CN110567918B - A Specular Quality Analysis Method Based on 2D Structured Light - Google Patents

Abstract

A mirror surface quality analysis method based on 2D structured light comprises the following steps: 1) installing equipment; 2) constructing a mirror surface initial model; 3) analyzing the imaging process of the camera by using a pinhole model; 4) determining the corresponding relation among the pixel point P, the reflection point M and the target point T; 5) color checkerboard coding; 6) dividing and decoding the color checkerboard; 7) optimizing mirror equation description; 8) fitting a new mirror model according to the normal vector, wherein the fitting mode is a least square method; 9) calculating whether the model converges according to the new mirror model; 10) calculating a normal vector of a mirror reference point according to the obtained converged mirror model; 11) comparing whether the fitted normal vector at the reference point is the same as the designed normal vector; 12) calculating the focal length of the fitted parabolic mirror in the direction X, Y, calculating the standard deviation of the normal vector of the mirror surface from the ideal value, and representing the flatness of the mirror surface. The invention can obtain more accurate measurement results and higher measurement efficiency.

Description

A Specular Quality Analysis Method Based on 2D Structured Light

technical field

The invention belongs to the field of specular quality analysis and detection, relates to a technology of concentrator specular analysis in concentrating thermal power generation, and is a specular quality analysis method based on 2D structured light.

Background technique

At present, the rapid consumption of fossil energy and serious environmental pollution have become the focus of the world's common concerns, and the development of new energy has also become the core of research in various countries. Among them, solar energy has attracted more and more attention and research due to its sustainability and cleanliness. The main way of using solar energy is to convert light energy into electrical energy. The existing technologies mainly include photovoltaic power generation and concentrating thermal power generation. The main factors of quality are the quality of the mirror surface of the condenser and the accuracy of the mirror surface installation. The analysis of specular quality is the core of the present invention.

In order to ensure the quality of the concentrator so that it can gather sunlight to generate a heat source with uniform energy distribution, the optical parameters of the mirror surface need to be highly consistent with the design model. However, machining errors may change the focal length of the mirror, and even cause distortion and surface unevenness, so the quality of the mirror needs to be further inspected. The mirror surface has the characteristics of high reflectivity, which makes it difficult for 3D scanning technology to directly measure the shape of the mirror surface. In recent years, reflection-based measurement methods have received increasing attention, such as SOFAST. The method uses the LCD screen to display the phase-shifted fringe image as the target, and captures the mirror reflection image through a digital camera. Decoding the reflection fringe image can establish the mapping relationship between the camera pixel and the reflection target point, and the corresponding specular reflection point and normal vector can be determined according to the optical reflection principle and ray tracing method. Using the selected second-order equation to fit the normal vector data, the parameters related to the optical properties of the mirror surface can be obtained through the equation parameters. But the method is still in the developmental stage.

SUMMARY OF THE INVENTION

In order to overcome the deficiencies of the existing mirror surface quality analysis, the present invention provides a mirror surface analysis method capable of obtaining more accurate measurement results and higher measurement efficiency, and does not require complex equipment and high cost.

The technical scheme adopted by the present invention to solve its technical problems is:

A specular quality analysis method based on 2D structured light, comprising the following steps:

1) Install the equipment, the measurement system uses the color-coded checkerboard image as the target, and uses a digital camera to capture the reflection image of the mirror. The shooting distance of the camera is 5.5-6.5m. The distance between the mirror and the target is 0.9-1.1m;

2) Build a specular initial model

In order to efficiently concentrate sunlight onto the receiver, the condenser mirror is designed in the shape of a paraboloid of revolution, and the mirror is described by the following equation:

Z=A×X ² +B×Y ² +C×XY+D×X+E×Y+F

Among them, X, Y and Z are the three coordinate axes of the mirror coordinate system, and A, B, C, D, E and F are the parameters to be fitted;

3) Using the pinhole model to analyze the camera imaging process

M为镜面反射点，它的空间位置通过光线与镜面(初始采用设计模型进行描述)的交点进行计算，对应的反射目标点T通过棋盘格图像解码算法得到，TM确定了光线的入射方向

根据光线反射原理，M点处法向量通过以下方程计算：Point C is the transmission center of the camera, P is the image point, and PC determines the direction vector of the reflected light

M is the mirror reflection point, its spatial position is calculated by the intersection of the light and the mirror (initially described by the design model), the corresponding reflection target point T is obtained by the checkerboard image decoding algorithm, TM determines the incident direction of the light

According to the principle of light reflection, the normal vector at point M is calculated by the following equation:

