Rapid online detection method and equipment for reflective glass bead production

Abstract

The invention discloses a rapid online detection method and equipment for reflective glass bead production, and relates to the technical field of material detection. Comprising the steps of obtaining threshold images of reflective glass beads through irradiation of a spectrum source, determining mark points in the threshold images according to the characteristics of the reflective glass beads, collecting reflection images formed by two adjacent reflective glass beads through spectrums under time sequences, and judging the contact state of two adjacent reflective glass beads based on the positions of two mark points in the reflection images, performing retrieval matching in the multiple reflection image features according to threshold image features, and judging the quality of the reflective glass beads according to a matching result. According to the invention, the mark points during detection are determined according to the characteristics of the reflective glass beads, so that the position of the reflective glass beads can be positioned by identifying the mark points instead of identifying the whole reflective glass beads on a high-speed transmission reflective glass bead production line.

Description

Technical Field

[0001] The present invention relates to the technical field of material detection, and specifically to a rapid on-line detection method and device for the production of reflective glass beads. Background Art

[0002] In the production process of reflective glass beads, since the produced reflective glass beads are sticky, when the reflective glass beads come into contact with each other, to a certain extent, it will have an adverse impact on the overall quality of the reflective glass beads. Therefore, there is an urgent need for an effective detection scheme that can monitor and control the production situation of reflective glass beads in real time and accurately on the production line, so as to ensure that the produced reflective glass beads can meet the corresponding quality standards.

[0003] After retrieval, the Chinese invention patent with the publication number "CN117805131A" discloses a "glass ball detection system and method". In this application, the light source device is located below the glass ball transmission device, and the photographing device is located above the glass ball transmission device. Such a position distribution can ensure that the upper part of the glass ball can be clearly photographed by the photographing device. The glass ball transmission device can control the glass ball to rotate in different directions, so that the images captured by the photographing device can cover the entire glass ball, reducing the measurement error. The data processing device is connected to the glass ball transmission device and can control the glass ball to rotate in different directions. At the same time, the data processing device is also connected to the photographing device and can realize the automatic analysis of the images, enabling batch detection of glass balls.

[0004] In addition, the Chinese invention with the publication number "CN116718568A" discloses a "reflective performance detection device and method for a reflective material". This application measures the reflection coefficient under multiple states by controlling variables such as the monitoring angle, the incident angle, and the monitoring distance, greatly improving the detection accuracy and detection effect.

[0005] In the production process of reflective glass beads, due to their large output and the stickiness of the produced glass beads, when they come into contact with each other, it will have an adverse impact on the overall quality of the glass beads. However, in the actual detection process, the above-mentioned disclosed device methods and similar patents usually detect the whole product and do not independently detect the mutual positional relationship between products. Therefore, when facing a large number of products, such detection methods are not only inefficient, but also long-term detection may lead to large errors in the detection results and increase the actual workload of the detection equipment. In view of this, the present invention proposes a rapid on-line detection method for the production of reflective glass beads and its related equipment to solve the above problems. Summary of the Invention

[0006] The purpose of the present invention is to provide a rapid on-line detection method and device for the production of reflective glass beads to solve the problems raised in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] In a first aspect, a rapid on-line detection method for the production of reflective glass beads is proposed, including:

[0009] Using a spectral source to irradiate to obtain a threshold image of the reflective glass beads;

[0010] Based on the characteristics of the reflective glass beads, mark points are determined in the threshold image;

[0011] Collect the reflection images formed by the spectra of two adjacent reflective glass beads in a time series, and the reflection images in each time series are different;

[0012] Based on the positions of two mark points in the reflection image, determine the contact state of two adjacent reflective glass beads;

[0013] Among the characteristics of multiple reflection images, retrieve and match according to the characteristics of the threshold image, determine the quality of the reflective glass beads according to the matching result, and obtain the correlation between the contact state between two adjacent reflective glass beads and the quality of the reflective glass beads;

[0014] Associate the matching result with the corresponding time series to obtain the abnormal time points of the reflective glass bead production line.

