Tempered glass stress spot generation and evaluation method, device and system

Abstract

The invention relates to the technical field of automatic detection, and discloses a tempered glass stress spot generation and evaluation method, device and system, and the method comprises the steps: collecting an image through an area-array camera array, and carrying out the transverse and longitudinal splicing through employing the external parameters of a camera and a matching error minimization model, and generating a complete stress spot image; then, segmenting a single glass instance by adopting a corner detection algorithm based on a gradient and an autocorrelation matrix; and finally, multiplying a difficulty coefficient determined according to process information such as glass size and thickness by the statistical score of the stress spot image to obtain a comprehensive evaluation result. According to the method, continuous, complete and seamless imaging of tempered glass on a production line is realized, the accuracy of splicing and segmentation is ensured through an accurate algorithm, a set of comprehensive evaluation system combined with process difficulty is established, and the detection automation level and the industrial guidance value of an evaluation result are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of automated testing technology, specifically to a method, apparatus, and system for generating and evaluating stress spots in tempered glass. Background Technology

[0002] During the heat treatment process of tempered glass, due to the inherent limitations of the processing equipment and the differences in heat conduction efficiency at different locations within the glass, unevenly distributed residual stress will form inside. This stress distribution not only affects the optical performance of the product but is also a key factor determining its mechanical strength and safety. Therefore, detecting this internal stress distribution is a core aspect of quality control in the glass deep processing industry.

[0003] Photoelasticity is an important experimental mechanical method for observing this stress distribution. Its physical basis lies in the stress birefringence effect that occurs when isotropic transparent materials like tempered glass are subjected to stress; that is, the difference in principal stresses within the material is proportional to the resulting optical path difference. Under a polarized light field, this effect forms specific interference fringe patterns, thus visualizing the internal stress.

[0004] However, traditional quantitative photoelastic analysis methods based on the precise counting of microscopic interference fringes have significant technical limitations in industrial applications. These methods typically require precise optical system setup and rigorous sample pretreatment, such as surface polishing and application of refractive index matching solutions, resulting in high equipment investment and operating costs. Furthermore, single measurements can only acquire localized stress data, making it difficult to perform macroscopic, continuous stress characterization on large-size glass. More importantly, its invasive and time-consuming sample preparation process fundamentally clashes with the high-speed, continuous operation environment of tempering production lines, making it unsuitable for online 100% inspection.

[0005] "Stress cloud," as an integrated representation of interference fringe patterns generated by photoelasticity on a macroscopic scale, provides a feasible technical approach for online inspection. Observation of stress cloud is a non-contact measurement method, requiring no sample pretreatment, and its detection speed can match the pace of modern production lines, demonstrating significant application value in real-time process monitoring and closed-loop quality control.

[0006] While stress cloud observation has the potential for online detection, efficient and accurate generation of complete stress cloud images and objective evaluation remain critical technical challenges in practical applications. A key challenge is stitching together discrete images acquired by multiple cameras at different times into a single, comprehensive image accurately reflecting the true stress distribution when dealing with continuously transported large-sized glass. Existing stitching methods lack accuracy and robustness in calculating relative displacement between images, easily leading to misalignment or deformation in the final synthesized image, thus compromising the data's accuracy. Furthermore, after obtaining the images, current technologies lack effective models capable of accurately segmenting individual glass instances from complex backgrounds and comprehensively quantifying the evaluation based on product manufacturing complexity, limiting the fairness of the evaluation results and their guiding value for production processes. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method, apparatus, and system for generating and evaluating stress spots in tempered glass, which solves the problems of low automation in stress spot detection, insufficient splicing and segmentation accuracy, and poor objectivity in evaluation in existing technologies.

[0008] To achieve the above objectives, the present invention provides a method, apparatus, and system for generating and evaluating stress spots in tempered glass through the following technical solution.

[0009] The first aspect of this invention provides a method for generating and evaluating stress spots in tempered glass, the method comprising the following steps:

[0010] First, adjust the angle of the linear polarizer on each lens of the area array camera array so that the polarization direction of each linear polarizer is orthogonal to the polarization direction of the linear polarizer on the surface light source.

