Glass curtain wall disease detection method and system based on polarization imaging

Abstract

The invention belongs to the field of glass curtain wall safety detection, and particularly discloses a glass curtain wall disease detection method and system based on polarization imaging, and the method comprises the steps: collecting images of a glass curtain wall in a plurality of polarization directions in an illumination environment of a polarization light source, and superposing the images in the plurality of polarization directions to obtain a polarization image; obtaining a linear polarization degree and a polarization angle of the reflected light according to the polarization image, determining an incident angle according to the linear polarization degree of the reflected light, and determining an incident plane azimuth angle according to the polarization angle; obtaining a normal vector of each pixel point of the polarization image according to the incident angle and the incident plane azimuth angle; respectively comparing the normal vector of each pixel point with a theoretical normal vector, if the deviation between all the normal vectors and the theoretical normal vector is smaller than a preset threshold value, judging that the glass curtain wall is safe, otherwise, judging that the glass curtain wall has a disease risk. According to the invention, the self-explosion hidden danger of the glass curtain wall can be detected under the condition that a destructive experiment is not carried out.

Description

Technical Field

[0001] The present invention belongs to the field of glass curtain wall safety detection, and more specifically, relates to a glass curtain wall disease detection method and system based on polarization imaging. Background Art

[0002] Curtain wall is the outer wall of a building. It is a lightweight wall with decorative effect commonly used in modern large and high-rise buildings. It is composed of a structural frame and inlaid panels. It is a building enclosure structure that does not bear the load and function of the main structure. In the main composition of the building glass curtain wall, glass is the most important material. The performance of glass materials largely determines the comprehensive performance of the building glass curtain wall. At present, glass curtain walls occupy a dominant position in the field of building curtain walls.

[0003] There are many reasons for the spontaneous explosion of glass curtain walls, but the fundamental reason is the local stress concentration in the stress layer inside the glass. According to statistics, after the glass is installed, one case of spontaneous explosion occurs in every four tons of glass curtain walls, that is, the probability of spontaneous explosion is about 3 / 1000 to 5 / 1000.

[0004] Although the existing curtain wall detection technology has proposed a variety of theoretical methods, most of them adopt a sampling safety detection method. With the existing sampling inspection method, it is difficult to use it in the daily operation and maintenance of existing glass curtain walls, and it is difficult to comprehensively judge the safety of existing glass curtain walls. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a glass curtain wall disease detection method and system based on polarization imaging, the purpose of which is to comprehensively judge the safety of the glass curtain wall without conducting destructive experiments.

[0006] To achieve the above object, according to one aspect of the present invention, a glass curtain wall defect detection method based on polarization imaging is proposed, comprising the following steps:

[0007] Collect images of the glass curtain wall in several polarization directions under the illumination of a polarized light source, and superimpose the images in several polarization directions to obtain a polarization image;

[0008] The linear polarization degree and polarization angle of the reflected light are obtained according to the polarization image, and then the incident angle is determined according to the linear polarization degree of the reflected light, and the azimuth of the incident surface is determined according to the polarization angle;

[0009] The normal vector of each pixel point of the polarization image is obtained according to the incident angle and the azimuth of the incident surface;

[0010] The normal vector of each pixel point is compared with the theoretical normal vector. If the deviation between all normal vectors and the theoretical normal vector is less than the preset threshold, the glass curtain wall is judged to be safe; otherwise, the glass curtain wall is judged to be in danger of disease.

[0011] As a further preference, the incident angle is determined according to the linear polarization degree of the reflected light, and the calculation formula is:

[0012]

[0013] where D represents the linear polarization degree of the reflected light, and θ represents the incident angle.

[0014] As a further preference, when calculating the incident angle, the corresponding calculation formula is transformed into a non-linear equation, and the bisection method is used for solution.

[0015] As a further preference, the azimuth angle of the incident plane is determined according to the polarization angle, and the calculation formula is:

[0016]

[0017] where represents the azimuth angle of the incident plane, and ψ represents the polarization angle.

