CN101398940B - True object material quality reflecting attribute modeling approach based on linear light - Google Patents

Abstract

A modeling method of material reflection properties of real objects based on linear light sources: (1) Data collection and processing by data acquisition equipment based on linear light sources; linear light sources, as the light source system of data collection equipment; On the collected plane object, the control system drives the automatic translation system to move forward, and at the same time, the industrial camera collects the image of the collected plane, and the computer performs subsequent data processing on the collected image data to obtain the pixel value track curve and store it; (2) Generate reflection table; (3) BRDF attribute modeling, according to the pixel value trajectory curve obtained in step (1) and the reflection table generated in step (2), obtain ρ _d of the Ward illumination model, then obtain α, and finally obtain ρ _s , where ρ _d is the diffuse reflection parameter, ρ _s the specular reflection parameter, and α is the material roughness coefficient. The invention can efficiently collect data, quickly solve BRDF model parameters, and is simple to implement.

Description

Modeling method of real object material reflection property based on linear light source

technical field

The invention belongs to the technical field of computer virtual reality, and specifically relates to a method for modeling reflection properties of flat object materials using linear light sources, which is used for modeling and drawing material BRDF in a realistic virtual environment.

Background technique

3D modeling of real objects is an important part of computer graphics. 3D modeling is divided into geometric modeling and attribute modeling. Geometric modeling can be well realized by modeling software and 3D scanners, while attribute modeling has become a difficult point in computer graphics modeling. Accurate acquisition and modeling of real object surface reflection properties is a hot issue in computer graphics. The appearance of an object is the result of light and the material of the object. Modeling reflection properties is all about describing how light interacts with surfaces. At present, the most commonly used model to describe reflection properties is Bi-directional Reflectance Distribution Function BRDF (Bi-directional Reflectance Distribution Function). Document 1-Nicodemus, F.E., Richmond, J.C., Hsia, J.J., Ginsberg, I.W., and Limperis, T.Geometric considerations and nomenclature for reflection.National Bureau of Standards Monograph 160, 1977 first fully described the BRDF definition, which describes The ratio of incident light reaching the surface of an object to the outgoing light reflected from that location. If the direction of the incident light and outgoing light is fixed, the material sample is rotated around the surface normal, and the BRDF value remains unchanged, the material is an isotropic material; otherwise, it is called an anisotropic material. BRDF conforms to the basic principles of physics and has important properties such as non-negativity, energy conservation and reciprocity. In the real world, many objects are composed of multiple materials. For this reason, it is necessary to use a spatially varying BRDF (Spatially Varying BRDF, SVBRDF) to describe the surface reflection properties of the object.

Early BRDF data acquisition equipment was based on goniorreflectometer (gonioreflectometer) acquisition equipment, such as literature 2-K.Torrance, E.Sparrow, and R.Birkebak, "Polarization, directional distribution, and off-specular peak phenomenona in light reflected from roughened surfaces," Optical Society of America, vol.56, no.7, pp.916-925, 1966 and document 3-F.Sing-Choong, "A Gonioreflectometer for Measuring the Bidirectional Reflectance of Material for Use in Illumination Computation," The equipment used in Computer Science, Cornell University, 1997, this type of acquisition equipment has high accuracy, but the entire data acquisition process is relatively time-consuming. What is more commonly used now is the collection equipment based on the digital camera. The two-dimensional images collected by the camera can obtain a large amount of BRDF data. This type of equipment can collect BRDF data of a single material, but it is limited by the collected material and is not universal, such as literature 4-W.Matusik, P.Hanspeter, B. Matthew et al., "Efficient isotropic BRDF measurement," in Proceedings of the 14th Eurographics workshop on Rendering, Leuven, Belgium, pp.241-247, 2003 The equipment used can also collect multi-material SVBRDF data, document 5-K. Dana, J., "BRDF/BTF measurement device." Applied in pp.460-466, an acquisition device that can quickly acquire SVBRDF, using parabolic mirrors instead of complex mechanical controls to expand the range of the incident field and the observation field , the cost of the above acquisition equipment is too high, time-consuming and inefficient.

