CN110060335A - There are the virtual reality fusion methods of mirror article and transparent substance in a kind of scene - Google Patents
There are the virtual reality fusion methods of mirror article and transparent substance in a kind of scene Download PDFInfo
- Publication number
- CN110060335A CN110060335A CN201910332095.XA CN201910332095A CN110060335A CN 110060335 A CN110060335 A CN 110060335A CN 201910332095 A CN201910332095 A CN 201910332095A CN 110060335 A CN110060335 A CN 110060335A
- Authority
- CN
- China
- Prior art keywords
- light source
- point
- value
- transparent substance
- photon
- 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
- 239000000126 substance Substances 0.000 title claims abstract description 67
- 238000007500 overflow downdraw method Methods 0.000 title claims abstract description 8
- 238000005286 illumination Methods 0.000 claims abstract description 51
- 230000000694 effects Effects 0.000 claims abstract description 19
- 230000004927 fusion Effects 0.000 claims abstract description 16
- 238000009877 rendering Methods 0.000 claims abstract description 7
- 239000000463 material Substances 0.000 claims abstract description 4
- 238000000034 method Methods 0.000 claims description 35
- 230000000007 visual effect Effects 0.000 claims description 20
- 238000005457 optimization Methods 0.000 claims description 16
- 230000005855 radiation Effects 0.000 claims description 15
- 238000005070 sampling Methods 0.000 claims description 11
- 239000003518 caustics Substances 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000002310 reflectometry Methods 0.000 claims description 5
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000003709 image segmentation Methods 0.000 claims description 3
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 238000013507 mapping Methods 0.000 claims 1
- 230000000149 penetrating effect Effects 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 2
- 230000003287 optical effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 4
- 239000003550 marker Substances 0.000 description 3
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 230000011514 reflex Effects 0.000 description 1
- 238000009738 saturating Methods 0.000 description 1
- 239000011800 void material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/50—Lighting effects
-
- 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
- G06T19/006—Mixed reality
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Graphics (AREA)
- Software Systems (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- Image Generation (AREA)
Abstract
There are the virtual reality fusion method category computer virtual reality technology fields of mirror article and transparent substance in a kind of scene, using the shooting of RGB-D camera, there are the scenes of mirror article and transparent substance first by the present invention, it identifies the position of mirror article and transparent substance, and three-dimensional reconstruction is carried out to scene.The reflection between object is considered in primary light source estimation, and has estimated the material parameters of mirror article and transparent substance, is carried out difference rendering using the illumination result and model parameter estimated, is obtained virtual reality fusion effect picture.The present invention solves the problems, such as the virtual reality fusion illumination consistency in scene there are mirror article and transparent substance to obtain virtual reality fusion effect more true to nature by estimation mirror article BRDF model parameter, the refractive index of transparent substance and color attenuation coefficient.Meanwhile the present invention when estimating light source position from optical principle, it is contemplated that the reflection case between object has obtained more accurate light source position.
Description
Technical field
The invention belongs to computer virtual reality technology fields, and in particular to light source position and intensity, mirror surface in a kind of scene
The reflection coefficient of object and the estimation method of transparent substance refractive index, color attenuation coefficient.
Background technique
Augmented reality is that the dummy object of generation is combined by computer to be presented on user with actual scene
At the moment, for more life-like virtual reality fusion effect, it is necessary to show dummy object and be imitated with the consistent illumination of actual scene
Fruit.Illumination consistency mainly consider the problems of be in scene real light sources and real-world object to dummy object surface patches institute
Caused by color and brightness change.
Existing solution illumination consistency virtual reality fusion method is broadly divided into three classes: by auxiliary sign object space method, by
Ancillary equipment method and without marker or the method for ancillary equipment.Wherein auxiliary sign object is shade and the mark manually placed
Object obtains the light conditions in real scene by the information of marker offer.Ancillary equipment has depth camera, light-field camera
And the special capture apparatus such as fisheye camera is capable of providing the information such as depth, light field and full view image, provides for illumination estimation
New solution.Illumination in scene is obtained by way of image analysis without the method for marker or ancillary equipment to believe
Breath.
What existing illumination estimation method considered is only the shade pair of real-world object in real light sources or scene in scene
Variation caused by dummy object surface patches, for the caustic hot spot of mirror article and transparent substance generation in actual scene
Influence to dummy object does not account for but.
