Glass bow detection system and detection method

Abstract

A glass bow detection system and method are disclosed. The system includes a linear array camera for acquiring edge image information of the glass to be tested and an industrial computer for storing and processing the information. In the detection method, the linear array camera is moved in a direction parallel to the edge of the glass to be tested to acquire complete edge image information of the glass to be tested. The acquired image I is preprocessed to eliminate background noise excluding the target glass in the image, and features of the glass edge in the background noise-eliminated image I_g are extracted. The grayscale values ​​of the image are scanned row by row using an algorithm. When the grayscale value suddenly increases or decreases, the position of the corresponding feature point is recorded, and the bow of the glass is obtained based on the extracted feature point. The method calculates the bow of the glass by directly detecting the curved edge contour of the glass. It has the advantages of simple operation, high accuracy, and intuitive detection variables. The bow of the glass is directly reflected in the measured variable.

Description

Technical Field

[0001] The present invention relates to the technical field of glass parameter detection, and in particular to a glass bow detection system and a detection method thereof. Background Art

[0002] Detecting the bow of the glass surface is crucial for ensuring the quality of glass production. Among the methods for detecting the bow of the glass surface, there are mainly contact measurement methods and non-contact measurement methods. Among them, contact measurement requires the detection probe to be in direct contact with the glass surface to obtain glass surface information through the probe. This method has a slow detection speed and is prone to damage to the glass surface. Non-contact measurement is mainly based on visual methods for detection. One type of method uses low-coherence light to detect the flatness of the glass. This method requires the use of special low-coherence light instruments, has high requirements for equipment layout, and is complex to operate. Another type of visual detection method projects a pattern onto the surface of the glass to be tested through a transparent plate, and then detects the bow of the glass by observing the deformation of the pattern on the glass surface. Although this method is simple to operate, it is easily affected by the lighting environment. In addition, this method measures the bow of the glass through indirect measurement, and errors are easily accumulated during the calculation process, resulting in low accuracy of the detection results. Summary of the Invention

[0003] The present invention aims to provide a glass bow detection system and a detection method thereof, which can realize non-contact, rapid and accurate detection of glass surface bow parameters.

[0004] The technical solution adopted by the present invention is: a glass bow detection system, comprising:

[0005] A linear array camera, wherein the linear array camera obtains edge image information of the glass to be measured while moving along the sliding track, and the coaxial light source moves synchronously with the linear array camera;

[0006] The industrial computer stores the edge image information collected by the line array camera in the industrial computer for processing.

[0007] As a preferred solution, the moving direction of the linear array camera on the sliding track is parallel to the edge of the glass to be measured, and the sliding track is slightly longer than the length of the glass to ensure that the edge of the glass can be completely captured.

[0008] As a preferred solution, the detection system is further provided with a bracket for supporting the glass to be tested.

[0009] Also included is a method for detecting glass bow, comprising the following steps:

[0010] Step 201: Adjust the coaxial light source and the line scan camera so that the line scan camera can clearly and completely capture the glass edge image I;

[0011] Step 202: Acquire an image I containing the edge of the glass to be measured, which is captured by a linear array camera, and store the image I in an industrial computer;

[0012] Step 203: pre-process the image I to eliminate background noise outside the target glass, thereby obtaining a noise-eliminated image I_g;

[0013] Step 204: extract features of the glass edge in the image I_g and record feature point information of the glass edge;

[0014] Step 205: Calculate the bow degree of the glass to be tested based on the extracted feature points.

[0015] As a preferred solution, in step 201, the coaxial light source moves synchronously with the linear array camera, and the linear array camera moves in a direction parallel to the edge of the glass to be tested and obtains complete edge image information of the glass to be tested.

[0016] As a preferred solution, in step 203, for the collected image I, an image I_part of the glass-free portion is captured, and then the grayscale values ​​of I_part are averaged by column to obtain a row of grayscale values. The grayscale of this row is recorded as the average grayscale value of the background noise. The average grayscale value is expanded according to the height of the image I to obtain an average grayscale image I_bg of H*W, where H and W are the length and width of the image I, respectively. The original image I is subtracted from the average grayscale image I_bg to obtain a preprocessed image I_g with the background noise removed.