For the data obtained through the acquisition equipment, the next step is usually to enter the BRDF attribute modeling stage to process the data to obtain a mathematical model or an empirical model. There are two main types of BRDF modeling methods: one is the modeling method based on parametric mathematical expressions, through sparse sampling of material samples, using the acquired data to match the known analytical BRDF model parameters, in the drawing process The reflection phenomenon is simulated by analytical real-time calculation, for example, the simple cosine lobe (cosine lobe) is used in the literature 6-B.T.Phong, Illumination for computer generated pictures, Communications ofACM vol. ) Phong model for approximate specular reflection, literature 7-E.Lafortune, S.-C.Foo, etc.Non-linear approximation of reflectance functions, Computer Graphics, 1997, no.Annual Conference Series: pp.117-126 in Phong model Based on the improved, physically more realistic Lafortune model and the literature 8-Ward, G.J. Measuring and modeling anisotropic reflection. In SIGGRAPH, 1992, pp.265-272 uses Gaussian instead of cosine Ward model; the other is based on data-driven (Data-Driven) BRDF modeling method, such as literature 4, densely samples materials, uses linear and nonlinear analysis, generates new BRDF data through interpolation and extrapolation methods, and provides some intuitive parameters for users to modify interactively BRDF, this type of method is complex in modeling and needs to collect a large amount of data.

Contents of the invention

The technology of the present invention solves the problem: In order to overcome the above-mentioned shortcomings, the present invention provides a method for modeling the reflection properties of real object materials based on linear light sources. This method builds a collection device based on linear light sources, which can efficiently collect data. Quickly solve BRDF model parameters.

The present invention adopts the real object material reflection attribute modeling method based on linear light source, and its characteristic is that the steps are as follows:

(1) Data acquisition and processing based on linear light source data acquisition equipment

The data collection equipment of described linear light source comprises:

Lego robot, as the control system and automatic translation system of the data acquisition equipment;

Industrial cameras as image data capture systems;

Computer, as a data processing system for data acquisition equipment;

Linear light source, as a light source system for data acquisition equipment;

The linear light source emits linear light and irradiates the collected plane object. The control system drives the automatic translation system to move forward. At the same time, the industrial camera collects the image of the collected plane, and the computer performs subsequent data processing on the collected image data to obtain the pixel value. Trajectory curve Image _{u, v} (x) and store, wherein x represents the current image data index, the range of x is [1, n], n is the number of images, u and v represent the position of the current pixel in the image;

(2) Utilize the linear light source in the step (1) to generate a reflection table;

(3) BRDF attribute modeling;

According to the pixel value trajectory curve obtained in step (1) and the reflection table generated in step (2), get the ρ _d of the Ward illumination model, then get α, and finally get ρ _s , where ρ _d diffuse reflection parameters, ρ _s specular surface Reflection parameter, α is the material roughness coefficient.

The method that described step (1) obtains pixel value track curve is: a series of image data is defined as Image _{u, v} (x), wherein x represents current image data index, and the scope of x is [1, n], n is the number of images, u and v represent the position of the current pixel in the image.

The step of generating reflective table image in described step (2):

(a) Assuming that the diffuse reflection parameter ρ _d = 1 in the Ward model, use a linear light source to perform linear integration of the light source for the diffuse reflection part, and draw a diffuse reflection table D(x);

(b) Assuming that the specular reflection parameter ρ _s = 1 in the Ward model, for the specular reflection part, the linear light source is used to perform linear integration of the light source, and the specular reflection table S _α (x) is drawn, where the roughness coefficient of the α material has a value range of 0—0.3;

(c) For each current image data index x, find the mathematical expectation μ and variance σ;

(d) optimize the two reflection tables generated in steps (a) and (b), the optimization formula is as follows:

S _α ′(x)=0 0<α′<0.005, α′ is the optimized material roughness coefficient.