Summary of the invention
The purpose of the present invention is being directed to the deficiency of existing method, proposing one kind, there are mirror articles and saturating suitable for scene
The illumination consistency virtual reality fusion method of bright object, solves the caustic phenomenon pair of mirror article and transparent substance in actual scene
The influence of dummy object, technical solution used by this method is:
1.1 using RGB-D camera shooting there are the scenes of mirror article and transparent substance, obtain different perspectives depth image
And color image;Three-dimensional reconstruction is carried out to scene, and obtains the threedimensional model position of mirror article and transparent substance, including following
Step:
1.1.1 three-dimensional reconstruction is carried out using depth image of the KinectFusion algorithm to different perspectives, obtains truncation symbol
Number distance function (TSDF) model and camera posture;
1.1.2 the depth image for using different perspectives, identifies the approximate region of mirror article and transparent substance, by the area
Domain respectively splits mirror article and transparent substance from color image as initial position, in conjunction with image segmentation algorithm,
Three-dimensional reconstruction is carried out to mirror article and transparent substance using visual shell method;
1.1.3 TSDF model is merged with mirror article and transparent substance model;
1.2 primary light source positions and intensity estimation:
Primary light source estimation does not consider mirror article and transparent substance model, and remaining model material assumes that as lambert's table
Face, reflection coefficient value are 1;K point light source is evenly distributed on the hemisphere face centered on object scene, hemispherical diameter
For 2 times of hemispherical diameter for just surrounding scene;Each point light source emits the photon of q different directions into scene, tool
Body estimation method includes the following steps:
1.2.1 each photon energy emitted from j-th of point light source is calculated:
Wherein: ΔΦ (ωp) it is energy entrained by each photon, IjFor the intensity value of j-th of point light source;
1.2.2 the photon that k point light source emits is tracked respectively and by point of impingement coordinate, incident photon energy and incident photon
Direction is stored in k photon figure;
1.2.3 the reflected radiation brightness L of point x under any visual angle is calculatedr(x):
Wherein: n is the photon numbers collected near point x using photon figure obtained in step 1.2.2;D (x) is point x
At a distance from photon farthest in n photon being collected into;
1.2.4 by the energy of each photon of j-th of the point light source found out in step 1.2.1 transmitting, step 1.2.3 is substituted into
The reflected radiation brightness-formula of midpoint x obtains j-th of point light source in point x reflected radiation brightness Lrj(x):
Wherein: dj (x) be using j-th of photon figure under j-th of point light source in point x and n photon being collected into farthest
The distance of photon;
1.2.5 the color image of different perspectives collected in step 1.1 is converted into gray level image;
1.2.6 by the reflected radiation brightness L of the gray value of gray level image and step 1.2.4rj(x) objective function is formed:
Wherein: m is the gray level image quantity for participating in calculating;SiIt (x) is pixel grey scale of the i-th width gray level image at point x
Value;djiIt (x) is the lower i-th width gray level image midpoint x of j-th of point light source at a distance from photon farthest in n photon being collected into;k
It is point light source quantity on hemisphere;Objective function is solved using non-negative linearity least square method, makes Ei(x) minimum, obtain I1,
I2,...Ij... Ik;
1.2.7 light source estimated result I1,I2... Ij...IkOptimization include the following steps:
1.2.7.1 in I1,I2... Ij...IkIn select the maximum light source L of intensity value1, will be with L1Adjacent non-zero intensities
The intensity of value light source is added to L1The intensity value a of light source1On, by L1And a1It is added in illumination estimation results set;
1.2.7.2 if intensity value is all 0 in remaining light sources, optimization terminates, output illumination estimated result set;
If it is maximum 1.2.7.3 to select intensity value in remaining light sources there are the light source that intensity value is not 0 in remaining light sources
Light source Lm, will be with LmThe intensity of adjacent non-zero intensities value light source is added to LmThe intensity value a of light sourcemOn;By LmAnd amIt is added to light
By estimate in results set;
If 1.2.7.4 am<0.5a1, optimization terminates, output illumination estimated result set;
If 1.2.7.5 am≥0.5a1, by LmAnd amIt is added in illumination estimation results set, goes to step 1.2.7.2;
Mirror article obtained in 1.3 couples of step 1.1.2 and transparent substance model distinguish, including the following steps:
1.3.1 the model generated in a step 1.1.2 is chosen, and is assumed to be Lambert surface, reflection coefficient is taken as 1, makes
With light source position and intensity in the illumination estimation results set estimated, it is rendered under different perspectives, obtains difference
The sum of the grayscale values of the object, is respectively as follows: b under visual angle1, b2... bh, wherein h is the different perspectives number for participating in calculating;Again by mould
Type is assumed to be transparent substance, and refractive index value is 1.2, using light source position in the illumination estimation results set estimated and by force
Degree, it is rendered under identical h visual angle, the sum of the grayscale values of the object under different perspectives is obtained, is respectively as follows: t1,