4) Determine the corresponding relationship between the pixel point P, the reflection point M and the target point T, and the corresponding relationship between P and M is determined by geometric analysis; the structured light image (color checkerboard) encoded in the space is used as the reflection target image, and the target The target is simplified as a flat plate, the image is generated by the checkerboard coding algorithm, and the decoding image is to establish the mapping relationship between the pixel P and the target point T;

5) Color checkerboard coding

Seven color codes are used to generate the target image, and different numbers are used to replace different colors. The checkerboard image is converted into a corresponding digital matrix, and the coding of the matrix is completed, that is, the corresponding checkerboard image is obtained. The encoding process needs to first generate the corresponding horizontal sequence and state transition sequence, and then combine the two to obtain the target matrix, which is divided into: encoding the horizontal sequence, generating the state transition sequence, and generating the target digital matrix;

6) Color checkerboard segmentation and decoding

The decoding process of the checkerboard is divided into three steps: color recognition, checkerboard segmentation and codeword matching;

Convert the RGB image to the HSV color domain image. In this image format, the color information is independent of the brightness information and only included in the hue H (hue) channel value. First, determine whether the pixel is color by saturation, and then use the color distance. It is normalized to the standard color; for achromatic pixels, if its picture brightness value is large, it is judged as white, otherwise it is black, and the color recognition method is used to traverse the image, and the color of all pixels is normalized to the standard color (6 color, white and black) to complete the color identification;

Divide the original image into 7 sub-images of different colors, complete the noise filtering through the erosion and expansion algorithm, and finally merge the filtered 7 images to obtain the target image;

7) Specular equation description optimization

The fitting of the mirror equation is carried out in the mirror coordinate system, so it is necessary to convert the coordinates of the pixel plane and the color checkerboard target into the mirror coordinate system, and complete the conversion of the coordinate system through the following three steps:

7.1) Convert the pixel plane coordinate system to the camera coordinate system (X'Y'Z');

7.2) Convert the checkerboard target coordinate system (X"Y"Z") to the camera coordinate system (X'Y'Z');

7.3) Convert the camera coordinate system (X'Y'Z') to the mirror coordinate system (XYZ). Since the position of the mirror corner point is estimated according to the mirror design model, there is a certain error, which will lead to the calibration deviation of the coordinate system XYZ. In order to obtain accurate specular quality analysis results, this method uses gradient descent to minimize the slope error (RMS) of the specular equation.

8) Fit a new mirror model according to the normal vector, the fitting method is the least square method, and the model formula of the mirror is as follows:

Z=A×X ² +B×Y ² +C×XY+D×X+E×Y+F

Among them, X, Y and Z are the three coordinate axes of the mirror coordinate system, and A, B, C, D, E and F are the parameters to be fitted;

According to the partial derivative formula, the normal vector at the point (x,y,z) is (n ₁ ,n ₂ ,n ₃ )

n ₁ =2A×X+C×Y+D

n ₂ =2B×Y+C×X+E

n ₃ = -1

According to the normal vectors of each point calculated in step 3) 4) 5) 6), all parameters are fitted, and the C value is the average of the two fitting values, which realizes the fitting of the new mirror model;

9) According to the new mirror surface model, calculate whether the model converges, if not, then repeat steps 3) 4) 5) 6) 8) 9) re-fit the mirror surface until the model converges, if so, enter step 10);

10) Calculate the normal vector at the mirror reference point according to the converged mirror model obtained above, and the reference point is on the bisector of the two short sides of the mirror, from the smaller short side to the larger short side 3/4;

11) Compare whether the normal vector fitted at the reference point is the same as the designed normal vector, if it is the same, go to the next step; if it is different, rotate the mirror coordinate system around the reference point to make the two normal vectors coincide, and then repeat the steps 5) to 11);

12) According to the above fitted mirror model formula, calculate the focal length of the fitted parabolic mirror in the X and Y directions, and calculate the standard deviation of the deviation between the normal vector of the mirror surface and the ideal value, so as to characterize the flatness of the mirror surface.

The method in the present invention only needs to collect one picture, the normal vector calculation process of the mirror surface adopts a parallel method, and uses double-layer iteration to ensure the fitting accuracy of the mirror surface model, thereby greatly improving the efficiency of mirror surface detection, which can be directly used for Pipeline work.

The beneficial effects of the present invention are mainly manifested in that the measuring equipment of the mirror surface is simplified into a flat plate and a digital camera. The mirror surface can be measured by a single image, and the sampling resolution of about 3000 can be achieved, which is suitable for the fast detection scene of the mirror surface.

Description of drawings

FIG. 1 is a flowchart of a specular quality analysis method based on 2D structured light.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings.