[0018] As a further preference, after obtaining the normal vectors of each pixel point of the polarization image according to the incident angle and the azimuth angle, the normal vectors are first denoised according to the geometric relationship, and then the denoised normal vectors are compared with the theoretical normal vectors.

[0019] As a further preference, denoising the normal vectors according to the geometric relationship includes:

[0020] Traverse all pixel points in the polarization image, for any pixel point:

[0021] Determine the standard normal vector of this pixel point according to the normal vectors of all adjacent pixel points of this pixel point;

[0022] If the deviation between the normal vector of this pixel point and its standard normal vector is less than the preset threshold, it is considered that this pixel point is valid and is directly retained; otherwise, it is considered that this pixel point is a noise point, and its original normal vector is replaced by the standard normal vector to achieve denoising.

[0023] As a further preference, determining the standard normal vector of this pixel point according to the normal vectors of all adjacent pixel points of the pixel point includes:

[0024] If the pixel point is a non-edge point, the average value of the normal vectors of the eight pixel points around this pixel point is taken as its standard normal vector;

[0025] If the pixel point is an edge point, first expand each side of the polarization image by one row / one column outward, and set the normal vector of the expanded pixel point to be equal to the normal vector of its nearest edge point, and the normal vector of the new corner point after expansion is equal to the normal vector of the corresponding original corner point; then take the average value of the normal vectors of the eight pixel points around this pixel point as its standard normal vector.

[0026] As a further preference, several polarization directions are specifically set as: 0°, 45°, 90° and 135°.

[0027] As a further preference, the method for determining the theoretical normal vector is:

[0028] Based on the first-order differentials of the deflection function of the curved curtain wall glass in the length and width directions of the glass curtain wall, the theoretical normal vector is determined.

[0029] According to another aspect of the present invention, there is provided a glass curtain wall disease detection system based on polarization imaging, including a processor, and the processor is used to execute the above-mentioned glass curtain wall disease detection method based on polarization imaging.

[0030] Generally speaking, compared with the prior art by the above technical solutions conceived by the present invention, the following technical advantages are mainly possessed:

[0031] 1. By collecting polarization images of the glass curtain wall in multiple directions, the present invention obtains the polarization information on the glass surface, then analyzes the normal vectors of each reflection unit according to the physical properties of polarized light, and further analyzes the actual safety situation in combination with the safety threshold of the glass curtain wall. The present invention can comprehensively detect the self-explosion hidden danger of the glass curtain wall without conducting destructive experiments.

[0032] 2. The present invention obtains the degree of linear polarization and the polarization angle from the polarization image, then determines the incident angle and the azimuth angle of the incident plane, so as to obtain the normal vector of each pixel point of the polarization image; compared with the existing detection method based on visible light images, the method of the present invention can effectively avoid the influence of poor imaging caused by glass projection, and can effectively measure the displacement of the glass curtain wall, with an accuracy of 0.1 mm.

[0033] 3. Considering that dust and other particulate matters adhere to the glass surface, resulting in the obtained vector being the normal vector of the reflection surface at the dust location, at this time, there are large deviations in the direction and magnitude between the normal vector at the dust location and the normal vectors of adjacent points, which is not conducive to the overall analysis. The present invention designs a denoising processing method to further improve the detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flowchart of the glass curtain wall disease detection method based on polarization imaging according to an embodiment of the present invention;

[0035] Figure 2 It is a layout diagram of the glass curtain wall disease detection device based on polarization imaging according to an embodiment of the present invention;

[0036] Figure 3 It is a schematic diagram of the optical model according to an embodiment of the present invention;

[0037] Figure 4 It is a graph of the linear polarization degree analysis function according to an embodiment of the present invention;

[0038] Figure 5 Among them, (a)-(d) are the images collected on the surface of the glass curtain wall at each polarization angle in the embodiments of the present invention;