Described step (3) diffuse reflection parameter ρ _d obtains as follows:

The value of ρ _d is as follows:

${\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}\mathrm{<span\; class="notranslate">\; dard</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\frac{{\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{{\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}\mathrm{<span\; class="notranslate">\; dard</span>}}}$

ρ _standard is a known diffuse reflection parameter, L _standard is the reflection peak value of the known diffuse reflection parameter ρ _standard , and L _d is the diffuse reflection peak value of the pixel value track curve obtained after computer data processing;

The method for obtaining the material roughness coefficient α of the step (3) is: when the linear light source moves to the specular reflection direction x _m , the specular reflection peak value of this point is produced, and according to statistical knowledge, the specular reflection locus curve of this point is obtained The mathematical expectation μ and variance σ of the upper pixel value, as well as the sum S of all pixel values, are obtained according to the following three formulas:

$\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}\frac{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}$

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}\frac{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}$

$\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>$

Among them, x _m is the position of the specular reflection peak, x _d is the position of the diffuse reflection peak, Image(x) is the xth image data collected, and the pixel value in the area symmetrical to x _m is taken, that is, from x _m = 2x _m -x _d to x _d , because of the influence of the projection transformation of the camera on the area, different pixels correspond to different widths, so all values need to be scaled, and the scaling factor is as follows:

$\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}$

After obtaining these statistical sample data, from the reflection table of the linear light source drawn before, look up the table to get the σ in the reflection table corresponding to the σ calculated by the statistical sample, and because the linear light source reflection table σ and α are One-to-one correspondence, then α can be obtained, and it is unique.

The formula of the step (3) specular reflection parameter ρ _s :

${\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}<span\; class="notranslate">\; \&prime;</span>}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}\mathrm{<span\; class="notranslate">\; dard</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\frac{\mathrm{<span\; class="notranslate">\; ls</span>}}{{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; \&prime;</span>}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}\mathrm{<span\; class="notranslate">\; dard</span>}}}$

Among them: ρ' _standard is the known specular diffuse reflection parameter, L' _standard is the reflection peak value of the known specular diffuse reflection parameter ρ' _standard , and Ls is the specular reflection peak value of the pixel value trajectory curve obtained after computer data processing.

Compared with the prior art, the present invention has the advantages of constructing a set of high-efficiency collection equipment based on linear light source, which can collect material attribute data in all directions of plane real objects only by collecting hundreds of sets of data images; The light source and reflector quickly solve the parameters of the BRDF model, avoiding the time-consuming, large calculation and numerical instability problems that are solved by nonlinear numerical optimization methods.

Description of drawings

Fig. 1 is the overall structural diagram of the present invention;

Fig. 2 is a schematic structural diagram of a data acquisition device based on a linear light source of the present invention;

Fig. 3 is the light source linearization reflection table of the present invention;

Fig. 4-a, Fig. 4-b, Fig. 4-c are three parameters of the Ward model of the present invention, and Fig. 4-a is the parameter p _d of the Ward model obtained, Fig. 4-b is the Ward model obtained Parameter ρ _s , Figure 4-c is the obtained parameter α of the Ward model;

Fig. 5-a and Fig. 5-b are the experimental results of the present invention, Fig. 5-a is the real image of the object, and Fig. 5-b is the rendering effect diagram of the object.

Detailed ways

The invention mainly aims at the BRDF modeling of the material reflection property of the flat real object surface.

As shown in Figure 1, the specific implementation process of the present invention includes three main steps: data collection and processing, BRDF modeling and GPU real-time rendering.