[0011] In one specific implementation, the adjustment process can be achieved as follows: turn on the surface light source, receive the images from each area array camera, and sequentially rotate and adjust the installation angle of the linear polarization analyzer until the extinction degree of the corresponding area array camera image on the surface light source reaches its maximum.

[0012] Next, the area array camera array is calibrated. This process includes calibrating the distortion parameters of each area array camera in the array, as well as the pose transformation between the cameras in the array to obtain external parameters.

[0013] In one alternative technical solution, the external parameters can be obtained using a calibration method based on manually placed markers. Specifically, the surface light source is turned on, a tempered glass is statically placed on the conveyor, and the calibration is completed by directly matching the stress spot characteristics of the tempered glass acquired by each camera.

[0014] Then, when at least one piece of tempered glass moves on the conveyor, the stress pattern images of the tempered glass are continuously and synchronously acquired by the area array camera array. Based on the external parameters obtained in the foregoing steps, the single-frame area array camera images triggered synchronously are horizontally stitched to generate a single-frame horizontally stitched image.

[0015] Subsequently, the horizontally stitched image is vertically stitched. This step is implemented according to the vertical displacement between two adjacent single-frame horizontally stitched images in time sequence. In a specific embodiment, the vertical displacement is calculated by matching the stress pattern features of two consecutive frames of horizontally stitched images to obtain the actual vertical displacement, and the calculation process is as follows:

[0016] For the (k - 1)-th frame horizontally stitched image I

[0020] , , k , , , start , , , , , k ,

[0017] , k , k-1 , , end ,

[0019] ,

[0024] ,

[0018] , k ,

[0023] , ,

[0022] ,

[0021] , k-1 , and the k-th frame horizontally stitched image I k , calculate the matching error E(Δy k ) at the vertical displacement amount Δy k :

[0017]

[0018] where W is the image width; h k is the overlapping area height; y start and y end are the vertical boundaries of the overlapping area; I k (x, y) is the specific value of the pixel at the coordinate (x, y) in the k-th frame horizontally stitched image; I k-1 (x, y - Δy) is the specific value of the pixel at the coordinate (x, y - Δy) in the (k - 1)-th frame image; |I k (x, y) - I k-1 (x, y - Δy)| is the absolute pixel difference;

[0019] The actual vertical displacement is obtained by solving the following optimization problem

[0020]

[0021] where argmin is the parameter corresponding to finding the minimum value, and |Δy k | < H is the constraint condition.

[0022] In another alternative technical solution, the vertical displacement can also be obtained by receiving an external signal input of the conveyor roller rotation speed and calculating based on this signal.

[0023] Repeat the horizontal and vertical stitching processes until an image that completely contains the stress pattern information of the at least one piece of tempered glass is generated.

[0024] Next, the complete image is segmented to generate stress spot images of individual tempered glass instances. In one specific implementation, this segmentation step is accomplished using corner detection, specifically including:

[0025] Calculate the gradient I in the x and y directions of the complete image containing stress spot information of the at least one piece of tempered glass. x and I y And construct the autocorrelation matrix M:

[0026]

[0027] Where w(x,y) is the Gaussian weight function, and (x,y) are the pixel coordinates in the image. Let be the square of the gradient in the x-direction. I is the square of the gradient in the y-direction. x I y It is the product of the gradient in the x-direction and the gradient in the y-direction;

[0028] Calculate the corner response function R:

[0029] R = det(M) - k·trace(M) 2 ;

[0030] Where det(M) is the determinant of matrix M, det(M) = λ1λ2; trace(M) is the trace of matrix M, that is, the sum of the elements on the main diagonal of the matrix, trace(M) = λ1 + λ2; λ1, λ2 are the eigenvalues ​​of M, and k is an empirical constant;

[0031] The local maxima of the corner response function R are found as candidate corners, and rectangle fitting is performed on the candidate corners to complete the segmentation of the single tempered glass instance.

[0032] In another alternative technical solution, the segmentation step can also be assisted by additionally setting up a lidar or thermal imager and using it to detect the physical boundaries of the tempered glass.

[0033] Finally, stress spot evaluation is performed. This step integrates the process information of a single tempered glass sample to determine its difficulty coefficient, and calculates the score of its stress spot image. The difficulty coefficient is then multiplied by the score to obtain the final stress spot evaluation. The process information may include the dimensions, thickness, and tempering properties of the single tempered glass sample.