t2... th;IfThen object is mirror article, wherein ciIt is object under i-th of visual angle in step
The sum of corresponding pixel points gray value on the gray level image that rapid 1.2.5 is generated;IfThen object is
Transparent substance;
The isotropism Ward bidirectional reflectance distribution function (Ward BRDF) of the optimization of 1.4 light source positions and mirror article is joined
Number estimation, includes the following steps:
1.4.1 each light source position near zone estimated is sampled on hemisphere face, each light source is nearby adopted
Sampling point number is g, and for each sampled point as a sampling point light source, sampled point intensity of light source value is all illumination in step 1.2.7
Corresponding intensity of light source value in estimated result set;
1.4.2 under the corresponding each sampling point light source of each light source in illumination estimation results set, on mirror article
The reflected radiation brightness of point xAre as follows:
Wherein: s is the number of light source in illumination estimation results set;I is intensity of light source value;I is point light source to the direction point x
Vector;For the angle where the vector and point x of d-th of point light source to the direction point x between the normal of plane;f(ρd,ρs,σ)
For isotropism Ward BRDF model, expression formula are as follows:
Wherein: o is the vector of direction of visual lines;Half-angle vector (h=(i+o)/| i+o |) of the h between vector i and o,
WithAngle respectively where the vector sum half-angle vector of direction of visual lines and point x between the normal of plane;ρdIt is irreflexive
Reflectivity;ρsFor the reflectivity of mirror-reflection;σ is roughness parameter;It is solved using branch and bound method and Second-order cone programming excellent
Change problem:
Obtain corresponding ρ when optimal solutiond、ρs, σ and e;Wherein: M is by the corresponding mirror of grayscale image obtained in step 1.2.5
Column vector M=[the M of the pixel value composition of face object1 M2 ... MN]T;It is reflected at difference for mirror article
The column vector of radiance composition
1.4.3 to g sampling point light source near the s light source and corresponding each light source in illumination estimation results set,
Mirror article Ward BRDF model parameter is estimated using the method for step 1.4.2 respectively, obtains (g+1)sGroup ρd、ρs、σ
With the value of e;Corresponding light source position and ρ when e value minimumd、ρsWith the light source position and mirror article Ward that σ is after optimization
BRDF model parameter estimation value;
The refractive index and color attenuation coefficient of 1.5 transparent substances are estimated, including the following steps:
1.5.1 using the light source position and intensity after optimizing in step 1.4.3, transparent substance is rendered, is reflected using photon
The rendering mode penetrated only changes transparent substance refractive index in scene;Refractive index changes to 2 from 1.2, can recognize that with human eye
The smallest variations in refractive index 0.01 that bright object caustic effect changes is step-length increase, calculates corresponding field under different refractivity
Scape gray value and z1, z2... z80;Estimate the calculation formula of refractive index are as follows:
S.t.i=1,2 ... 80
Wherein: μ is the sum of grayscale image respective pixel value obtained in step 1.2.5, the corresponding refractive index of calculated i value
For transparent substance refractive index;
1.5.2 transparent substance color attenuation coefficient σr、σgAnd σbCalculation formula are as follows:
Wherein: σr、σgAnd σbRespectively red, green, blue channel attenuation coefficient;H is the pixel sum for participating in calculating;diFor
The transmission range of light;WithIt is in d respectivelyiWhen=0, the refractive index and step estimated in step 1.5.1 is used
1.4.3 the red, green, blue channel gray value that the light source position and intensity estimated in renders transparent substance;ri、gi
And biThe red, green, blue channel gray value of the color image respectively shot;
1.6 carry out difference rendering using the illumination result and model parameter that estimate, obtain virtual reality fusion effect picture.
The features of the present invention and beneficial effect
Compared with existing algorithm, the present invention not only considers light source directly to the shadow of object when primary light source is estimated
It rings, reflex of the light between object is also simulated, obtained more accurate initial illumination estimation result.
By estimating Ward BRDF model parameter, the refractive index of transparent substance and the color attenuation coefficient of mirror article, well solve
Influence of the caustic hot spot that mirror article and transparent substance in actual scene generate to dummy object.
Detailed description of the invention
Fig. 1 is that there are the processes of the illumination consistency virtual reality fusion method of mirror article and transparent substance suitable for scene
Figure
Fig. 2 is that there are mirror article virtual reality fusion experiment effect figures in scene
Fig. 3 is that there are transparent substance virtual reality fusion experiment effect figures in scene
In Fig. 2 and Fig. 3: (a) indicating actual scene image, (b) indicate to utilize the effect after the method for the present invention virtual reality fusion
Figure
Specific embodiment
Core of the invention content is: considering the reflection between object in primary light source estimation, has obtained more
Accurate estimated result.Mirror article Ward BRDF model parameter, the refractive index of transparent substance and color attenuation coefficient are carried out
Estimation, and light source position is optimized simultaneously.It renders to have obtained using light source and model parameter the progress difference estimated and more force
Genuine virtual reality fusion effect.