[0039] Figure 6 It is the polarization information map of the surface of the glass curtain wall in the embodiments of the present invention;

[0040] Figure 7 Among them, (a) and (b) are the schematic diagrams of the unit normal vectors on the surface of the glass curtain wall in the embodiments of the present invention. Specific Embodiments

[0041] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0042] A method for detecting diseases of a glass curtain wall based on polarization imaging provided by an embodiment of the present invention, as Figure 1 shown, includes the following steps:

[0043] S1. Calibrate the black and white polarization camera on the split focal plane in advance to determine the camera parameters, including the external parameters corresponding to the camera position and attitude, the internal parameters of the internal optical imaging system of the camera, and the distortion coefficient of the lens, and initialize the global coordinate system, and set the camera coordinate system as the world coordinate system.

[0044] S2. Collect images of the glass curtain wall in several polarization directions under the illumination environment of a polarization light source. In the split focal plane polarization imaging system, the world position corresponding to each pixel position in each direction is determined, so the image information collected in each polarization direction can be superimposed to obtain a polarization image that fully reflects the polarization information on the glass surface.

[0045] Specifically, set the positions of the polarization light source and the polarization camera, as Figure 2 shown, and keep the positions unchanged during the image acquisition process; preferably, set both the polarization light source and the polarization camera perpendicular to the glass curtain wall.

[0046] In this embodiment, images of the glass curtain wall in four polarization directions (0°, 45°, 90°, 135°) are collected; after the image information is superimposed, the Stokes vector S' = [S0'S1' S2' S3'] corresponding to the reflected light formed by the incident light reflected from the surface of the glass curtain wall is obtained T = [I0 + I 90 I0 - I 90 I 45 -I 135 IQ,45 -I Q,135 ] T , where I 0° ,I 45° ,I 90° and I 135° They represent the average light intensity received by the detector under the corresponding polarization direction of the polarizer, I Q,45° and I Q,135° It is the corresponding intensity value received by the wave plate detector placed in front of the polarizer at the corresponding rotation angle.

[0047] S3. Based on the polarization image and in combination with the reflection model established according to the physical properties of polarized light, the strain of each reflection unit is analyzed, specifically, the normal vector of each pixel corresponding to the reflection unit.

[0048] Furthermore, the linear polarization degree D and polarization angle ψ of the reflected light are obtained according to the polarization image, and then the incident angle θ is determined according to the linear polarization degree D of the reflected light, and the azimuth of the incident surface is determined according to the polarization angle ψ. According to the incident angle θ and the incident surface azimuth Get the normal vector a of each pixel in the polarization image i,j .

[0049] The details are as follows:

[0050] (1) Take any tiny unit on the target curtain wall glass as a reflection unit for analysis. In the reflection unit, its surface is a plane, such as Figure 3 As shown in the schematic diagram of the optical model, the reflected light is taken as the Z-axis direction, the incident surface is taken as the YZ plane, and the direction perpendicular to the incident surface is taken as the X-axis direction to establish a local three-dimensional Cartesian coordinate system. The normal vector of a space is represented by spherical coordinates, and its gradient direction is determined by the incident angle and the azimuth of the incident surface, such as Figure 3 As shown, the light reflection model is established. According to the law of light reflection, the incident angle and the reflection angle on both sides of the normal are equal and coplanar with the normal. The Cartesian coordinate form of the normal vector is the projection on the coordinates X, Y, and Z.

[0051] Normalize the vector and set the normal vector projection in the Z direction to 1, then the normal vector n of the reflection surface is:

[0052]

[0053] in is the corresponding gradient of the target point at (x, y).