[0034] A second aspect of the present invention provides an apparatus for generating and evaluating stress spots in tempered glass, the apparatus being applied to the method described in any of the preceding claims, the structure of which includes:

[0035] The machine body contains a conveyor consisting of multiple sets of rollers;

[0036] Mounting bracket 1 is set on the machine body and located at the bottom of the conveyor. At least one surface light source is set on its top. The surface light source is located at the bottom of the gap between the two sets of rollers, and a linear polarizing mirror is set on the top of the surface light source.

[0037] Mounting bracket two is set on the machine body and located at the top of the conveyor. At least two area scan cameras are set at the bottom of the bracket. The area scan cameras are located at the top of the gap between the two sets of rollers, and a linear polarizing analyzer is set on the lens of the area scan camera.

[0038] A third aspect of the present invention provides a system for generating and evaluating stress spots in tempered glass, the system being applied to the method described in any of the preceding claims, comprising:

[0039] The stress cloud imaging module is used to image the tempered glass passing through the conveyor to reveal its stress spots by using an orthogonally polarized surface light source, a linear polarizing mirror, a surface array camera array, and a linear polarizing analyzer.

[0040] The image acquisition module is used to continuously and synchronously acquire images of stress spots on tempered glass captured by the area array camera array;

[0041] The image stitching module is used to stitch the acquired stress spot images of the tempered glass horizontally and vertically to generate a complete image;

[0042] The glass instance segmentation module is used to segment stress spot images of each individual tempered glass instance from the complete image;

[0043] The stress cloud evaluation module is used to calculate and obtain stress spot evaluation based on the process information of the single tempered glass instance and the score of its stress spot image.

[0044] This invention provides a method, apparatus, and system for generating and evaluating stress spots in tempered glass. It offers the following advantages:

[0045] 1. By employing an area array camera and combining horizontal and vertical image stitching technology, this invention achieves complete and seamless imaging of stress spots on the surface of large-size tempered glass during continuous conveying. This solves the technical problems of traditional single-point or local detection being unable to cover the entire area and adapt to continuous production processes, and significantly improves the level of automation and the integrity of image information.

[0046] 2. This invention employs a mathematical model based on minimizing matching error to accurately calculate longitudinal displacement and utilizes a corner detection algorithm based on gradient and autocorrelation matrix for glass instance segmentation. This approach ensures that both image stitching and segmentation are based on objective quantitative calculations, rather than relying on human experience or simple threshold judgments. This effectively improves the seamlessness of image stitching and the accuracy of instance segmentation, providing a reliable and repeatable data foundation for subsequent quantitative evaluation.

[0047] 3. This invention introduces a difficulty coefficient derived from process information such as glass size and thickness in the final evaluation stage, and multiplies it with the statistical score of stress spot images to establish a more comprehensive and fair evaluation system. The evaluation results not only reflect the physical distribution of stress spots, but also objectively reflect the manufacturing process difficulty of the corresponding products, making it more practical industrial guidance value and helping to accurately optimize and control the tempering process. Attached Figure Description

[0048] Figure 1 This is a perspective view of a tempered glass stress spot generation and evaluation device according to an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram showing the positional relationship between the polarizing mirror and the linear polarizing analyzer of the tempered glass stress spot generation and evaluation device according to an embodiment of the present invention.

[0050] Figure 3 This is a module flowchart of a tempered glass stress spot generation and evaluation system according to an embodiment of the present invention;

[0051] Figure 4 This is a flowchart of a method for generating and evaluating stress spots in tempered glass according to an embodiment of the present invention.

[0052] The components are: 1. Main body; 2. Conveyor; 3. Mounting frame one; 4. Surface light source; 5. Linear polarizer; 6. Mounting frame two; 7. Area array camera; 8. Linear polarizer analyzer. Detailed Implementation

[0053] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Please refer to the appendix. Figure 1 and attached Figure 2 , Figure 1This is a perspective view of a tempered glass stress spot generation and evaluation apparatus according to an embodiment of the present invention. Figure 2 This is a schematic diagram showing the positional relationship between the polarizing polarizer 5 and the linear polarizing analyzer 8 of a tempered glass stress spot generation and evaluation device according to an embodiment of the present invention. The tempered glass stress spot generation and evaluation device provided by the present invention may include: a body 1, a conveyor 2, a mounting frame 1 3, a surface light source 4, a linear polarizing polarizer 5, a mounting frame 2 6, a surface array camera 7, and a linear polarizing analyzer 8.