To make the purpose of the present invention, technical solution and advantage are clearer, and with reference to the accompanying drawing and example is done further
Narration in detail:
1.1 using RGB-D camera shooting there are the scenes of mirror article and transparent substance, obtain different perspectives depth image
And color image;Three-dimensional reconstruction is carried out to scene, and obtains the threedimensional model position of mirror article and transparent substance, including following
Step:
1.1.1 three-dimensional reconstruction is carried out using depth image of the KinectFusion algorithm to different perspectives, obtains truncation symbol
Number distance function (TSDF) model and camera posture;
1.1.2 the depth image for using different perspectives, identifies the approximate region of mirror article and transparent substance, by the area
Domain respectively splits mirror article and transparent substance from color image as initial position, in conjunction with image segmentation algorithm,
Three-dimensional reconstruction is carried out to mirror article and transparent substance using visual shell method;
1.1.3 TSDF model is merged with mirror article and transparent substance model;
1.2 primary light source positions and intensity estimation:
Primary light source estimation does not consider mirror article and transparent substance model, and remaining model material assumes that as lambert's table
Face, reflection coefficient value are 1;K point light source is evenly distributed on the hemisphere face centered on object scene, hemispherical diameter
For 2 times of hemispherical diameter for just surrounding scene;Each point light source emits the photon of q different directions into scene, tool
Body estimation method includes the following steps:
1.2.1 each photon energy emitted from j-th of point light source is calculated:
Wherein: ΔΦ (ωp) it is energy entrained by each photon, IjFor the intensity value of j-th of point light source;
1.2.2 the photon that k point light source emits is tracked respectively and by point of impingement coordinate, incident photon energy and incident photon
Direction is stored in k photon figure;
1.2.3 the reflected radiation brightness L of point x under any visual angle is calculatedr(x):
Wherein: n is the photon numbers collected near point x using photon figure obtained in step 1.2.2;D (x) is point x
At a distance from photon farthest in n photon being collected into;
1.2.4 by the energy of each photon of j-th of the point light source found out in step 1.2.1 transmitting, step 1.2.3 is substituted into
The reflected radiation brightness-formula of midpoint x obtains j-th of point light source in point x reflected radiation brightness Lrj(x):
Wherein: dj(x) for using j-th of photon figure under j-th of point light source in point x and n photon being collected into farthest
The distance of photon;
1.2.5 the color image of different perspectives collected in step 1.1 is converted into gray level image;
1.2.6 by the reflected radiation brightness L of the gray value of gray level image and step 1.2.4rj(x) objective function is formed:
Wherein: m is the gray level image quantity for participating in calculating;SiIt (x) is pixel grey scale of the i-th width gray level image at point x
Value; djiIt (x) is the lower i-th width gray level image midpoint x of j-th of point light source at a distance from photon farthest in n photon being collected into;k
It is point light source quantity on hemisphere;Objective function is solved using non-negative linearity least square method, makes Ei(x) minimum, obtain I1,
I2,...Ij... Ik;
1.2.7 light source estimated result I1,I2... Ij...IkOptimization include the following steps:
1.2.7.1 in I1,I2... Ij...IkIn select the maximum light source L of intensity value1, will be with L1Adjacent non-zero intensities
The intensity of value light source is added to L1The intensity value a of light source1On, by L1And a1It is added in illumination estimation results set;
1.2.7.2 if intensity value is all 0 in remaining light sources, optimization terminates, output illumination estimated result set;
If it is maximum 1.2.7.3 to select intensity value in remaining light sources there are the light source that intensity value is not 0 in remaining light sources
Light source Lm, will be with LmThe intensity of adjacent non-zero intensities value light source is added to LmThe intensity value a of light sourcemOn;By LmAnd amIt is added to light
By estimate in results set;
If 1.2.7.4 am<0.5a1, optimization terminates, output illumination estimated result set;
If 1.2.7.5 am≥0.5a1, by LmAnd amIt is added in illumination estimation results set, goes to step 1.2.7.2;
Mirror article obtained in 1.3 couples of step 1.1.2 and transparent substance model distinguish, including the following steps:
1.3.1 the model generated in a step 1.1.2 is chosen, and is assumed to be Lambert surface, reflection coefficient is taken as 1, makes
With light source position and intensity in the illumination estimation results set estimated, it is rendered under different perspectives, obtains difference
The sum of the grayscale values of the object, is respectively as follows: b under visual angle1, b2... bh, wherein h is the different perspectives number for participating in calculating;Again by mould
Type is assumed to be transparent substance, and refractive index value is 1.2, using light source position in the illumination estimation results set estimated and by force
Degree, it is rendered under identical h visual angle, the sum of the grayscale values of the object under different perspectives is obtained, is respectively as follows: t1,
t2... th;IfThen object is mirror article, wherein ciIt is object under i-th of visual angle in step
The sum of corresponding pixel points gray value on the gray level image that rapid 1.2.5 is generated;IfThen object is
Transparent substance;
The isotropism Ward bidirectional reflectance distribution function (Ward BRDF) of the optimization of 1.4 light source positions and mirror article is joined
Number estimation, includes the following steps:
1.4.1 each light source position near zone estimated is sampled on hemisphere face, each light source is nearby adopted
Sampling point number is g, and for each sampled point as a sampling point light source, sampled point intensity of light source value is all illumination in step 1.2.7
Corresponding intensity of light source value in estimated result set;