Step 1: Data collection and processing

The data acquisition and processing steps are the basis for subsequent steps, including the following stages:

(1) Construction of data acquisition equipment

As shown in FIG. 2 , the acquisition equipment used in the present invention is a linear light source data acquisition equipment. The linear light source device is a halogen tube lamp for car decoration, 50 cm long and 2 cm in diameter, wrapped with black non-reflective tape, so that one side of the light source becomes a linear light source with a length of 50 cm and a width of about 3 mm to ensure the width Can be narrow enough to be ignored as a theoretical linear light source.

Use LEGO MindStroms Robotics NXT #8527 as the control system and automatic translation system, and use its accessories to make double wheels that carry linear light sources. The track under the double wheels is made of two U-shaped aluminum pieces about 1 meter long and 1 centimeter wide One side of the automatic translation system hangs a heavy object with a fishing line, and the other side is connected to the rotating pulley of Lego robot #8527 with a fishing line. The automatic translation system can control the horizontal movement of the linear light source. The automatic translation system fixes the linear light source at a height of about 5 cm from the plane of the object to be collected, and can pull the linear light source to move in parallel at a constant speed.

The industrial camera is composed of Daheng industrial camera DH-SV1410 and camera control program. The camera is placed in a direction at an angle of 60 degrees to the upward normal vector of the collection plane, facing the collection plane.

(2) Collection of BRDF data

First define the acquisition space, that is, the acquisition plane is regarded as a plane uniquely determined by the X-axis and the Z-axis, the linear light source is parallel to the Z-axis, and moves in a direction parallel to the X-axis, the center of the acquisition plane is the origin, and the far direction of the camera is the X-axis Positive direction, the Y axis vertically upwards is the three-dimensional space in the positive direction.

In the actual collection, put the flat object to be collected on the desktop of the collection device, and flatten the two sides with black matte metal strips. Mount the camera to a point on the plane X=0, and the angle between the vector formed by the camera point and the origin and the positive direction of the Y axis is about 60 degrees. This angle must ensure that the data collected at the camera position is enough to separate the brightness values of the pixels from the trace. Reflection Peak and Specular Peak.

Then set its parameters through the API of Lego robot #8527, so that it can move along the X-axis at a constant speed and pull the linear light source forward. The number of gear rotations must ensure sufficient collection and movement distance. The so-called sufficient acquisition movement distance is that the end position is the position where the angle vector is reflected by the edge point of the acquired plane in the positive direction of the X axis, that is, the edge point of the acquired plane object in the positive direction of the X axis is relative to the angle of view vector of the camera. The position through which the reflection vector of the point normal vector passes. Set the camera parameters, adjust its focus, and make the camera collect data every 5 seconds.

The data acquisition process is as follows: move the linear light source to the negative side of the X-axis of the collection plane, so that the projection line segment projected from the starting position of the linear light source to the Y=0 plane is on the side of the negative edge of the collected plane object on the X-axis, so that The linear light source moves to the positive direction of the X axis at a constant speed, and at the same time collects data every 5 seconds until the linear light source reaches the end position, and the data collection ends. The displacement of approximately two adjacent data linear light sources is 1 mm, and the entire acquisition process takes about half an hour, and about 300 images are collected.

(3) Industrial camera calibration and data processing

In order to establish the projection transformation relationship between the image coordinate system and the acquisition space, it is necessary to determine the internal and external reference calibration of the industrial camera. The application document of the present invention 9-Zhengyou Zhang, Gang Xu: A general expression of the fundamental matrix for both perspective and affine cameras. IJCAI1997: pP The method of .1502-1510 collects the checkerboard data used for calibrating the internal and external parameters of the camera. Put a black and white checkerboard and make its center coincide with the center of the acquisition space, collect an image, then move the angle of the checkerboard, collect multiple images from different angles, use these images to restore the internal parameters of the Daheng camera and the relative acquisition of the camera Spatial extrinsics.

For the data collected in step (2), data processing is performed by a computer, that is, the collected data is expressed as a curve in the form of a pixel value track and stored. The method is: define a series of images as Image _{u, v} (x), where x represents the current image data index, assuming that the number of images is n, then the range of x is [1, n]; u and v represent the position of the current pixel in the image.