[0054] Take θ m =θ1-θ2,θ n =θ1+θ2, θ1 is the incident angle, θ2 is the refraction angle, considering the rotation angle α of the polarizer in front of the detector, the Stokes vector of the reflected light is:

[0055]

[0056] Let the degree of linear polarization (DOLP) of the reflected light be D, then we have:

[0057]

[0058] According to the law of refraction of light, we can get Substituting into the above formula, we have:

[0059]

[0060] From the above formula, it can be obtained that the degree of linear polarization is related to the refractive index n of the target object and the incident angle θ. For a definite and uniform target object, its refractive index n is determined. Given the refractive index, the degree of linear polarization is only related to the incident angle θ.

[0061] Therefore, when solving the incident angle θ, the relationship between the degree of linear polarization D and the incident angle θ is: This formula cannot obtain a theoretical solution, so a numerical solution is used for approximation. Let the function on the right side of the equation be f(θ), that is, to solve the nonlinear equation f(θ) - DOLP = 0. From the above analysis, f(θ) is continuous and monotonically increasing with respect to θ, and the bisection method is used for solution.

[0062] (2) According to the Stokes vector of the reflected light, let the polarization angle of the reflected light be ψ, then we have

[0063]

[0064] According to the optical model, the azimuth angle of the incident plane The relational expression with the polarization angle ψ is: For the azimuth angle of the incident plane of each target point, there are also two solutions.

[0065] Considering the actual engineering application, the measurement method is normal incidence with a small incident angle and unilateral light irradiation. By the known azimuth of the incident light relative to the measured target, the ambiguous solutions of the azimuth angle are eliminated, that is, only positive values exist, namely

[0066] (3) According to the obtained incident angle and azimuth angle, the normal vector of the reflection unit where the corresponding point of each pixel point is located can be obtained, and the normal vectors corresponding to each point are denoised and corrected according to the geometric relationship.

[0067] Further, considering that dust and other particulate matters adhere to the glass surface, resulting in the obtained vector being the normal vector of the reflection surface at the dust position. At this time, there are significant deviations in the direction and magnitude between the normal vector at the dust position and the normal vectors of adjacent points, which is not conducive to the overall analysis and requires denoising and correction. Therefore, a preset threshold ε is set to determine whether a vector is regarded as noise. If the difference between a certain vector and its adjacent vectors exceeds the preset threshold, then this vector is marked as a noise point. Specifically as follows:

[0068] Traverse all pixel points. For any pixel point A (the normal vector corresponding to the pixel point A in the i-th row and j-th column is a i,j ):

[0069] If A is a non-edge point, at this time i, j≠1, n, take the average value of the normal vectors of the eight points around the pixel point to obtain the standard normal vector

[0070] If A is an edge point (including the four corner points), then first expand the polarization image. Expand each of the four edges of the original m×n size image outward by one row / column, and transform it into a matrix of size (m + 2)×(n + 2). And set the normal vector of the expanded pixel point to be equal to the normal vector of its nearest edge point, and the normal vector of the new corner point after expansion to be equal to the normal vector of the corresponding original corner point. Then, similarly according to the above formula, take the average value of the normal vectors of the eight pixel points around pixel point A as its standard normal vector.

[0071] When , the point is considered valid, otherwise it is noise; for the noise point a i,j , then take its standard normal vector as its true value.

[0072] S4. According to the elastic mechanics model of the glass curtain wall, analyze the safety threshold of the glass curtain wall, and then judge the actual safety situation of the glass curtain wall according to the normal vector.

[0073] Specifically, the elastic mechanics model of the glass curtain wall is a thin plate bending model, and the safety threshold is determined by analyzing the theoretical maximum lateral displacement. According to the elastic mechanics model, for the glass curtain wall of a flat quadrilateral subjected to a uniform load, the deflection function corresponding to the bending of the curtain wall glass is:

[0074]

[0075] Among them, a and b are the length and width of the glass curtain wall respectively, q is the load, and in the glass curtain wall is its own gravity, is the bending stiffness, which is only related to the properties of the glass itself; take a corner point of the glass curtain wall as the origin, the long side direction as the x-axis, and the short side direction as the y-axis to establish a two-dimensional coordinate system; x and y are the coordinate positions of the points on the glass, and m and n are the summation terms.