1.4.2 under the corresponding each sampling point light source of each light source in illumination estimation results set, on mirror article
The reflected radiation brightness of point xAre as follows:
Wherein: s is the number of light source in illumination estimation results set;I is intensity of light source value;I is point light source to the direction point x
Vector;For the angle where the vector and point x of d-th of point light source to the direction point x between the normal of plane;f(ρd,ρs,
It σ) is isotropism Ward BRDF model, expression formula are as follows:
Wherein: o is the vector of direction of visual lines;Half-angle vector (h=(i+o)/| i+o |) of the h between vector i and o,
WithAngle respectively where the vector sum half-angle vector of direction of visual lines and point x between the normal of plane;ρdIt is irreflexive
Reflectivity;ρsFor the reflectivity of mirror-reflection;σ is roughness parameter;It is solved using branch and bound method and Second-order cone programming excellent
Change problem:
Obtain corresponding ρ when optimal solutiond、ρs, σ and e;Wherein: M is by the corresponding mirror of grayscale image obtained in step 1.2.5
Column vector M=[the M of the pixel value composition of face object1 M2 ... MN]T;It is anti-at difference for mirror article
Penetrate the column vector of radiance composition
1.4.3 to g sampling point light source near the s light source and corresponding each light source in illumination estimation results set,
Mirror article Ward BRDF model parameter is estimated using the method for step 1.4.2 respectively, obtains (g+1)sGroup ρd、ρs、σ
With the value of e;Corresponding light source position and ρ when e value minimumd、ρsWith the light source position and mirror article Ward that σ is after optimization
BRDF model parameter estimation value;
The refractive index and color attenuation coefficient of 1.5 transparent substances are estimated, including the following steps:
1.5.1 using the light source position and intensity after optimizing in step 1.4.3, transparent substance is rendered, is reflected using photon
The rendering mode penetrated only changes transparent substance refractive index in scene;Refractive index changes to 2 from 1.2, can recognize that with human eye
The smallest variations in refractive index 0.01 that bright object caustic effect changes is step-length increase, calculates corresponding field under different refractivity
Scape gray value and z1, z2... z80;Estimate the calculation formula of refractive index are as follows:
S.t.i=1,2 ... 80
Wherein: μ is the sum of grayscale image respective pixel value obtained in step 1.2.5, the corresponding refractive index of calculated i value
For transparent substance refractive index;
1.5.2 transparent substance color attenuation coefficient σr、σgAnd σbCalculation formula are as follows:
Wherein: σr、σgAnd σbRespectively red, green, blue channel attenuation coefficient;H is the pixel sum for participating in calculating;diFor
The transmission range of light;WithIt is in d respectivelyiWhen=0, the refractive index and step estimated in step 1.5.1 is used
1.4.3 the red, green, blue channel gray value that the light source position and intensity estimated in renders transparent substance;ri、gi
And biThe red, green, blue channel gray value of the color image respectively shot;
1.6 carry out difference rendering using the illumination result and model parameter that estimate, obtain virtual reality fusion effect picture.
With specific test, to verify one kind provided by the invention, there are mirror articles and transparent suitable for scene below
The feasibility of the illumination consistency virtual reality fusion method of object.The primary light source estimated result of the method for the present invention and Chen are proposed
Only consider a light source illumination estimation algorithm of object contributions is compared, and illustrate in scene there are mirror article and thoroughly
The virtual reality fusion effect picture of bright object (test sample is shot by RGB-D camera).
1. operating condition:
Experiment porch of the invention uses Intel Core i74.2GHz CPU 4.20GHz 4.20GHz, inside saves as
16GB, runs the PC machine of Windows 7, and programming language is MATLAB language and C Plus Plus.
2. experiment content and interpretation of result:
Table 1 is that estimation primary light source algorithm carries out the illumination estimation method of object contributions with light source is only considered in the present invention
Comparison, wherein error angle is that light-source angle and the corner dimension of light-source angle estimated, unit are degree in actual scene,
The accuracy of the light source position estimated is evaluated with it, the method for the present invention error angle compares bibliography as can be seen from Table 1
Error angle in method lower 8.2 ± 7.3.
As shown in Fig. 2, Fig. 2 (a) is real scene object, the box among desktop is mirror-reflection object, is shown on desktop
Show its caustic hot spot effect.Fig. 2 (b) is using the method for the present invention virtual reality fusion effect picture, and wherein arrow meaning is void
Quasi- object.It can be seen that influence of the mirror article to dummy object of real scene from the hot spot in Fig. 2 (b) on dummy object.
As shown in figure 3, Fig. 3 (a) is real scene object, the caustic hot spot effect of transparent substance is shown on desktop.Fig. 3
(b) for using the method for the present invention virtual reality fusion effect picture, wherein arrow meaning is dummy object.The dummy object from Fig. 3 (b)
On hot spot can be seen that influence of the transparent substance to dummy object of real scene.
The error angle (unit: degree) of 1 illumination estimation result of table
The method of the present invention | Bibliography method | |
Error angle | 11.4±2.7 | 19.6±10 |
The experimental results showed that, the present invention passes through estimation mirror article Ward BRDF model parameter, transparent substance by above
Refractive index and color attenuation coefficient solve in scene that there are mirror articles to obtain virtual reality fusion effect more true to nature
With the virtual reality fusion illumination consistency problem of transparent substance.Meanwhile the present invention when estimating light source position from optical principle,
The reflection case between object is considered, more accurate light source position has been obtained, this is also relative to other illumination estimation methods
Superior place.