The trajectory curve of each pixel contains two peaks: the first relatively wide and low peak is the diffuse reflection peak, the second relatively narrow and high peak is the specular reflection peak, and the two peaks obtained by the solution are modeled as BRDF The basis of recovering parameters, in the present invention, x is set to x _d when the linear light source arrives directly above the pixel point (ie, the peak point of diffuse reflection), and x is set to x _m when the linear light source reaches the direction of specular emission (ie, the peak point of specular reflection).

Step 2: BRDF attribute modeling

The process of BRDF attribute modeling is as follows:

(1) Establish Ward model

For attribute modeling of the collected data, the present invention adopts the isotropic elliptic Gaussian sharpening function BRDF model proposed by Ward. Because the Ward model is simple and accurate, it can better restore the surface properties of real objects. _The Ward model describes the equation _of the BRDF value at any point on the surface, that is, the _{ratio f r (} θ _i , φ _i ; θ _r , φ _r ) as follows:

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ;</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&pi;</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\sqrt{\mathrm{<span\; class="notranslate">\; cos</span>}{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; cos</span>}{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}}}<span\; class="notranslate">\; \&CenterDot;</span>\frac{\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; [</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; tan</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; \&delta;</span>}<span\; class="notranslate">\; /</span>{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; ]</span>}{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

In the above formula, ρ _d is the diffuse reflection lighting parameter, ρ _s is the specular reflection lighting parameter, α is the material reflection roughness coefficient, and δ is the angle between the normal vector n and the half-angle vector h between the incident light and the viewing angle direction.

(2) Light source linearization and drawing reflection table

The light source used by the acquisition device of the present invention is a linear light source, and in the Ward model, the BRDF value of a certain observation point corresponding to the light emitted from a certain sampling point in a certain direction is calculated, so it is necessary to understand the linear light source of the present invention Integral summation of point lights. Integral approximation is carried out for the Lambert reflection model of the diffuse reflection part and the specular reflection part respectively. When performing integral summation of light sources on the Ward model of the present invention, if it is assumed that the viewpoint vector (θ _r , φ _r ) is perpendicular to the linear light source, then θ _r = 0, and φ _r is a constant. Since the camera that collects data is on the X=0 plane, the assumption is true, and the final calculation formula for the color value of each pixel is as follows: Image(x)=ρ _d D(x)+ρ _s S _α (x)

The present invention stores the BRDF integral value after integrating the light source into a reflection table. Drawing reflective table as shown in Figure 3 The present invention can pass through four steps:

(a) Assuming that the diffuse reflection parameter ρ _d = 1 in the Ward model, the light source is linearly integrated for the diffuse reflection part, and the application regards the point light source as a trace, and performs an integral solution for the entire linear light source in the overall area, and the final result is stored in In the image, the diffuse reflection table D(x) is drawn, as shown in the D(x) part on the left side of Figure 3.

(b) Assuming that the specular reflection parameter ρ _s = 1 in the Ward model, the light source is linearly integrated for the specular reflection part, and the linear integration is performed for each x, and all the values obtained in the definition domain of x are stored in the image, and drawn Specular Table S. (x), where the value range of the material roughness coefficient α is (0, 0.3).

(c) Find the mathematical expectation μ and variance σ for each x.

(d) Optimize the above two reflective tables, that is, obtain empirical values through a large number of experimental results, so that they can be well continuous when α' is close to zero, as follows:

S _α ′(x)=0 0<α′<0.005, α′ is the optimized material roughness coefficient.