[0015] As a further preference of this technical solution, the method for obtaining the threshold image includes:

[0016] Collect a primary spectral image, which is formed by irradiating a qualified reflective glass bead with the maximum imaging angle of the spectral source;

[0017] Collect a secondary spectral image, which is formed by irradiating a qualified reflective glass bead with the minimum imaging angle of the spectral source;

[0018] Gray-scale process the primary spectral image and the secondary spectral image, and obtain a primary gray-scale histogram and a secondary gray-scale histogram respectively;

[0019] Stack the primary gray-scale histogram and the secondary gray-scale histogram, and obtain a threshold gray-scale histogram;

[0020] The threshold gray-scale histogram determines the threshold image based on binary processing.

[0021] As a further preference of this technical solution, the method for determining the mark points includes:

[0022] According to the vision machine, identify the attribute characteristics of the reflective glass beads;

[0023] Collect the spectral images generated by irradiating the reflective glass beads with the spectral source at different angles;

[0024] Classify according to attribute characteristics and locate corresponding matching points in the spectral image;

[0025] Select two spectral images with the largest difference in the irradiation angle of the spectral source, and retrieve the positional relationship between the corresponding matching points in the two spectral images;

[0026] Based on the evaluation method, determine the identification points among multiple matching points.

[0027] As a further preference of this technical solution, the evaluation method includes:

[0028] Construct a two-dimensional coordinate system to represent the spectral image specifications, and the horizontal and vertical coordinates of the two-dimensional coordinate system respectively correspond to the size specifications of the spectral image;

[0029] Map the spectral image to the constructed two-dimensional coordinate system;

[0030] Extract the coordinate positions of each matching point in the spectral image in the two-dimensional coordinate system;

[0031] Select two spectral images with the largest difference in the irradiation angle of the spectral source, and associate each corresponding matching point in the two spectral images;

[0032] Analyze the relationship between the two associated corresponding matching points based on the spacing algorithm;

[0033] Identify the identification points among many matching points according to the analysis results.

[0034] As a further preference of this technical solution, the spacing algorithm includes:

[0035] , where represents the spacing, respectively represent two points in the n-dimensional space, and respectively represent the coordinate values of point and point on the i-th dimension, represents the number of dimensions of the space, i represents the dimension index, ranging from 1 to n, represents that the parameter of the Minkowski distance is a positive real number parameter, and the value range is any integer from 1 to 2. When = 1, is the Manhattan distance. When p = 2, degenerates to the Euclidean distance.

[0036] As a further preference of this technical solution, the spacing algorithm includes: , where represents the spacing, where respectively represent two points in an n-dimensional space, represents the covariance matrix of the data.

[0037] As a further preference of this technical solution, the method for determining the contact state includes:

[0038] Set the specification limits of the threshold image according to the physical properties of the retroreflective glass beads;

[0039] Obtain a comparison image formed by splicing two threshold images according to the positioning of the marked points within the specification limits;

[0040] Extract the distance between two marked points from the comparison image as the discrimination distance range;

[0041] Use the comparison between the discrimination distance range and the distance between the positions of two marked points in the reflection image to judge the contact condition of adjacent retroreflective glass beads;

[0042] If the distance between the positions of two marked points in the reflection image is within the discrimination distance range, it is determined that the adjacent retroreflective glass beads are in a contact state;

[0043] If the distance between the positions of two marked points in the reflection image exceeds the discrimination distance range, it is determined that the adjacent retroreflective glass beads are in a non-contact state.

[0044] As a further preference of this technical solution, the method for discriminating the quality of retroreflective glass beads according to the matching result includes:

[0045] Construct an image standard feature library based on the features of the threshold image;

[0046] Real-time collect the features of the reflection image and match them with the image standard feature library, and use the matching algorithm to obtain the matching result;

[0047] Formulate a quality grading rule according to the production indexes of the retroreflective glass beads;

[0048] According to the corresponding relationship between the matching result and the quality grading rule, conduct a quality grading determination on the retroreflective glass beads;

[0049] The matching algorithm is ; respectively represent the pixel average values of image x and image y, characterizing the brightness, respectively represent the pixel variances of image x and image y, characterizing the contrast, represents the covariance of image x and image y, characterizing the structural similarity, is a stability constant, used to avoid the denominator being 0, where , , where L is the pixel dynamic range, with a value of 255, .