Claims (1)
1. there are the virtual reality fusion methods of mirror article and transparent substance in a kind of scene, it is characterised in that include the following steps:
1.1 using RGB-D camera shooting there are the scenes of mirror article and transparent substance, obtain different perspectives depth image and coloured silk
Chromatic graph picture;Three-dimensional reconstruction is carried out to scene, and obtains the threedimensional model position of mirror article and transparent substance, including following step
It is rapid:
1.1.1 three-dimensional reconstruction is carried out to the depth image of different perspectives using KinectFusion algorithm, obtain unblind away from
From function (TSDF) model and camera posture;
1.1.2 the depth image for using different perspectives, identifies the approximate region of mirror article and transparent substance, which is made
For initial position, mirror article and transparent substance are split in conjunction with image segmentation algorithm respectively from color image, is used
Visual shell method carries out three-dimensional reconstruction to mirror article and transparent substance;
1.1.3 TSDF model is merged with mirror article and transparent substance model;
1.2 primary light source positions and intensity estimation:
Primary light source estimation does not consider mirror article and transparent substance model, and remaining model material assumes that as Lambert surface, instead
Penetrating coefficient value is 1;K point light source is evenly distributed on the hemisphere face centered on object scene, and hemispherical diameter is rigid
2 times for surrounding the hemispherical diameter of scene well;Each point light source emits the photon of q different directions into scene, specifically estimates
Meter method includes the following steps:
1.2.1 each photon energy emitted from j-th of point light source is calculated:
Wherein: ΔΦ (ωp) it is energy entrained by each photon, IjFor the intensity value of j-th of point light source;
1.2.2 the photon that k point light source emits is tracked respectively and by point of impingement coordinate, incident photon energy and incident photon direction
It is stored in k photon figure;
1.2.3 the reflected radiation brightness L of point x under any visual angle is calculatedr(x):
Wherein: n is the photon numbers collected near point x using photon figure obtained in step 1.2.2;D (x) is point x and receives
The distance of farthest photon in the n photon collected;
1.2.4 by the energy of each photon of j-th of the point light source found out in step 1.2.1 transmitting, the midpoint step 1.2.3 is substituted into
The reflected radiation brightness-formula of x obtains j-th of point light source in point x reflected radiation brightness Lrj(x):
Wherein: dj(x) for using j-th of photon figure under j-th of point light source farthest photon in point x and n photon being collected into
Distance;
1.2.5 the color image of different perspectives collected in step 1.1 is converted into gray level image;
1.2.6 by the reflected radiation brightness L of the gray value of gray level image and step 1.2.4rj(x) objective function is formed:
Wherein: m is the gray level image quantity for participating in calculating;SiIt (x) is grey scale pixel value of the i-th width gray level image at point x;dji
It (x) is the lower i-th width gray level image midpoint x of j-th of point light source at a distance from photon farthest in n photon being collected into;K is hemisphere
Upper point light source quantity;Objective function is solved using non-negative linearity least square method, makes Ei(x) minimum, obtain I1,I2,
...Ij...Ik;
1.2.7 light source estimated result I1,I2... Ij...IkOptimization include the following steps:
1.2.7.1 in I1,I2... Ij...IkIn select the maximum light source L of intensity value1, will be with L1Adjacent non-zero intensities value light
The intensity in source is added to L1The intensity value a of light source1On, by L1And a1It is added in illumination estimation results set;
1.2.7.2 if intensity value is all 0 in remaining light sources, optimization terminates, output illumination estimated result set;
If 1.2.7.3 there are the light sources that intensity value is not 0 in remaining light sources, the maximum light source of intensity value is selected in remaining light sources
Lm, will be with LmThe intensity of adjacent non-zero intensities value light source is added to LmThe intensity value a of light sourcemOn;By LmAnd amIt is added to illumination to estimate
It counts in results set;
If 1.2.7.4 am<0.5a1, optimization terminates, output illumination estimated result set;
If 1.2.7.5 am≥0.5a1, by LmAnd amIt is added in illumination estimation results set, goes to step 1.2.7.2;
Mirror article obtained in 1.3 couples of step 1.1.2 and transparent substance model distinguish, including the following steps:
1.3.1 the model that generates in a step 1.1.2 is chosen, and is assumed to be Lambert surface, reflection coefficient is taken as 1, using estimating
Light source position and intensity in the illumination estimation results set counted out render it under different perspectives, obtain different perspectives
The sum of the grayscale values of the lower object, is respectively as follows: b1, b2... bh, wherein h is the different perspectives number for participating in calculating;Again by model vacation
It is set as transparent substance, refractive index value is 1.2, right using light source position and intensity in the illumination estimation results set estimated
It is rendered under identical h visual angle, is obtained the sum of the grayscale values of the object under different perspectives, is respectively as follows: t1, t2... th;
IfThen object is mirror h body, wherein ciIt is object under i-th of visual angle in step 1.2.5
The sum of corresponding pixel points gray value on the gray level image of generation;IfThen object is transparency
Body;