(3) Diffuse reflection parameter ρ _d

The present invention first obtains the diffuse reflection parameter ρ _d of the Ward model, as shown in FIG. 4-a. According to the pixel locus curve Image _{u, v} (x), the diffuse reflection peak value of each channel can be obtained. Because the present invention collects an ideal flat material, it can be seen that when the linear light source runs vertically above a certain point, that is, x=x _d , the pixel value obtained at this point is the peak value of diffuse reflection at this point, which is set as L _d . Here we need to pay attention to placing the linear light source at a high enough position so that the diffuse reflection peak and the specular reflection peak are far enough apart so that the influence of the specular reflection peak on the diffuse reflection peak can be ignored.

In order to obtain _ρd, it is also necessary to know the light energy of the pixel (u, v), the present invention uses a color calibration plate (ColorChecker) to obtain the diffuse reflection peak value L _standard whose diffuse reflection parameter is known as ρ _standard , and then obtains according to the proportional relationship The value of ρ _d is as follows:

${\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}\mathrm{<span\; class="notranslate">\; dard</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\frac{{\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{{\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}\mathrm{<span\; class="notranslate">\; dard</span>}}}$

(4) Material roughness coefficient α

After the diffuse reflection parameter ρ _d is known, it is first considered that the influence of the diffuse reflection peak on the specular reflection peak cannot be ignored because the diffuse reflection peak is wider than the specular reflection peak. Remove the known diffuse reflection peaks that have been obtained to obtain a trajectory curve with only specular reflection peaks. Therefore, the specular reflection trajectory curve minus the diffuse reflection part can be regarded as independent peak.

When the linear light source moves to the specular reflection direction x _m , the specular reflection peak value of this point is generated, and the present invention calculates the mathematical expectation μ and variance σ of the pixel value on the specular reflection locus curve of this point, and all pixels The sum S of values is obtained according to the following three formulas:

$\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}\frac{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}$

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}\frac{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}$

$\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; image</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>$

Where x _m is the position of the specular reflection peak, and the pixel value in the area symmetrical to x _m is taken, that is, from x _m =2x _m -x _d to x _d , because of the influence of the projection transformation of the industrial camera on this area, Different pixels correspond to different widths, so all values need to be scaled, and the scaling factor is as follows:

$\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}$

After obtaining these statistical sample data, from the reflection table of the linear light source drawn before, look up the table to get the σ in the reflection table corresponding to the σ calculated by the statistical sample, and because σ and α in the linear light source reflection table are one One corresponding, then α can be obtained, and unique, as shown in Figure 4-c.

(5) Specular reflection parameter ρ _s

The method of obtaining the specular reflection parameter ρ _s is similar to that of the diffuse reflection parameter ρ _d , as shown in Figure 4-b. But first of all, it is necessary to use the known α to find the corresponding specular reflection parameter ρ _s in the corresponding linear light source reflection table. In order to find ρ _s , it is necessary to calculate the statistical sample summation S, and then look up the reflection table to find out the corresponding value of the same α. S _α (x), and finally apply the same method as for solving ρ _d to obtain ρ _s , as follows:

${\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&rho;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}\mathrm{<span\; class="notranslate">\; dard</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\frac{\mathrm{<span\; class="notranslate">\; ls</span>}}{{\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}\mathrm{<span\; class="notranslate">\; dard</span>}}}$

Step 3: GPU real-time rendering

Draw according to the modeling results of the Ward model, store the restored parameters in the form of texture, load the texture data in the Fragmentshader stage of the GPU, and perform real-time calculation on the pixel color value, the experimental results of drawing are shown in Figure 5-b, and Figure 5-a is The real photos of the collected objects are taken, and the lower part of the image screenshot for drawing the experimental results realistically reproduces the lower part of the real image.