[0055] The main body 1 provides overall structural support for the device. A conveyor 2, consisting of multiple sets of rollers, is located inside the main body 1 and is used to transport tempered glass. A mounting bracket 3 is fixed to the main body 1 and located at the bottom of the conveyor 2. At least one surface light source 4 is located on top of the mounting bracket 3, precisely below the gap between the two sets of rollers. A linear polarizing mirror 5 covers the top of the linear polarizing mirror 5.

[0056] Mounting bracket 2 6 is fixed to the body 1 and located on top of the conveyor 2. An array of at least two area scan cameras 7 is set at the bottom of mounting bracket 2 6 and above the gap between the rollers, corresponding vertically to the area light source 4 below. Each area scan camera 7 has a linear polarizing analyzer 8 mounted on its lens.

[0057] In this embodiment, the polarization direction of the linear polarizing analyzer 8 is set to be orthogonal to the polarization direction of the linear polarizing lens 5. During operation, the light emitted from the surface light source 4 first passes through the linear polarizing lens 5 to form linearly polarized light, which then passes vertically upwards through the tempered glass on the conveyor 2. Due to the internal stress of the tempered glass, according to the photoelastic effect, the stress distribution in the glass causes birefringence, thereby changing the polarization state of the linearly polarized light passing through it. The light with the changed polarization state continues upwards, passes through the linear polarizing analyzer 8, and is received by the area array camera 7. Since the analyzer 8 is orthogonal to the polarizing lens 5, the change in polarization state caused by stress is converted into a change in light intensity, thus allowing the stress distribution inside the glass to be clearly captured by the area array camera 7 in the form of visible bright and dark fringes, i.e., stress spot images.

[0058] Please refer to the appendix. Figure 3 , Figure 3 This is a modular flowchart of a tempered glass stress spot generation and evaluation system according to an embodiment of the present invention. The present invention also provides a tempered glass stress spot generation and evaluation system, which can be implemented by a processing unit (e.g., an industrial computer, embedded system, or PLC), and includes multiple functional modules:

[0059] The stress cloud imaging module 10 is used to image the tempered glass passing through the conveyor 2 to reveal its stress spots. The physical entity of this module is the aforementioned orthogonal polarization optical system, which includes a surface light source 4, a linear polarizing mirror 5, an array of area cameras 7, and a linear polarizing analyzer 8.

[0060] Image acquisition module 20, which is electrically connected to the area array camera 7, is used to continuously and synchronously acquire images of stress spots on tempered glass generated by stress cloud imaging module 10, and provide the acquired digital image data to subsequent modules for processing.

[0061] The image stitching module 30 receives multiple, time-division images output by the image acquisition module 20. Its function is to first horizontally stitch the multiple images acquired at the same time to generate a single-frame wide-format image; then, based on the longitudinal displacement between adjacent frames of wide-format images in time sequence, vertically stitch these wide-format images to finally generate a stress pattern map representing the complete glass surface.

[0062] The glass instance segmentation module 40 is used to identify and segment the image region of each independent single tempered glass from the complete stress spot image generated by the image stitching module 30, that is, to generate the stress spot image of a single instance of tempered glass.

[0063] The stress cloud evaluation module 50 is used to quantitatively evaluate each segmented individual tempered glass instance. It determines a difficulty coefficient based on preset process information for the individual tempered glass instance, such as size and thickness, and analyzes the stress spot image of the instance to obtain a quality score. Finally, it calculates a comprehensive stress spot evaluation result.