Isotropism Ward bidirectional reflectance distribution function (Ward BRDF) parameter of the optimization of 1.4 light source positions and mirror article is estimated
Meter, includes the following steps:
1.4.1 each light source position near zone estimated is sampled on hemisphere face, sampled point near each light source
Number is g, and for each sampled point as a sampling point light source, sampled point intensity of light source value is all illumination estimation in step 1.2.7
Corresponding intensity of light source value in results set;
1.4.2 under the corresponding each sampling point light source of each light source in illumination estimation results set, point x on mirror article
Reflected radiation brightnessAre as follows:
Wherein: s is the number of light source in illumination estimation results set;I is intensity of light source value;I be point light source to point x direction to
Amount;For the angle where the vector and point x of d-th of point light source to the direction point x between the normal of plane;f(ρd,ρs, σ) and it is each
To same sex Ward BRDF model, expression formula are as follows:
Wherein: o is the vector of direction of visual lines;Half-angle vector (h=(i+o)/| i+o |) of the h between vector i and o,With
Angle respectively where the vector sum half-angle vector of direction of visual lines and point x between the normal of plane;ρdFor irreflexive reflection
Rate;ρsFor the reflectivity of mirror-reflection;σ is roughness parameter;Optimization is solved using branch and bound method and Second-order cone programming to ask
Topic:
Obtain corresponding ρ when optimal solutiond、ρs, σ and e;Wherein: M is by the corresponding mirror of grayscale image obtained in step 1.2.5
Column vector M=[the M of the pixel value composition of body1 M2 ... MN]T;Spoke is reflected at difference for mirror article
Penetrate the column vector of brightness composition
1.4.3 to g sampling point light source near the s light source and corresponding each light source in illumination estimation results set, respectively
Mirror article WardBRDF model parameter is estimated using the method for step 1.4.2, obtains (g+1)sGroup ρd、ρs, σ and e
Value;Corresponding light source position and ρ when e value minimumd、ρsWith the light source position and mirror article WardBRDF model ginseng that σ is after optimization
Number estimated value;
The refractive index and color attenuation coefficient of 1.5 transparent substances are estimated, including the following steps:
1.5.1 using the light source position and intensity after optimizing in step 1.4.3, transparent substance is rendered, Photon Mapping is used
Rendering mode only changes transparent substance refractive index in scene;Refractive index changes to 2 from 1.2, can recognize that transparency with human eye
The smallest variations in refractive index 0.01 that body caustic effect changes is step-length increase, calculates corresponding scene ash under different refractivity
Angle value and z1, z2... z80;Estimate the calculation formula of refractive index are as follows:
S.t.i=1,2 ... 80
Wherein: μ is the sum of grayscale image respective pixel value obtained in step 1.2.5, and the corresponding refractive index of calculated i value is
Bright object refractive index;
1.5.2 transparent substance color attenuation coefficient σr、σgAnd σbCalculation formula are as follows:
Wherein: σr、σgAnd σbRespectively red, green, blue channel attenuation coefficient;H is the pixel sum for participating in calculating;diFor light
Transmission range;WithIt is in d respectivelyiWhen=0, the refractive index and step 1.4.3 estimated in step 1.5.1 is used
In the light source position that estimates and intensity red, green, blue channel gray value that transparent substance is rendered;ri、giAnd biPoint
The red, green, blue channel gray value for the color image that Wei do not shoot;
1.6 carry out difference rendering using the illumination result and model parameter that estimate, obtain virtual reality fusion effect picture.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910332095.XA CN110060335B (en) | 2019-04-24 | 2019-04-24 | Virtual-real fusion method for mirror surface object and transparent object in scene |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910332095.XA CN110060335B (en) | 2019-04-24 | 2019-04-24 | Virtual-real fusion method for mirror surface object and transparent object in scene |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110060335A true CN110060335A (en) | 2019-07-26 |
CN110060335B CN110060335B (en) | 2022-06-21 |
Family
ID=67320396
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910332095.XA Expired - Fee Related CN110060335B (en) | 2019-04-24 | 2019-04-24 | Virtual-real fusion method for mirror surface object and transparent object in scene |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110060335B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110689514A (en) * | 2019-10-11 | 2020-01-14 | 深圳大学 | Training method and computer equipment for new visual angle synthetic model of transparent object |
CN110717979A (en) * | 2019-08-19 | 2020-01-21 | 北京航空航天大学 | Atmospheric and three-dimensional earth surface coupling radiation simulation method based on photon tracking |
CN111028597A (en) * | 2019-12-12 | 2020-04-17 | 塔普翊海(上海)智能科技有限公司 | Mixed reality foreign language scene, environment and teaching aid teaching system and method thereof |
CN113593049A (en) * | 2021-07-27 | 2021-11-02 | 吉林大学 | Virtual-real fusion method for geometric consistency of real object and virtual object in scene |
WO2022142311A1 (en) * | 2020-12-30 | 2022-07-07 | 浙江商汤科技开发有限公司 | Method for segmenting target object in image, three-dimensional reconstruction method, and related apparatus |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130044213A1 (en) * | 2011-08-18 | 2013-02-21 | Massachusetts Institute Of Technology | System and method for diffuse imaging with time-varying illumination intensity |