[0017] As a preferred solution, in step 204, for the pre-processed image I_g, the grayscale values ​​of the image are scanned row by row, and when the grayscale value suddenly increases or decreases, the position of the corresponding feature point is recorded.

[0018] As a preferred solution, in this step, the specific method is: when scanning the image line by line, record the feature points p_a and p_b corresponding to the sudden increase and decrease of the grayscale value from left to right, record the distance between p_a and p_b as L_a, and calculate the confidence α=|(L-L_a) / L| based on this value, where L is the preset glass width value. When the confidence satisfies 0≤α≤0.5, retain the feature point.

[0019] As a preferred solution, this step includes the step of removing abnormal points, specifically: record all feature points p_a into the point set P_a, fit a straight line every 5 points in a row, and calculate the distance d_i of each feature point from the straight line in turn. When (d_i-d_mean) / d_std>10, remove the feature points p_a and p_b corresponding to the row, where d_mean is the average value of the 5 points from the straight line, and d_std is the standard deviation of the distance; perform the same operation on the other side p_b.

[0020] As a preferred solution, in step 205, the glass length L and the glass bow height H in the image I_g are obtained according to the extracted feature points, and the glass bow value is calculated according to the formula: value = H / L.

[0021] The beneficial effects of the present invention are:

[0022] Based on the defects of the existing technology, a detection system and method are provided for non-contact, rapid and accurate detection of the bow degree of the glass surface. In this method, a linear array camera is moved in a direction parallel to the edge of the glass to be tested and complete edge image information of the glass to be tested is obtained. The obtained image I is preprocessed to eliminate background noise outside the target glass in the image, and features of the glass edge in the background noise-eliminated image I_g are extracted. The grayscale value of the image is scanned row by row through an algorithm. When the grayscale value suddenly increases or decreases, the position of the corresponding feature point is recorded, and the bow degree parameters of the glass are calculated based on the extracted feature points. This method calculates the bow degree of the glass by directly detecting the curved edge contour of the glass. It has the advantages of simple operation, high accuracy, and intuitive detection variables. The bow degree of the glass is directly reflected in the measured variables. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 Schematic diagram of the structure of the detection system of the present invention;

[0025] Figure 2 Flow chart of the detection method of the present invention;

[0026] Figure 3 It is a structural diagram of processing images captured by a line array camera in the present invention.

[0027] Reference numerals: 101, line array camera, 102, coaxial light source, 103, sliding guide rail, 104, bracket, 105, industrial computer. DETAILED DESCRIPTION

[0028] The present invention is described in detail below by way of exemplary embodiments. However, it should be understood that elements, structures, and features in one embodiment may also be beneficially combined in other embodiments without further description.

[0029] It should be noted that: unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning understood by persons having ordinary skills in the field to which the invention belongs. The words "one", "an" or "the" and the like used in the patent application specification and claims of the present invention do not express a quantitative limitation, but rather indicate the presence of at least one. Words such as "include" or "comprise" indicate that the elements or objects appearing before "include" or "comprises" include the elements or objects listed after "include" or "comprises" and their equivalents, but do not exclude other elements or objects with the same function.

[0030] In order to more clearly describe the specific structure and detailed detection process of the glass bow detection system, Figure 1-3 The description is as follows:

[0031] like Figure 1 As shown, a glass bow detection system mainly consists of five devices, namely a line array camera 101, a coaxial light source 102, a sliding track 103, a bracket 104 and an industrial computer 105. Among them, the bracket 104 is used to support the glass to be tested. The sliding track 103 is arranged next to the bracket 104 to support the line array camera 101 and constrain the sliding direction of the line array camera 101. The movement direction of the line array camera 101 on the sliding track 103 is parallel to the edge of the glass to be tested. The length of the sliding track 104 is slightly longer than the length of the glass to ensure that the edge of the glass can be fully captured. The captured images are stored in the industrial computer 105 for batch processing.