[0064] Please refer to the appendix. Figure 4 , Figure 4 This is a flowchart of a method for generating and evaluating stress spots in tempered glass according to an embodiment of the present invention. The method may include the following steps:

[0065] Step S1 involves performing orthogonal calibration of the polarization system. Before image acquisition, the rotation angle of the linear polarizer 8, mounted on the lens of each camera in the array of area array cameras 7, is adjusted. This adjustment process assumes that the area light source 4 is turned on, and the processing unit receives and analyzes the images from each area array camera 7 in real time. By rotating each linear polarizer 8 sequentially, the overall brightness change of the corresponding camera image is observed until the overall image brightness is at its lowest, i.e., the extinction degree of the area light source 4 is at its maximum, at which point the angle of the linear polarizer 8 is locked. This step ensures that the polarization direction of the linear polarizer 8 is strictly orthogonal to the polarization direction of the linear polarizer 5, providing an optical basis for subsequent acquisition of high-contrast stress spot images.

[0066] Step S2 involves calibrating the camera array. This step is divided into intrinsic parameter calibration and extrinsic parameter calibration. Intrinsic parameter calibration is used to correct the lens distortion of each area array camera 7, and can be completed using conventional techniques such as the Zhang Zhengyou calibration method. Extrinsic parameter calibration is used to determine the relative spatial position and attitude relationship between each area array camera 7 in the camera array, i.e., pose transformation.

[0067] In this embodiment, the extrinsic parameter calibration can be achieved by statically placing a tempered glass with obvious stress spot characteristics on the conveyor 2, and simultaneously capturing images by each array camera 7. The processing unit calculates the precise pose transformation between the cameras by matching the stress spot feature points in the overlapping areas of each image, and stores this transformation as an extrinsic parameter.

[0068] Step S3: As the tempered glass moves on the conveyor 2, stress spot images are continuously and synchronously acquired. The processing unit sends a synchronization trigger signal to the area array camera 7, causing all cameras to expose at the same time and acquire a set of images.

[0069] Step S4: Perform horizontal stitching of single-frame images. For each group of images synchronously acquired in step S3, the processing unit calls the external parameters obtained in step S2 to perform coordinate transformation and pixel fusion on the images captured by each camera according to their precise relative positional relationship, stitching them into a seamless single-frame horizontal stitched image that covers the horizontal width of conveyor 2.

[0070] Step S5 involves vertically stitching multiple frames of images. This step is performed based on the vertical displacement between two adjacent horizontally stitched single-frame images in time sequence.

[0071] In one specific implementation, the core of the longitudinal stitching process described in step S5 lies in accurately calculating the actual longitudinal displacement between two adjacent horizontally stitched images in the time sequence. This calculation process will now be described in detail. Assume that the acquired horizontally stitched stress spot image sequence consists of N frames, with the width of any frame being W and the height being h. k Let I be the image of the k-th frame. k , where k is an integer between 1 and N. The computational goal of this process is to obtain the k-th frame image I. k Compared to the (k-1)th frame image I k-1 The resulting actual longitudinal displacement Δy k .

[0072] First, determine the search space for the longitudinal displacement. Let Y... max The actual longitudinal displacement Δy is calculated based on a maximum pixel offset preset according to the maximum operating speed of conveyor 2 and the acquisition frame rate of area scan camera 7. k The search space is limited to [-Y max ,Y maxWithin this interval. For any longitudinal displacement Δy to be tested within this interval. k It is possible to calculate the two frames of image I k and I k-1 The height h of the longitudinal overlapping area k Its calculation formula is: h k =H-|Δy k |

[0073] Next, calculate the longitudinal displacement Δy. k The matching error E(Δy) k This error is obtained by calculating the average absolute pixel difference between the two frames in the overlapping region. The specific calculation formula is as follows:

[0074]

[0075] Among them, I k (x,y) represents the pixel value at coordinates (x,y) of the k-th frame image; k-1 (x, y - Δy) represents the pixel value at the corresponding coordinates of the (k-1)th frame image after displacement compensation. The vertical boundary of the summation operation is determined by the boundary offset y. start and y end Determined, their calculation formulas are as follows: y start =max(1,1+Δy) k ),y end =min(H,H+Δy) k );|I k (x,y)-I k-1 (x,y-Δy)| represents the absolute pixel difference.