1, a specular quality analysis method based on 2D structured light, comprising the following steps:

1) installation equipment, the measuring equipment of the present invention is a color checkerboard flat panel and a digital camera, the digital camera is placed at a position about 6m away from the mirror surface, and the target target is about 1m away from the mirror surface; the mirror surface and the target target are at a certain angle to ensure that The camera can take a checkerboard image of the target through the reflection of the mirror. Adjust camera parameters, including focal length, aperture, ISO, white balance and exposure time, etc., to ensure clear images captured by the camera. Set the camera to remote control mode to avoid the risk of manual shooting destroying the coordinate relationship between the measurement components;

2) Calibration of camera internal parameters, using the checkerboard calibration board to calibrate the internal parameters of the camera, including calibration of focal length, image center and lens distortion;

3) Determine the initial coordinate system of the camera and the target, and use a high-quality plane mirror to calculate the coordinate transformation matrix of the color checkerboard target and the camera through the reflection calibration algorithm. Based on the corner information of the mirror, the LHM algorithm is used to determine the camera to the mirror. The transformation relationship of the coordinate system. Then, the conversion relationship between the pixel plane and the camera coordinate system is obtained according to the internal parameters of the camera obtained before;

4) Preliminarily establish the positional relationship between the camera and the mirror. The invention adopts the laser range finder to measure the distance from the camera to the mirror surface;

5) Estimate the distribution of the normal vector of the mirror surface according to the mirror initial model, firstly segment the checkerboard image, obtain the codeword of the checkerboard, establish the mapping relationship between the sampling point and the reflection target point, and decode the decoding according to the coordinate relationship obtained by the calibration measurement system. The coordinates of the sampling point and the target point obtained from the image are converted into the mirror coordinate system, and the gradient descent method is used to minimize the slope error (RMS) of the mirror equation, and the measurement results of the mirror optical parameters can be obtained;

6) Fit a new specular model according to the normal vector, and the fitting method is the least square method. The model formula for the mirror surface is as follows:

Z=A×X ² +B×Y ² +C×XY+D×X+E×Y+F

Among them, X, Y and Z are the three coordinate axes of the mirror coordinate system, and A, B, C, D, E and F are the parameters to be fitted;

According to the partial derivative formula, the normal vector at the point (x,y,z) is (n ₁ ,n ₂ ,n ₃ )

n ₁ =2A×X+C×Y+D

n ₂ =2B×Y+C×X+E

n ₃ = -1

According to the normal vector of each point calculated in step 5), all parameters are fitted, and the C value is the average of the two fitted values. In this way, the fitting of the new specular model is realized;

7) According to the new mirror surface model, calculate whether the model converges, if not, then repeat step 5) 6) 7) refit the mirror surface, until the model converges, if so, enter step 8);

8) Calculate the normal vector at the mirror reference point according to the converged mirror model obtained above, and the reference point is on the bisector of the two short sides of the mirror, from the smaller short side to 3/4 of the larger short side;

9) Compare whether the normal vector fitted at the reference point is the same as the designed normal vector. If they are the same, go to the next step. If they are different, rotate the mirror coordinate system around the reference point to make the two normal vectors coincide, and then repeat the steps. 5) to 9);

10) According to the above fitted mirror model formula, calculate the focal length of the fitted parabolic mirror in the X and Y directions, and calculate the standard deviation of the deviation between the normal vector of the mirror surface and the ideal value, so as to characterize the flatness of the mirror surface.

The method in the present invention only needs to take one picture, simplifies the measuring device of the mirror surface into a flat plate and a digital camera, and can complete the measurement of the mirror surface through a single picture, and can achieve a sampling resolution of about 3000, which is suitable for mirror surface measurement. Quickly detect scenes. The three-dimensional information calculation process of the mirror surface adopts a parallel method, and uses the double-layer iterative method to ensure the fitting accuracy of the mirror surface model, thereby greatly improving the efficiency of mirror surface detection and reaching the level that can be directly used in pipeline work.

The content described in the embodiments of the present specification is only an enumeration of the realization forms of the inventive concept, and the protection scope of the present invention should not be regarded as limited to the specific forms stated in the embodiments, and the protection scope of the present invention also extends to those skilled in the art. Equivalent technical means that can be conceived by a person based on the inventive concept.