Claims (4)

1. true object material quality reflecting attribute modeling approach based on linear light sorurce is characterized in that step is as follows:

(1) based on the data acquisition equipment of linear light sorurce to data acquisition and processing (DAP);

The data acquisition equipment of described linear light sorurce comprises:

The Le Gao robot is as the control system and the automatic translation system of data acquisition equipment;

Industrial camera is as the view data capture systems;

Computing machine is as the data handling system of data acquisition equipment;

Linear light sorurce is as the light-source system of data acquisition equipment;

Linear light sorurce sends linear light irradiation on the quilt planar object of gathering, control system drives automatic translation system and moves forward, industrial camera carries out image acquisition to the plane of being gathered simultaneously, and computing machine carries out follow-up data to the view data of gathering to be handled, and obtains pixel value geometric locus Image _{U, v}(x) also storage, wherein x represents the current images data directory, and the scope of x is [1, n], and n is the image number, and u and v represent the position of current pixel point in image;

(2) utilize linear light sorurce in the step (1) as light-source system, carry out data acquisition,, generate linear light sorurce reflection table according to the data of gathering;

(3) bidirectional reflectance distribution function BRDF attribute modeling;

According to pixel value geometric locus and the middle reflection table that generates of step (2) that step (1) obtains, obtain the ρ of Ward illumination model _d, and then obtain α, obtain ρ at last _s, ρ wherein _dThe diffuse reflection parameter, ρ _sThe direct reflection parameter, α is the material coefficient of roughness;

The method that generates linear light sorurce reflection table in the described step (2) is:

(a) suppose diffuse reflection parameter ρ in the Ward model _d=1, partly utilize linear light sorurce to carry out light source linearization integration for diffuse reflection, draw out diffuse reflection table D (x);

(b) suppose direct reflection parameter ρ in the Ward model _s=1, partly utilize linear light sorurce to carry out light source linearization integration for direct reflection, draw out direct reflection table S _α(x), wherein α is the material coefficient of roughness, and its span is 0-0.3;

(c) for each current images data directory x, ask mathematical expectation μ and variances sigma;

(d) to step (a) and two reflection tables that generate (b) be optimized, it is as follows to optimize formula:

S _{α '}(x)=and 00＜α '＜0.005, α ' is the material coefficient of roughness after optimizing.

2. the true object material quality reflecting attribute modeling approach based on linear light sorurce according to claim 1 is characterized in that described step (3) diffuse reflection parameter ρ _dObtain as follows:

ρ _dValue as shown in the formula:

ρ _StandardBe known diffuse reflection parameter, L _StandardBe known diffuse reflection parameter ρ _StandardReflection peak, L _dThe diffuse reflection peak value of the pixel value geometric locus that obtains after handling for computer data.

3. the true object material quality reflecting attribute modeling approach based on linear light sorurce according to claim 1 is characterized in that the method that obtains of the material coefficient of roughness α of described step (3) is: when linear light sorurce runs to direct reflection direction x _mDuring the place, this x _mThe direct reflection peak value at place produces, and according to knowledge of statistics, obtains this x _mThe mathematical expectation μ and the variances sigma of pixel value on place's direct reflection geometric locus, and all pixel value sum S, promptly try to achieve according to following three formulas:

X wherein _mBe the position at direct reflection peak value place, x _dBe diffuse reflection peak value position, Image (x) is the x width of cloth view data of gathering, and gets with respect to x _mPixel value in the zone of symmetry is promptly from x _m=2x _m-x _dTo x _d, because camera is to the influence of this regional projective transformation, the width that different pixels is corresponding different, so all need scaledly for all values, zoom factor is a following formula:

After obtaining these statistical sample data, from the reflection table of the linear light sorurce drawn out before, the σ in the corresponding reflection table of σ that obtains calculating with statistical sample of tabling look-up is again because σ and α are one to one in the linear light sorurce reflection table, so just can obtain α, and unique.

4. the true object material quality reflecting attribute modeling approach based on linear light sorurce according to claim 1 is characterized in that described step (3) direct reflection parameter ρ _sFormula:

Wherein: ρ ' _StandardBe known minute surface diffuse reflection parameter, L ' _StandardBe known minute surface diffuse reflection parameter ρ ' _StandardReflection peak, Ls is the direct reflection peak value of the pixel value geometric locus that obtains after computer data is handled.

Images