));
NzRange (3 :)=cos (thetaRange (3));
NxRange (4 :)=sin (thetaRange (4) * cos (phiRange));
NyRange (4 :)=sin (thetaRange (4) * sin (phiRange));
NzRange (4 :)=cos (thetaRange (4));
NxRange (5 :)=sin (thetaRange (5) * cos (phiRange));
NyRange (5 :)=sin (thetaRange (5) * sin (phiRange));
NzRange (5 :)=cos (thetaRange (5));
6) variance minimum combination is selected, the method is selected to swear the direction of normal as filter window central element, algorithm are as follows:
nij={ nxij∈nx,nyij∈ny,nzij∈nzs.t.varij=min (var) };
7) all direction of normal are sequentially found;
8) according to Green's function, gauging surface texture relative altitude, selects different path integrals and finally to calculate it flat
Mean value, specific algorithm are as follows:
Wherein, L1、L2、LnIt is any curve that point-to-point (x, y) is fixed from certain, c is integral constant.
9) its surface reconstruction result is as shown in fig. 6, its surface line after image after filtering being reconstructed using the above method
Reason has obtained preferable recovery, and the arch texture after milling tool has also obtained largely retaining.
Claims (3)
1. a kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity, it is characterised in that the following steps are included:
1) fixed suitable coaxial source of parallel light direction, acquisition one are reconstructed the digitized video on surface;
2) acquisition image is filtered using average curvature filtering;
3) filter window of n × n is created, wherein n is odd number, enumerates all half window groups by filter window central element
It closes, calculates all combined variances, select the smallest half window combination of variance, is i.e. brightness of image changes more gentle half-window
Mouthful;
4) the smallest combination of variance is selected, all micro- planar processes that may be present are enumerated according to the relationship of brightness of image and reflectogram
Swear direction, all possible direction of normal forms a circular cone around coaxial source of parallel light direction;
5) variance minimum combination is selected, the method is selected to swear the direction of normal as filter window central element, specific algorithm are as follows:
6) the corresponding direction of normal of all pixels point is sequentially found;
7) according to Green's function, gauging surface texture relative altitude selects different path integrals and finally calculates its average value,
Its specific algorithm are as follows:
Wherein, L1、L2、LnIt is any curve that point-to-point (x, y) is fixed from certain, c is integral constant.
2. a kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity as described in claim 1, it is characterised in that
In step 3), the specific algorithm of the smallest half window combination of selection variance are as follows:
Wherein, loc is that the smallest half window of variance combines position, varmFor the variance of m-th of half window combination, N is half window institute
Include number of pixels, XnFor half window nth pixel value, μ is half window average value.
3. a kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity as described in claim 1, it is characterised in that
In step 4), all possible direction of normal forms the specific algorithm of a circular cone around coaxial source of parallel light direction
Are as follows:
Wherein, θn=arctan (In/ max (I)) it is pixel InWith the angle between Image Acquisition direction,It is square with x-axis
To angle.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811368655.9A CN109472853A (en) | 2018-11-16 | 2018-11-16 | A kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811368655.9A CN109472853A (en) | 2018-11-16 | 2018-11-16 | A kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109472853A true CN109472853A (en) | 2019-03-15 |
Family
ID=65673853
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811368655.9A Pending CN109472853A (en) | 2018-11-16 | 2018-11-16 | A kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109472853A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110554054A (en) * | 2019-09-05 | 2019-12-10 | 西安多维机器视觉检测技术有限公司 | three-dimensional blade breakage detection system of mechanically-clamped indexable turning tool |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103065359A (en) * | 2013-01-14 | 2013-04-24 | 厦门大学 | Optical imaging three-dimensional contour reconstruction system and reconstruction method |
CN104778755A (en) * | 2015-03-27 | 2015-07-15 | 浙江理工大学 | Region-division-based three-dimensional reconstruction method for texture image |