[0050] In a second aspect, to improve the rapid on-line detection method for the production of reflective glass beads disclosed above, the present invention also provides a rapid on-line detection device for the production of reflective glass beads. A rapid on-line detection device for the production of reflective glass beads uses the rapid on-line detection method for the production of reflective glass beads disclosed above and includes:

[0051] An LED spectral lamp, which is responsible for emitting a stable spectral source to irradiate the reflective glass beads;

[0052] A vision machine, which is used to obtain a threshold image and a reflection image, and is responsible for collecting the reflection images formed by irradiating two adjacent reflective glass beads with spectra at different time series, and capturing the differences in the reflection images at each time series;

[0053] A positioning marking module, which accurately identifies the marking points in the obtained threshold image through the characteristics of the reflective glass beads themselves;

[0054] A contact state discriminator, which analyzes the positional relationship between two marking points in the reflection image to determine the contact state between two adjacent reflective glass beads;

[0055] A quality discrimination and correlation analyzer, which retrieves and matches according to the threshold image characteristics, discriminates the quality of the reflective glass beads according to the matching results, and obtains the internal correlation between the contact state between two adjacent reflective glass beads and the quality of the reflective glass beads;

[0056] A time series correlator, which is used to associate the matching results with the corresponding time series and obtain the time points when abnormal conditions occur in the production process of the reflective glass bead production line.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] The rapid on-line detection method and device for the production of reflective glass beads determine the marking points during detection according to the characteristics of the reflective glass beads, so that on the high-speed transmission reflective glass bead production line, the position of the reflective glass beads can be located by identifying the marking points instead of identifying the whole reflective glass beads;

[0059] In addition, by obtaining the distance between two adjacent marking points and the reflection image characteristics of two adjacent marking points in the time series, the contact state between two adjacent reflective glass beads is judged, thus avoiding misjudgment caused by mutual occlusion of reflective glass beads due to adhesion;

[0060] Meanwhile, by constructing an image standard feature library based on threshold image features and collecting reflection image features in real time for matching with the image standard feature library, the quality of reflective glass beads can be quickly judged, which to a certain extent ensures the detection efficiency and accuracy.

[0061] In addition, by obtaining the correlation between the contact state between adjacent two reflective glass beads and the quality of reflective glass beads, it provides strong support for the quality control of the reflective glass bead production line. Finally, through the time series correlator, the matching results are associated with the corresponding time series, and the time points of abnormal conditions during the production process of the reflective glass bead production line are obtained, providing an important reference basis for troubleshooting and maintenance of the production line. Brief Description of the Drawings

[0062] Figure 1 is the step flow chart of the disclosed method of the present invention;

[0063] Figure 2 is the assembly schematic diagram of the disclosed device of the present invention;

[0064] Figure 3 is the auxiliary explanatory diagram of step S205.A of the present invention;

[0065] Figure 4 is the auxiliary explanatory diagram of step S401 of the present invention. Detailed Embodiments

[0066] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0067] Before understanding the technical method proposed by the present invention, it should be clear that in the actual production process of reflective glass beads, since the material of reflective glass beads is glass, adjacent two reflective glass beads will have a slight adhesion phenomenon due to mutual contact after forming and cooling. This adhesion not only affects the appearance quality of reflective glass beads, but also may reduce their reflective performance and block the detection of the appearance of reflective glass beads by the detection device. Therefore, the present invention proposes a rapid on-line detection method and device for the production of reflective glass beads.

[0068] As a preferred implementation method, as Figure 1 shown, the present invention provides a technical solution: a rapid on-line detection method for the production of reflective glass beads, including: step S100 - step S600.

[0069] Step S100: Obtain the threshold image of the reflective glass beads by irradiating with a spectral source.

[0070] It should be clear that, referring to Figure 2 it can be known that there is an angle between the spectral source and the reflective glass beads.

[0071] Specifically, the method for obtaining the threshold image includes: Step S101 - Step S105.

[0072] Step S101: Collect the primary spectral image.