[0076] Finally, the actual longitudinal displacement is determined by solving a constrained optimization problem. This problem is solved by exhaustive search, that is, traversing the search space [-Y]. max ,Y max For all possible Δy_k values ​​within the range, calculate the corresponding matching error E(Δy_k). k ), and select E(Δy) k The Δy that reaches its minimum value k As the optimal solution, the solution process can be expressed by the following equation:

[0077]

[0078] Here, argmin is an operator that returns the value of the argument corresponding to the minimum value of the function. The result calculated from this... This refers to the displacement used for precise longitudinal splicing in step S5.

[0079] Step S6: Repeat the stitching process of S4 and S5. The processing unit continuously stitches the newly generated single-frame horizontal image according to the calculated vertical displacement. The image is precisely stitched below the generated image until it completely covers the area of ​​all target tempered glass on conveyor 2, ultimately generating a complete stress pattern map.

[0080] Step S7: Perform instance segmentation of individual glass pieces. For the complete image generated in S6, a corner detection-based segmentation method is used to separate each individual piece of tempered glass. This method utilizes the characteristic that the stress spots in the corners of a single piece of tempered glass have high gradients at the corner locations due to stress concentration.

[0081] In one specific implementation, the single-piece glass instance segmentation process described in step S7 is achieved through a clustering method that utilizes Harris corner detection combined with geometric constraints. This method leverages the stress concentration at the corners of a single piece of tempered glass, which causes the stress spots in the image to exhibit bright characteristics at the corresponding locations, thus resulting in significant gradient changes at these corner locations. The specific algorithm flow is as follows:

[0082] First, gradient calculation is performed on the complete stress patch image I generated in S6. The Sobel operator or other gradient operators can be used to calculate the gradient components I in the x and y directions of the image. x and I y The formula for its calculation is:

[0083]

[0084] in, Let I be the partial derivative of image I in the x-direction (horizontal direction), which is the rate of change of image intensity in the horizontal direction; It is the partial derivative of image I in the y-direction (vertical direction), that is, the rate of change of image intensity in the vertical direction;

[0085] Next, for each pixel in the image, a 2x2 autocorrelation matrix M is constructed within its neighborhood window. This matrix summarizes the gradient distribution characteristics within the neighborhood window, and its specific form is as follows:

[0086]

[0087] Where ∑ represents the summation of calculations for all pixels within the window centered on the current pixel; w(x,y) is a Gaussian weighting function used to assign different weights to pixels at different positions within the window; and (x,y) are the pixel coordinates in the image. Let be the square of the gradient in the x-direction. I is the square of the gradient in the y-direction. x Iy It is the product of the gradient in the x-direction and the gradient in the y-direction.

[0088] Subsequently, the corner response function value R for each pixel is calculated based on the autocorrelation matrix M to quantify the probability that the point is a corner. The calculation formula is as follows:

[0089] R = det(M) - k·trace(M) 2 ;

[0090] Where det(M) is the determinant of matrix M, and det(M) = λ1λ2. trace(M) is the trace of matrix M, and trace(M) = λ1 + λ2, where λ1 and λ2 are two eigenvalues ​​of matrix M. k is a preset empirical constant, whose value is usually between 0.04 and 0.06. After calculating the R-value distribution of the entire image, a set of candidate corner points can be obtained by finding the local maxima of the R-function.

[0091] Finally, rectangle fitting is performed on the candidate corner point set to complete instance segmentation. This process clusters the candidate corner points, and for each group consisting of four corner points... Geometric constraint verification is performed on the potential glass instance. First, the edge vector v formed by connecting the four corner points in sequence is calculated. i =p i+1 -p i Where p5 is defined as p1, and p i For set The i-th corner point in the array is represented by its coordinates.

[0092] Then, parallel and perpendicular constraints are applied to this set of edge vectors. The parallel constraint requires that two opposite edges must be approximately parallel, meaning their unit vectors have the same or opposite directions, which can be expressed as:

[0093]

[0094] Among them, v i Connect the corner point p i and p i+1 edge vectors; |v i | is the edge vector v i The modulus, that is, its length; For edge vector v i A unit vector represents only direction.