US20150222875A1 (en) * | 2012-07-31 | 2015-08-06 | Consejo Superior De Investigaciones Cientificas (Csic) | Device and method for obtaining densitometric images of objects by a combination of radiological systems and depth-sensing cameras |
CN106875468A (en) * | 2015-12-14 | 2017-06-20 | 深圳先进技术研究院 | Three-dimensional reconstruction apparatus and method |
CN107103637A (en) * | 2017-04-05 | 2017-08-29 | 南京信息工程大学 | One kind enhancing texture power method |
JP2017159444A (en) * | 2016-03-08 | 2017-09-14 | 新日鐵住金株式会社 | Polishing jig and sample polishing method and surface analysis method |
-
2018
- 2018-11-16 CN CN201811368655.9A patent/CN109472853A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150222875A1 (en) * | 2012-07-31 | 2015-08-06 | Consejo Superior De Investigaciones Cientificas (Csic) | Device and method for obtaining densitometric images of objects by a combination of radiological systems and depth-sensing cameras |
CN103065359A (en) * | 2013-01-14 | 2013-04-24 | 厦门大学 | Optical imaging three-dimensional contour reconstruction system and reconstruction method |
CN104778755A (en) * | 2015-03-27 | 2015-07-15 | 浙江理工大学 | Region-division-based three-dimensional reconstruction method for texture image |
CN106875468A (en) * | 2015-12-14 | 2017-06-20 | 深圳先进技术研究院 | Three-dimensional reconstruction apparatus and method |
JP2017159444A (en) * | 2016-03-08 | 2017-09-14 | 新日鐵住金株式会社 | Polishing jig and sample polishing method and surface analysis method |
CN107103637A (en) * | 2017-04-05 | 2017-08-29 | 南京信息工程大学 | One kind enhancing texture power method |
Non-Patent Citations (2)
Title |
---|
何昱超 等: "复小波域影致留形及金属铣削表面质量评估", 《国外电子测量技术》 * |
郭便 等: "基于SFS的加工表面三维形貌重建研究", 《机械制造》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110554054A (en) * | 2019-09-05 | 2019-12-10 | 西安多维机器视觉检测技术有限公司 | three-dimensional blade breakage detection system of mechanically-clamped indexable turning tool |
CN110554054B (en) * | 2019-09-05 | 2021-10-29 | 西安多维机器视觉检测技术有限公司 | Three-dimensional blade breakage detection system of mechanically-clamped indexable turning tool |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111429403B (en) | Automobile gear finished product defect detection method based on machine vision | |
US8581162B2 (en) | Weighting surface fit points based on focus peak uncertainty | |
CN106611160B (en) | Image hair identification method and device based on convolutional neural network | |
CN109557101A (en) | A kind of defect detecting device and method of nonstandard high reflection curve surface work pieces | |
CN108760747A (en) | A kind of 3D printing model surface defect visible detection method | |
CN109085178B (en) | Defect fingerprint accurate online monitoring and feedback method for additive manufacturing | |
CN108550160B (en) | Non-uniform light bar characteristic region extraction method based on light intensity template | |
CN111724341B (en) | Aviation blade air film hole quality detection method based on binocular stereo measurement | |
CN112907527B (en) | Infrared thermal imaging splicing detection method for large-size curved surface test piece | |
Babbar et al. | Comparative study of image matching algorithms | |
CN108280838A (en) | A kind of intermediate plate tooth form defect inspection method based on edge detection | |
US9696265B2 (en) | Computational wafer inspection filter design | |
CN109472853A (en) | A kind of lambert's body microcosmic surface reconstructing method based on image irradiation intensity | |
CN115272256A (en) | Sub-pixel level sensing optical fiber path Gaussian extraction method and system | |
CN110378916A (en) | A kind of TBM image based on multitask deep learning is slagged tap dividing method | |
CN104880803B (en) | Adjusting device and adjusting method for full-focus camera vision field | |
CN209486002U (en) | A kind of defect detecting device of nonstandard high reflection curve surface work pieces | |
WO2020158726A1 (en) | Image processing device, image processing method, and program | |
Chen et al. | Efficient Anti-Glare Ceramic Decals Defect Detection by Incorporating Homomorphic Filtering. | |
CN115115578B (en) | Defect detection method and system in additive manufacturing process | |
CN113592953B (en) | Binocular non-cooperative target pose measurement method based on feature point set | |
Schmugge et al. | Automatic detection of cracks during power plant inspection | |
Mikołajczyk et al. | Camera-Based Automatic System for Tool Measurements and Recognition | |
CN113592960B (en) | Method for screening images containing specific features from multiple images | |
Song et al. | Three-dimensional measurement and defect detection based on single image |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190315 |
|
WD01 | Invention patent application deemed withdrawn after publication |