[0073] It should be clear that, referring to Figure 2 it can be known that the primary spectral image is formed under the irradiation of the maximum imaging angle of the spectral source on the qualified reflective glass beads. In addition, it should be noted that the qualified reflective glass beads represent the standard samples of the reflective glass beads in the present invention, and they meet the production requirements in terms of physical properties, surface quality, and dimensional specifications.

[0074] Step S102: Collect the secondary spectral image.

[0075] It should be clear that, referring to Figure 2 it can be known that the secondary spectral image is formed under the irradiation of the minimum imaging angle of the spectral source on the qualified reflective glass beads.

[0076] Step S103: Grayscale the primary spectral image and the secondary spectral image, and obtain the primary grayscale histogram and the secondary grayscale histogram respectively.

[0077] It should be clear that the grayscale processing mentioned in Step S103 is to convert the color value of each pixel point in the primary spectral image and the secondary spectral image into a grayscale value through a computer. The grayscale value range is usually between 0 - 255, where 0 represents black, 255 represents white, and the intermediate values represent different degrees of gray. The primary grayscale histogram and the secondary grayscale histogram after computer grayscale processing can reflect the grayscale distribution of the qualified reflective glass beads at different imaging angles.

[0078] Step S104: Stack the primary grayscale histogram and the secondary grayscale histogram, and obtain the threshold grayscale histogram.

[0079] It should be clear that the stacking method of the primary grayscale histogram and the secondary grayscale histogram in Step S104 is to vertically stack the primary grayscale histogram and the secondary grayscale histogram on the grayscale value axis, so that the two are presented in the same coordinate system, thereby forming a threshold grayscale histogram. It should be added that the threshold grayscale histogram comprehensively reflects the grayscale distribution characteristics of the qualified reflective glass beads at the maximum and minimum imaging angles, providing a key data basis for subsequent analysis and determination.

[0080] Step S105: Determine the threshold image based on the threshold gray histogram through binarization processing.

[0081] It should be clear that the binarization processing in Step S105 is to convert each gray value in the threshold gray histogram into two optional values by a computer. Commonly, they are 0 or 255, corresponding to black and white respectively. The image obtained after such binarization processing is the threshold image, which can clearly present the gray distribution boundaries of qualified reflective glass beads at different imaging angles and provides a key visual reference for the subsequent determination of the quality of reflective glass beads.

[0082] Step S200: Determine the marking points in the threshold image according to the characteristics of the reflective glass beads.

[0083] It should be clear that the characteristics of the reflective glass beads in Step S200 specifically refer to the patterns inside the reflective glass beads or the built-in fillers used as decorations.

[0084] In addition, it should be supplemented that in the present invention, the method for determining the marking points in Step S200 includes: Step S201 - Step S206.

[0085] Step S201: Identify the attribute characteristics of the reflective glass beads according to the vision machine.

[0086] Specifically, the attribute characteristics identified by the vision machine for the reflective glass beads are the fillers used as decorations.

[0087] Step S202: Collect the spectral images generated by irradiating the reflective glass beads with a spectral source at different angles.

[0088] It should be clear that in the present invention, in Step S202, by collecting the spectral images generated by irradiating the reflective glass beads with a spectral source at different angles and classifying them according to the attribute characteristics of the reflective glass beads, the corresponding matching points are accurately located in the spectral images, which plays a paving role for the subsequent accurate determination of various states and quality conditions of the reflective glass beads, can present the relevant characteristic performances of the reflective glass beads under different lighting conditions, and provides the necessary data support for further in-depth analysis.

[0089] Step S203: Locate the corresponding matching points in the spectral image according to the classification of the attribute characteristics.

[0090] Step S204: Select two spectral images with the largest difference in the irradiation angle of the spectral source, and retrieve the positional relationship between the corresponding matching points in the two spectral images.

[0091] Step S205: Determine the identification points among multiple matching points based on the evaluation method.

[0092] It should be supplemented for step S205 that the evaluation method includes: step S205.A - step S205.F.

[0093] Step S205.A: Construct a two-dimensional coordinate system to characterize the spectral image specifications.