Claims (1)

1. A method for analyzing the quality of a mirror surface based on 2D structured light, which is characterized by comprising the following steps:

1) installing equipment, taking the color coded checkerboard image as a target by a measuring system, and shooting a reflection image of a mirror surface by using a digital camera, wherein the shooting distance of the camera is 5.5-6.5 m, and the distance between the mirror surface and the target is 0.9-1.1 m;

2) constructing a mirror initial model

In order to efficiently concentrate sunlight onto a receiver, the collector mirror is designed in the shape of a paraboloid of revolution, which is described by the following equation:

Z＝A×X²+B×Y²+C×XY+D×X+E×Y+F

wherein X, Y and Z are mirror coordinate system three coordinate axes, A, B, C, D, E and F are parameters needing fitting;

3) analyzing camera imaging process using pinhole model

Point C is the camera transmission center, point P is the image point, and PC determines the direction vector of the reflected light

M is a mirror reflection point, the spatial position of the mirror reflection point is calculated through the intersection point of the light and the mirror, the corresponding reflection target point T is obtained through a checkerboard image decoding algorithm, and TM determines the incident direction of the light

According to the principle of ray reflection, the normal vector at the point M is calculated by the following equation:

4) determining the corresponding relation among the pixel point P, the reflection point M and the target point T, wherein the corresponding relation between P and M is determined through geometric analysis; the method comprises the steps of adopting a spatial coding structured light image as a reflection target image, simplifying the target image into a flat plate, generating the image through a checkerboard coding algorithm, and decoding the image to establish a mapping relation between a pixel P and a target point T;

5) color checkerboard coding

Seven colors are adopted for coding to generate a target image, different colors are replaced by different numbers, the checkerboard image is converted into a corresponding number matrix, the coding of the matrix is completed, and the corresponding checkerboard image is obtained; the encoding process needs to generate a corresponding transverse sequence and a state transition sequence first, and then combine the transverse sequence and the state transition sequence to obtain a target matrix, which is divided into: coding a transverse sequence, generating a state conversion sequence and generating a target digital matrix;

6) color checkerboard segmentation and decoding

The checkerboard decoding process is divided into three steps: color identification, checkerboard segmentation and codeword matching;

converting the RGB image into an HSV color domain image, wherein in the image format, color information is independent of brightness information and only contained in hue H channel numerical values, firstly judging whether a pixel is colorful according to saturation, and then normalizing the pixel into a standard color according to color distance; aiming at an achromatic color pixel, if the picture brightness value of the achromatic color pixel is larger, the achromatic color pixel is judged to be white, otherwise, the picture is black, a color identification method is adopted to traverse the image, and the colors of all pixels are normalized to a standard color, so that the color identification can be completed;

dividing an original picture into 7 sub-pictures with different colors, completing noise filtering through corrosion and expansion algorithms, and finally combining 7 filtered pictures to obtain a target image;

7) mirror equation description optimization

The fitting of the mirror equation is performed in a mirror coordinate system, and therefore the coordinates of the pixel plane and the color checkerboard target need to be converted into the mirror coordinate system, and the conversion of the coordinate system is completed by the following three steps:

7.1) converting the pixel plane coordinate system to a camera coordinate system (X ' Y ' Z ');

7.2) converting the checkerboard target coordinate system (X 'Y' Z ') to a camera coordinate system (X' Y 'Z');

7.3) converting the camera coordinate system (X ' Y ' Z ') to the mirror coordinate system (XYZ) and minimizing the slope error of the mirror equation by using a gradient descent method;

8) and fitting a new mirror surface model according to the normal vector, wherein the fitting mode is a least square method, and the model formula of the mirror surface is as follows:

Z＝A×X²+B×Y²+C×XY+D×X+E×Y+F

wherein X, Y and Z are mirror coordinate system three coordinate axes, A, B, C, D, E and F are parameters needing fitting;

from the partial derivative formula, the normal vector at the point (x, y, z) is (n)₁,n₂,n₃)

n₁＝2A×X+C×Y+D

n₂＝2B×Y+C×X+E

n₃＝-1

Fitting all parameters according to the normal vectors of each point calculated in the step 3)4)5)6), wherein the C value is the average value of the two fitting values, so that the fitting of a new mirror model is realized;

9) calculating whether the model is converged according to the new mirror model, if not, repeating the step 3)4)5)6)7)8)9) to fit the mirror again until the model is converged, and if yes, entering the step 10);

10) calculating a normal vector of a mirror reference point according to the obtained converged mirror model, wherein the reference point is on a bisector of two short sides of the mirror and is 3/4 from the smaller short side to the larger short side;

11) comparing whether the normal vector fitted at the reference point is the same as the designed normal vector or not, and if so, entering the next step; if not, rotating around the reference point through the mirror coordinate system to enable the two normal vectors to be coincident, and then repeating the steps 5) to 11);

12) according to the fitted mirror model formula, the focal length of the fitted parabolic mirror in the direction X, Y is calculated, and the standard deviation of the normal vector of the mirror surface and the ideal value is calculated, so that the flatness of the mirror surface is represented.

Images