[0076] The position with the maximum deflection occurs at the center of the glass, which is the maximum displacement value of the theoretical displacement curve. The normal vector of the reflection unit corresponds to the first-order differentials in the x and y axis directions of the deflection function, that is The theoretical normal vector value is

[0077] For the normal vector of the glass curtain wall surface: If max(|a i,j - a' i,j |) < ε, it is considered that the deflections of all units on the glass surface are within the allowable range, and it is considered to meet the condition of the glass bearing the load under safe conditions, that is, this piece of curtain wall glass is safe; otherwise, the glass curtain wall has the hidden danger of self-explosion.

[0078] The following are specific embodiments:

[0079] S1. Obtain the parameters of the camera and the lens through Zhang Zhengyou's two-dimensional calibration method. The internal parameter matrix I, distortion coefficient d, and external parameter matrix (R, T) of the camera are respectively:

[0080]

[0081] d = (0.0402 -0.7214 2.5855 -0.0101 -0.0088)

[0082]

[0083] T = [37.7712 -83.4237 695.6190] T

[0084] Set a polarized light source, and set the light beam emitted by the polarized light source to be a right-handed circularly polarized light, which is incident in the positive direction. At this time, the Stokes vector corresponding to the incident light is S =

[1001] T .

[0085] S2. Set the positions of the polarized light source and the polarized camera, and collect images of the glass curtain wall in four polarization directions (0°, 45°, 90°, 135°), as Figure 5 shown.

[0086] S3. Analyze the normal vectors of the reflection units corresponding to each pixel point based on the polarized images.

[0087] The refractive index n of the glass curtain wall is 1.5, and the corresponding relationship between the degree of linear polarization and the incident angle θ can be obtained, as Figure 4 shown. Among them, there is a peak in the degree of linear polarization, and the corresponding incident angle θ B = 56.364°, and for any degree of linear polarization less than 1, there are two corresponding incident angles at θ B . Set the incident angle to be less than θ B, at this time, the degree of linear polarization corresponds one-to-one with the incident angle.

[0088] Solve the degree of linear polarization and the polarization angle according to the polarization images collected in four directions, and present the results in HSV format, where the hue H is the polarization angle, the value V is the degree of linear polarization, and the saturation S is set to 1 for all, to obtain an image as Figure 6 shown. Then, inversely solve the incident angle and the azimuth angle. According to the obtained incident angle and azimuth angle, the normal vector of the reflection unit where each pixel point is located can be obtained, and then the normal vectors corresponding to each point are corrected by geometric relationships.

[0089] In this embodiment, the preset threshold ε = tan5° = 0.087, and the denoised normal vectors are shown in Table 1:

[0090] Table 1 Denoised normal vectors

[0091]

[0092] Without considering the Z direction, two-dimensionalize the normal vector in the three-dimensional coordinate system. After visualization, it is as shown in Figure 7 (a) in, Figure 7 (b) in is the magnification of the vector in the selected range of Figure 7 (a).

[0093] S4. Judge the actual safety condition of the glass curtain wall according to the normal vector:

[0094] For a 36cm × 36cm double-layer insulating framed tempered glass, the thickness of each layer of glass is 6mm, the thickness of the insulating layer is 12mm, and the designed flexural stiffness of the glass curtain wall Based on the deflection function corresponding to the bending of the curtain wall glass, the theoretical maximum displacement value w = 4.85×10 -6 m, which is much smaller than 0.1mm. It is considered that the theoretical displacement of each point is 0, that is, the theoretical value a' of the micro-element normal vector where the target surface point of the corresponding plane is located in the glass world coordinate system i,j is all (0, 0, 1).