[0032] Using the above detection system, such as Figure 2-3 As shown, a glass bow detection method based on computer vision edge detection technology is adopted. The specific steps are as follows:

[0033] Step 201: Adjust the coaxial light source 102 and the line scan camera 101 so that the line scan camera 101 can clearly and completely capture the glass edge image I;

[0034] Step 202: Acquire an image I containing the edge of the glass to be measured, which is captured by the line array camera 101, and store the image I in an industrial computer;

[0035] Step 203: pre-processing the image I to remove background noise outside the target glass, thereby obtaining a noise-removed image I_g;

[0036] Step 204: extracting features of the glass edge in the image I_g and recording feature point information of the glass edge;

[0037] Step 205: Calculate the bow degree of the glass to be tested based on the extracted feature points.

[0038] In this embodiment, the image preprocessing portion in step 203 cuts off the glass-free portion of the captured image I (e.g., the first 200 rows of the image), then calculates the mean (or mode, maximum, median, etc.) of the grayscale values ​​of I_part by column to obtain one row of grayscale values. The grayscale value of this row is recorded as the average grayscale value of the background noise. This average grayscale value is expanded by the height of the image I to obtain an average grayscale image I_bg of H*W, where H and W are the length and width of the image I, respectively. After obtaining the I_bg image, the original image I is subtracted from the average grayscale image I_bg to obtain the preprocessed image I_g with the background noise removed.

[0039] In addition, it should be noted that: in the glass edge preprocessing part, since the change scale of glass thickness is negligible compared to the change scale of the glass itself, the glass thickness is approximately unchanged to a certain extent. In order to improve the accuracy of glass detection, a dual-edge detection method is used here to detect the glass edge; because the coaxial light source used makes the glass edge area of ​​interest in I_g higher in brightness, the glass edge can be located by looking for areas with higher grayscale values.

[0040] In this embodiment, in step 204, the algorithm scans the grayscale values ​​of the preprocessed image I_g row by row. Optionally, the image can be scanned every n rows to improve detection speed. For each selected row, from left to right, whenever the grayscale value suddenly changes and exceeds thre_1, the positions of the feature points p_1 and p_2 are recorded. More strictly, as the algorithm scans the image row by row, the corresponding feature points p_a and p_b corresponding to sudden increases in grayscale value by thre_a1 and sudden decreases by thre_a2 are recorded, respectively. The distance between p_a and p_b is recorded as L_a, and a confidence level α = |(L-L_a) / L| is calculated based on this value, where L is the preset glass width. This operation can be performed during camera initialization. If the confidence level satisfies 0≤α≤0.5, the feature point is retained; otherwise, it is not retained.

[0041] After scanning, outliers must be removed. All feature points p_a are recorded in the point set P_a. A straight line is fitted to every five points in a row. The distance d_i from each feature point to the line is calculated. When (d_i - d_mean) / d_std > 10, the corresponding feature points p_a and p_b in that row are removed. d_mean is the average distance of the five points from the line, and d_std is the standard deviation of the distance. The same operation is repeated for the other side p_b.

[0042] In step 205, the glass bow value = H / L, where L is the length of the glass and H is the highest height of the bowed glass. The reference plane for this height is the plane where the two sides of the glass are located.

[0043] In this embodiment, this value is calculated using value_a = H / L. For example, using the point set P_a, L = dist(P_ai, P_aj), where dist() is a function that calculates the distance between two points. P_ai and P_aj represent the first and last points in the point set P_a, respectively. H = max(D_a), where max() is a function that finds the maximum value of an array. D_a is an array consisting of the distance between each point in P_a and the line l_a, where the line l_a is a line fitted based on P_a. Similarly, vaule_b is calculated, ultimately yielding the glass bow v = (value_a + value_b) / 2.