[0094] It should be noted that referring to Figure 3 it can be known that the abscissa and ordinate of the two-dimensional coordinate system disclosed in step S205.A respectively correspond to the size specifications of the spectral image.

[0095] Step S205.B: Map the spectral image into the constructed two-dimensional coordinate system.

[0096] Step S205.C: Extract the coordinate positions of each matching point in the spectral image in the two-dimensional coordinate system.

[0097] Step S205.D: Select two spectral images with the largest difference in the illumination angle of the spectral source, and associate each corresponding matching point in the two spectral images.

[0098] Step S205.E: Analyze the relationship between the two corresponding matching points associated based on the spacing algorithm.

[0099] It should be clear that the relationship between the corresponding matching points in step S205.E can be divided into three types after the spacing algorithm: a close relationship, a distant relationship, and a general relationship. Among them, the close relationship indicates that the two matching points are very close in spatial position, that is, the value is between 2 units, and the spectral feature similarity is relatively high. The distant relationship is the opposite, indicating that the two matching points are relatively far away in spatial position, that is, the value exceeds 3 units, and the spectral feature similarity is relatively low. The general relationship is between the close relationship and the distant relationship, indicating that the two matching points have a certain similarity in spatial position and spectral features, but are not particularly close or far away.

[0100] Step S205.F: Identify the identification points among the numerous matching points according to the analysis results.

[0101] It should be clear that to enhance the accuracy of detection, step S205.F generally selects two corresponding matching points in a close relationship as the identification points for output.

[0102] As a preferred implementation manner, the spacing algorithm disclosed in step S205.E includes: , where represents the spacing, respectively represent two points in the n-dimensional space, and respectively represent the coordinate value of point and the coordinate value of point represents the number of dimensions of space, and i represents the index of the dimension, ranging from 1 to n. It is indicated that the parameter of the Minkowski distance is a positive real number parameter, and the value range is any integer in 1 - 2. When p = 1, it is the Manhattan distance. When p = 2, it degenerates to the Euclidean distance.

[0103] It should be noted that referring to Figure 3 it can be known that since it is a two - dimensional coordinate system, the value of n is 2. At this time, the corresponding coordinate values and are (1, 2) and (1, 3), and when p takes the value of 1, at this time d = |1 - 1|+|2 - 3| = 0 + 1 = 1. When p takes the value of 2, d = . At this time, it should be added that since the value of the spacing is 1, the relationship between two corresponding matching points is judged as a close relationship.

[0104] As a preferred implementation manner, in actual use, the spacing algorithm can also be , where d represents the spacing, where \(x_i\) and \(y_i\) respectively represent two points in the n - dimensional space. Σ represents the covariance matrix of the data.

[0105] It should be noted that referring to Figure 3 it can be known that the coordinate values of \(x_i\) are (1, 2) and (1, 3). At this time, the value of d is , so the relationship between two corresponding matching points is judged as a close relationship.

[0106] It should be further added that in actual application, the first spacing algorithm focuses on calculating the spacing based on the coordinate differences between two points in the n - dimensional space to clarify the relative position relationship between the two points. The first spacing algorithm is intuitive and simple and can play a good role in preliminary judgment. While the second spacing algorithm introduces the covariance matrix of the data. The second spacing algorithm more carefully and accurately grasps the relative position and its distribution characteristics between the data by considering factors such as the correlation between the data.

[0107] Step S300: Collect the reflection images formed by the spectra of two adjacent reflective glass beads under the time series.

[0108] It should be clear that in step S300, the reflected images under each time series are different.

[0109] Step S400: Determine the contact state of two adjacent reflective glass beads based on the positions of two marked points in the reflected image.

[0110] It should be clear that in step S400 of the present invention, the method for determining the contact state includes steps S401 - S406.

[0111] Step S401: Set the specification limits of the threshold image according to the physical properties of the reflective glass beads.

[0112] It should be clear that referring to Figure 4 it can be known that the physical properties in step S401 refer to the reflection range of the reflective glass beads plus the self - specifications of the reflective glass beads, and the specification limits of the threshold image are obtained by superimposing the reflection range and the self - specifications.

[0113] Step S402: Obtain a comparison image formed by splicing two threshold images according to the positioning of the marked points within the specification limits.