[0095] The perpendicular constraint requires that two adjacent edges must be approximately perpendicular, that is, the cosine of their included angle is close to 0, which can be expressed as: Where · represents the vector dot product operation, ∈ ⊥It is a preset low threshold for judging verticality. Only a group of corner points that simultaneously meets the above parallel and perpendicular constraints is identified as a valid single tempered glass instance, thus completing the accurate segmentation of its image region.

[0096] Step S8 involves quantitatively evaluating stress spots. For each individual tempered glass image segmented in S7, a difficulty coefficient is first determined based on its size, thickness, tempering properties, and other process information using a pre-defined quantification model or lookup table. Then, image analysis algorithms are used to statistically analyze the characteristic parameters of the stress spot image (such as stress spot density and uniformity) and calculate a quality score. Finally, the difficulty coefficient is multiplied by the quality score to obtain a comprehensive stress spot evaluation result.

[0097] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for generating and evaluating stress spots in tempered glass, characterized in that, Includes the following steps: S1. Install a linear polarizing analyzer (8) in the orthogonal direction of the optical path of the lens of the area array camera (7), and adjust the angle of the linear polarizing analyzer (8) on the lens of each area array camera (7) in the area array camera (7) so that the polarization direction of each linear polarizing analyzer (8) is orthogonal to the polarization direction of the linear polarizing mirror (5) on the surface light source (4). S2. Calibrate the distortion parameters of each area array camera (7) in the area array camera (7) array, and calibrate the pose transformation between each camera in the area array camera (7) array to obtain external parameters; S3. When at least one piece of tempered glass moves on the conveyor (2), the stress spot images of the tempered glass captured by the array of the area array camera (7) are continuously and synchronously acquired. S4. Based on the external parameters, the images of the synchronously triggered single-frame area array camera (7) are horizontally stitched together to generate a single-frame horizontally stitched image. S5. Based on the longitudinal displacement between two adjacent single-frame horizontally stitched images of the tempered glass in time sequence, the horizontally stitched images are vertically stitched together. S6. Repeat steps S4 and S5 until a complete image containing information about the stress spots of at least one piece of tempered glass is generated. S7. Segment the complete image containing the stress spot information of at least one tempered glass to generate stress spot images of each individual tempered glass instance; S8. Determine the difficulty coefficient based on the process information of the single tempered glass example, calculate the score of the stress spot image of the single tempered glass example, and multiply the difficulty coefficient by the score to obtain the stress spot evaluation.

2. The method for generating and evaluating stress spots in tempered glass according to claim 1, characterized in that, In step S1, there are two ways to install the linear polarizing analyzer in the orthogonal direction of the camera lens optical path: Installation method 1: Each area array camera (7) is independently equipped with a linear polarizing analyzer (8); Installation method 2: Install a complete linear polarizing analyzer (8) directly below the lenses of at least two area array cameras (7) in the area array camera (7).

3. The method for generating and evaluating stress spots in tempered glass according to claim 1, characterized in that, The method for adjusting the angle of the linear polarizer (8) in step S1 is as follows: turn on the surface light source (4), receive the image of each area array camera (7), and rotate and adjust the installation angle of the linear polarizer (8) in sequence until the extinction degree of the image of the corresponding area array camera (7) on the surface light source (4) reaches the maximum.

4. The method for generating and evaluating stress spots in tempered glass according to claim 1, characterized in that, There are two methods for obtaining the calibration of external parameters in step S2: Calibration Method 1: A calibration method using manually placed markers for matching. Specifically, a calibration plate printed with calibration markers is statically set on the conveyor (2). The calibration is completed by matching the marker features of the calibration plate captured by each camera. Calibration Method 2: The calibration method of directly matching with tempered glass. Specifically, the surface light source (4) is turned on, a piece of tempered glass is statically set on the conveyor (2), and the calibration is completed by directly matching the stress spot characteristics of the tempered glass collected by each camera.