[0044] The following description uses a glass with a side length of 538mm and a thickness of 4mm as an example: In this method, an image with a size of 63983*2048 is generally used for bow measurement. Under this condition, based on the method proposed in this patent, the bow detection accuracy can reach 10um; compared with the method of projecting a pattern onto the surface of the glass to be measured through a transparent plate, this method directly calculates the bow required for measuring the glass length L and bow height H through the image, without the need to indirectly calculate the glass bow by observing the deformation of the pattern on the glass surface, and is not easily affected by ambient light; compared with low-coherence light detection instruments, the measurement process of this method is simple and intuitive, and does not require additional equipment such as fiber couplers.

[0045] Parts not described in detail in this embodiment are prior art.

[0046] It should be noted that although the present invention has been described with reference to the above embodiments, the present invention may also have other various embodiments. Without departing from the spirit and scope of the present invention, it is obvious that those skilled in the art may make various corresponding changes and modifications to the present invention, and such changes and modifications shall fall within the scope of protection of the appended claims and their equivalents.

Claims

1. A method for detecting a glass bow degree detection system, characterized in that: Here are the steps: Step 201: Adjust the coaxial light source and the line scan camera so that the line scan camera can clearly and completely capture the glass edge image I; Step 202: Acquire an image I containing the edge of the glass to be measured, which is captured by a linear array camera, and store the image I in an industrial computer; Step 203: pre-process the image I to eliminate background noise outside the target glass, thereby obtaining a noise-eliminated image I_g; Step 204: extract features of the glass edge in the image I_g and record feature point information of the glass edge; Step 205: Calculate the bow degree of the glass to be tested based on the extracted feature points; In step 203, for the captured image I, an image I_part without glass is captured. The grayscale values ​​of I_part are then averaged by column to obtain a row of grayscale values. The grayscale value of this row is recorded as the average grayscale value of the background noise. This average grayscale value is expanded by the height of the image I to obtain an average grayscale image I_bg of H*W, where H and W are the length and width of the image I, respectively. The original image I is subtracted from the average grayscale image I_bg to obtain a preprocessed image I_g with the background noise removed. In step 204, the preprocessed image I_g is scanned row by row through its grayscale values. When the grayscale value suddenly increases or decreases, the position of the corresponding feature point is recorded. The specific method in this step is as follows: when scanning the image row by row, the feature points p_a and p_b corresponding to the sudden increase and decrease of the grayscale value are recorded from left to right, respectively. The distance between p_a and p_b is recorded as L_a. Based on this value, a confidence level α = |(L-L_a) / L| is calculated, where L is the preset glass width value. When the confidence level satisfies 0≤α≤0.5, the feature point is retained. This step includes the step of removing abnormal points, specifically: record all feature points p_a into the point set P_a, fit a straight line every 5 points in a row, and calculate the distance d_i of each feature point from the straight line in turn. When (d_i-d_mean) / d_std>10, remove the feature points p_a and p_b corresponding to the row, where d_mean is the average value of the distance between the 5 points and the straight line, and d_std is the standard deviation of the distance; perform the same operation on the other side p_b.

2. The detection method according to claim 1, wherein: In step 201 , the coaxial light source moves synchronously with the linear array camera, and the linear array camera moves in a direction parallel to the edge of the glass to be tested and obtains complete edge image information of the glass to be tested.

3. The detection method according to claim 1, wherein: In step 205, the glass length L and the glass bow height H in the image I_g are obtained based on the extracted feature points, and the glass bow value is calculated according to the formula: value = H / L.

4. The detection system used in the detection method according to any one of claims 1 to 3, characterized in that: include: A linear array camera, wherein the linear array camera obtains edge image information of the glass to be measured while moving along the sliding track, and the coaxial light source moves synchronously with the linear array camera; The industrial computer stores the edge image information collected by the line array camera in the industrial computer for processing.

5. The detection system according to claim 4, characterized in that: The moving direction of the linear array camera on the sliding track is parallel to the edge of the glass to be measured, and the sliding track is slightly longer than the length of the glass to ensure that the edge of the glass can be completely captured.

6. The detection system according to claim 4, wherein: The detection system is also provided with a bracket for supporting the glass to be tested.

Images