US20140055570A1 (en) * | 2012-03-19 | 2014-02-27 | Fittingbox | Model and method for producing 3d photorealistic models |
CN106803267A (en) * | 2017-01-10 | 2017-06-06 | 西安电子科技大学 | Indoor scene three-dimensional rebuilding method based on Kinect |
US20180047208A1 (en) * | 2016-08-15 | 2018-02-15 | Aquifi, Inc. | System and method for three-dimensional scanning and for capturing a bidirectional reflectance distribution function |
CN108364344A (en) * | 2018-02-08 | 2018-08-03 | 重庆邮电大学 | A kind of monocular real-time three-dimensional method for reconstructing based on loopback test |
CN108364292A (en) * | 2018-03-26 | 2018-08-03 | 吉林大学 | A kind of illumination estimation method based on several multi-view images |
US20180255283A1 (en) * | 2017-03-03 | 2018-09-06 | Sony Corporation | Information processing apparatus and information processing method |
-
2019
- 2019-04-24 CN CN201910332095.XA patent/CN110060335B/en not_active Expired - Fee Related
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130044213A1 (en) * | 2011-08-18 | 2013-02-21 | Massachusetts Institute Of Technology | System and method for diffuse imaging with time-varying illumination intensity |
US20140055570A1 (en) * | 2012-03-19 | 2014-02-27 | Fittingbox | Model and method for producing 3d photorealistic models |
US20180047208A1 (en) * | 2016-08-15 | 2018-02-15 | Aquifi, Inc. | System and method for three-dimensional scanning and for capturing a bidirectional reflectance distribution function |
CN106803267A (en) * | 2017-01-10 | 2017-06-06 | 西安电子科技大学 | Indoor scene three-dimensional rebuilding method based on Kinect |
US20180255283A1 (en) * | 2017-03-03 | 2018-09-06 | Sony Corporation | Information processing apparatus and information processing method |
CN108364344A (en) * | 2018-02-08 | 2018-08-03 | 重庆邮电大学 | A kind of monocular real-time three-dimensional method for reconstructing based on loopback test |
CN108364292A (en) * | 2018-03-26 | 2018-08-03 | 吉林大学 | A kind of illumination estimation method based on several multi-view images |
Non-Patent Citations (3)
Title |
---|
YU ZHANG ET AL: "3D Reconstruction in the Presence of Glass and Mirrors by Acoustic and Visual Fusion", 《JOURNAL OF LATEX CLASS FILES》 * |
张颂等: "基于灭点理论的目标姿态角单站测量", 《应用光学》 * |
胡正乙等: "基于RGB-D的室内场景实时三维重建算法", 《东北大学学报(自然科学版)》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110717979A (en) * | 2019-08-19 | 2020-01-21 | 北京航空航天大学 | Atmospheric and three-dimensional earth surface coupling radiation simulation method based on photon tracking |
CN110689514A (en) * | 2019-10-11 | 2020-01-14 | 深圳大学 | Training method and computer equipment for new visual angle synthetic model of transparent object |
CN110689514B (en) * | 2019-10-11 | 2022-11-11 | 深圳大学 | Training method and computer equipment for new visual angle synthetic model of transparent object |
CN111028597A (en) * | 2019-12-12 | 2020-04-17 | 塔普翊海(上海)智能科技有限公司 | Mixed reality foreign language scene, environment and teaching aid teaching system and method thereof |
WO2022142311A1 (en) * | 2020-12-30 | 2022-07-07 | 浙江商汤科技开发有限公司 | Method for segmenting target object in image, three-dimensional reconstruction method, and related apparatus |
CN113593049A (en) * | 2021-07-27 | 2021-11-02 | 吉林大学 | Virtual-real fusion method for geometric consistency of real object and virtual object in scene |
CN113593049B (en) * | 2021-07-27 | 2023-08-04 | 吉林大学 | Virtual-real fusion method for geometric consistency of real object and virtual object in scene |
Also Published As
Publication number | Publication date |
---|---|
CN110060335B (en) | 2022-06-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110060335A (en) | There are the virtual reality fusion methods of mirror article and transparent substance in a kind of scene | |
Lindemann et al. | About the influence of illumination models on image comprehension in direct volume rendering | |
CN104952063B (en) | For indicating the method and system of virtual objects in the view of true environment | |
Narasimhan et al. | Acquiring scattering properties of participating media by dilution | |
CN104346420B (en) | Method and system for digital into appearance data | |
CN101454806B (en) | Method and apparatus for volume rendering using depth weighted colorization | |
CN107644453B (en) | Rendering method and system based on physical coloring | |
CN105447906A (en) | Method for calculating lighting parameters and carrying out relighting rendering based on image and model | |
CN106535736B (en) | Image processing apparatus, image processing method and image processing program | |
Šoltészová et al. | Chromatic shadows for improved perception | |
Devlin et al. | Realistic visualisation of the pompeii frescoes | |
CN107016719B (en) | A kind of Subsurface Scattering effect real-time drawing method of screen space | |
CN107734267A (en) | Image processing method and device | |
Zhang et al. | A systematic approach to testing and predicting light-material interactions | |
Grosch et al. | Consistent interactive augmentation of live camera images with correct near-field illumination | |
Englund et al. | Evaluating the perception of semi-transparent structures in direct volume rendering techniques | |
Retzlaff et al. | Physically based computer graphics for realistic image formation to simulate optical measurement systems | |
CN112907758B (en) | Data determination method and device and electronic equipment | |
Ropinski et al. | Advanced volume illumination with unconstrained light source positioning | |
Wang et al. | Capturing and rendering geometry details for BTF-mapped surfaces | |
Gigilashvili et al. | Appearance manipulation in spatial augmented reality using image differences | |
Pereira et al. | Photorealism in mixed reality: a systematic literature review | |
JP4357997B2 (en) | Fast estimation method of bidirectional reflection distribution function of objects | |
JP6432882B2 (en) | Image simulation method | |
Qi | Measuring perceived gloss of rough surfaces |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220621 |