5. The method for generating and evaluating stress spots in tempered glass according to claim 1, characterized in that, There are two methods to obtain the longitudinal displacement in step S5: The first method: calculates the actual longitudinal displacement by matching the stress spot features of two consecutive horizontally stitched images. The The calculation process is as follows: For the (k-1)th frame horizontally stitched image I k*1 And the horizontally stitched image I of the kth frame k Calculate the longitudinal displacement Δy k The matching error E(Δy) k ): Where W is the image width; h k y represents the height of the overlapping region. start and y end For the longitudinal boundary of the overlapping region; I k (x, y) represents the specific value of the pixel located at coordinates (x, y) in the horizontally stitched image of the k-th frame; k-1 (x, y - Δy) represents the pixel value at the corresponding coordinates of the (k-1)th frame image after displacement compensation; |I k (x,y)-I k-1 (x,y-Δy)| represents the absolute pixel difference; The actual longitudinal displacement is obtained by solving the following optimization problem. where argmin is the parameter corresponding to the minimum value, and |Δy k |<H is the constraint condition; The second method involves receiving an external signal input of the conveyor roller speed and calculating it based on that signal.

6. The method for generating and evaluating stress spots in tempered glass according to claim 1, characterized in that, The method for segmenting and generating stress spot images of each individual tempered glass instance in step S7 is as follows: segmentation is performed using Harris corner detection and its clustering method. The specific segmentation steps include: Calculate the gradient I in the x and y directions of the complete image containing stress spot information of the at least one piece of tempered glass. x and I y And construct the autocorrelation matrix M: Where w(x,y) is the Gaussian weight function, and (x,y) are the pixel coordinates in the image. Let be the square of the gradient in the x-direction. I is the square of the gradient in the y-direction. x I y It is the product of the gradient in the x-direction and the gradient in the y-direction; Calculate the corner response function R: R=det(M)-k·trace(M) 2 ; Where det(M) is the determinant of matrix M, det(M) = λ1λ2; trace(M) is the trace of matrix M, that is, the sum of the elements on the main diagonal of the matrix, trace(M) = λ1 + λ2; λ1, λ2 are the eigenvalues ​​of M, and k is an empirical constant; The local maxima of the corner response function R are found as candidate corners, and rectangle fitting is performed on the candidate corners to complete the segmentation of the single tempered glass instance.

7. The method for generating and evaluating stress spots in tempered glass according to claim 1, characterized in that, The method for segmenting and generating stress spot images of each individual tempered glass instance in step S7 is as follows: additionally setting up a lidar or thermal imager, and using the lidar or thermal imager to detect the physical boundaries of the tempered glass to assist in completing the segmentation.

8. The method for generating and evaluating stress spots in tempered glass according to claim 1, characterized in that, The process information in step S8 includes the dimensions, thickness, and tempering properties of the single tempered glass instance.

9. A device for generating and evaluating stress spots in tempered glass, characterized in that, The method for generating and evaluating stress spots in tempered glass according to any one of claims 1-8 includes: The machine body (1) is equipped with a conveyor (2) inside the machine body (1). The conveyor (2) is composed of multiple sets of rollers. A mounting frame (3) is provided on the bottom of the conveyor (2) on the machine body (1). At least one surface light source (4) is provided on the top of the mounting frame (3). The surface light source (4) is located at the bottom of the gap between the two sets of rollers. A linear polarizing mirror (5) is provided on the top of the surface light source (4). The machine body (1) is provided with a mounting frame two (6) at the top of the conveyor (2). At least two area array cameras (7) are provided at the bottom of the mounting frame two (6). The area array cameras (7) are located at the top of the gap between the two sets of rollers. A linear polarizing analyzer (8) is provided on the lens of the area array camera (7).

10. A system for generating and evaluating stress spots in tempered glass, characterized in that, The method for generating and evaluating stress spots in tempered glass according to any one of claims 1-8 includes: The stress cloud imaging module is used to image the tempered glass passing through the conveyor (2) to reveal its stress spots by means of an orthogonally polarized surface light source (4), a linear polarizing mirror (5), an array of surface cameras (7) and a linear polarizing analyzer (8). The image acquisition module is used to continuously and synchronously acquire images of stress spots on tempered glass captured by the array of the area array camera (7); The image stitching module is used to stitch the acquired stress spot images of the tempered glass horizontally and vertically to generate a complete image; The glass instance segmentation module is used to segment stress spot images of each individual tempered glass instance from the complete image; The stress cloud evaluation module is used to calculate and obtain stress spot evaluation based on the process information of the single tempered glass instance and the score of its stress spot image.

Images