[0095] Take the threshold ε = tan5° = 0.087. For the obtained normal vector max(|a i,j - a' i,j |) = 0.085 < ε, it is considered that the deflections of all units on the glass surface are within the allowable range, and it is considered to meet the condition of the glass bearing the load under the safe condition, that is, this piece of curtain wall glass is safe.

[0096] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A glass curtain wall defect detection method based on polarization imaging, characterized in that: The steps include: Collect images of the glass curtain wall in several polarization directions under the illumination of a polarized light source, and superimpose the images in several polarization directions to obtain a polarization image; The linear polarization degree and polarization angle of the reflected light are obtained according to the polarization image, and then the incident angle is determined according to the linear polarization degree of the reflected light, and the azimuth of the incident surface is determined according to the polarization angle; The normal vector of each pixel point of the polarization image is obtained according to the incident angle and the azimuth of the incident surface; The normal vector of each pixel point is compared with the theoretical normal vector. If the deviation between all normal vectors and the theoretical normal vector is less than the preset threshold, the glass curtain wall is judged to be safe; otherwise, the glass curtain wall is judged to be in danger of disease.

2. The glass curtain wall defect detection method based on polarization imaging according to claim 1, characterized in that: The incident angle is determined according to the linear polarization degree of the reflected light, and the calculation formula is: Where D represents the linear polarization degree of the reflected light, and θ represents the incident angle.

3. The glass curtain wall defect detection method based on polarization imaging according to claim 2, characterized in that: When calculating the incident angle, the corresponding calculation formula is converted into a nonlinear equation and solved using the dichotomy method.

4. The glass curtain wall defect detection method based on polarization imaging according to claim 1, characterized in that: Determine the azimuth of the incident surface according to the polarization angle, and the calculation formula is: in, represents the azimuth of the incident plane, and ψ represents the polarization angle.

5. The glass curtain wall defect detection method based on polarization imaging according to claim 1, characterized in that: After obtaining the normal vector of each pixel of the polarization image according to the incident angle and azimuth, the normal vector is first denoised according to the geometric relationship, and then the denoised normal vector is compared with the theoretical normal vector.

6. The glass curtain wall defect detection method based on polarization imaging according to claim 5, characterized in that: Denoise normal vectors based on geometric relationships, including: Traverse all pixels in the polarization image, for any pixel: Determine the standard normal vector of the pixel point based on the normal vectors of all the pixel points adjacent to the pixel point; If the deviation between the normal vector of the pixel point and its standard normal vector is less than the preset threshold, the pixel point is considered valid and retained directly; otherwise, the pixel point is considered to be a noise point, and its original normal vector is replaced by the standard normal vector to achieve denoising.

7. The glass curtain wall defect detection method based on polarization imaging according to claim 6, characterized in that: Determine the standard normal vector of the pixel point based on the normal vectors of all the pixels adjacent to the pixel point, including: If the pixel is not an edge point, the average of the normal vectors of the eight pixels around the pixel is taken as its standard normal vector; If the pixel is an edge point, first expand the four sides of the polarized image outward by one row / column, and set the normal vector of the expanded pixel point equal to the normal vector of its nearest edge point, and the normal vector of the new corner point after expansion is equal to the normal vector of the corresponding original corner point; then take the average of the normal vectors of the eight pixels around the pixel point as its standard normal vector.

8. The glass curtain wall defect detection method based on polarization imaging according to claim 1, characterized in that: Several polarization directions are specifically set to: 0°, 45°, 90° and 135°.

9. The glass curtain wall defect detection method based on polarization imaging according to any one of claims 1 to 8, characterized in that: The method for determining the theoretical normal vector is: The theoretical normal vector is determined based on the first-order differential of the deflection function of the curtain wall glass bending in the length and width directions of the glass curtain wall.

10. A glass curtain wall defect detection system based on polarization imaging, characterized in that: It includes a processor, and the processor is used to execute the glass curtain wall disease detection method based on polarization imaging as described in any one of claims 1 to 9.

Images