[0114] Step S403: Extract the distance between the two marked points from the comparison image as the discrimination distance range.

[0115] Step S404: Use the comparison between the discrimination distance range and the distance between the positions of the two marked points in the reflected image to judge the contact condition of adjacent reflective glass beads.

[0116] Step S405: If the distance between the positions of the two marked points in the reflected image is within the discrimination distance range, it is determined that the adjacent reflective glass beads are in a contact state.

[0117] Step S406: If the distance between the positions of the two marked points in the reflected image exceeds the discrimination distance range, it is determined that the adjacent reflective glass beads are in a non - contact state.

[0118] Step S500: In multiple reflected image features, perform retrieval and matching according to the threshold image features.

[0119] It should be clear that in step S500, the matching result can distinguish the quality of the reflective glass beads and obtain the correlation between the contact state between two adjacent reflective glass beads and the quality of the reflective glass beads.

[0120] Specifically, the method for distinguishing the quality of the reflective glass beads according to the matching result includes steps S501 - step S504.

[0121] Step S501: Construct an image standard feature library based on the threshold image features.

[0122] It should be clear that the threshold image features in step S501 include: the geometric features (diameter, roundness) of the reflective glass beads, the optical features (reflectivity, light intensity distribution), and the texture features (surface defects and internal decorative fillers).

[0123] Step S502: Real-time collect the reflection image features and match them with the image standard feature library, and use the matching algorithm to obtain the matching result.

[0124] Step S503: Formulate the quality grading rules according to the production indicators of the reflective glass beads.

[0125] It should be clear that in step S503, based on the production indicators, the quality grading rules are divided into qualified products (matching degree greater than or equal to 90%), repaired products (matching degree between 80% and 90%), and unqualified products (matching degree less than 80%).

[0126] Step S504: Make a quality grading determination for the reflective glass beads according to the corresponding relationship between the matching result and the quality grading rules.

[0127] As a preferred implementation manner, the matching algorithm in step S502 is ; are respectively used to represent the pixel average values of image x and image y, characterizing the brightness, are respectively used to represent the pixel variances of image x and image y, characterizing the contrast, is used to represent the covariance of image x and image y, characterizing the structural similarity, is a stability constant, used to avoid the denominator being 0, where , ,where L is the pixel dynamic range, with a value of 255, 。

[0128] It should be noted that the value range of the value obtained by the matching algorithm is (-1, 1). Therefore, the value obtained is corresponded to the quality grading rules by the percentage of the difference with the value range of (-1, 1), and then the quality grade of the reflective glass beads is determined. For example, if the value obtained is greater than or equal to the preset qualified threshold (such as the difference percentage corresponding to 0.9), it can be determined as a qualified product; if the value is within the preset repaired product threshold interval (such as the difference percentage range corresponding to 0.8 to 0.9), it is determined as a repaired product; if the value is less than the preset unqualified product threshold (such as less than the difference percentage corresponding to 0.8), it is determined as an unqualified product.

[0129] Specifically, when the pixel average values of image x and image y are 200, it characterizes moderate brightness. When the pixel variances of image x and image y are 50, it characterizes moderate contrast. When the covariance of image x and image y is 30, at this time , = 30, = = 50. Therefore , , so , so the obtained similarity is 74.75%, which is a defective product.

[0130] Step S600: Associate the matching result with the corresponding time series to obtain the abnormal time points of the reflective glass bead production line.

[0131] As a preferred embodiment, referring to Figure 2 it can be known that the present invention also proposes a rapid on-line detection device for the production of reflective glass beads. It should be clear that a rapid on-line detection device for the production of reflective glass beads uses the above-disclosed rapid on-line detection method for the production of reflective glass beads and includes:

[0132] LED spectral lamp, responsible for emitting a stable spectral source to irradiate the reflective glass beads;

[0133] Vision machine, used to obtain the threshold image and the reflection image, and responsible for collecting the reflection images formed by spectral irradiation of two adjacent reflective glass beads under different time series, and capturing the differences of the reflection images under each time series;

[0134] Positioning mark module, accurately identifying the mark points in the obtained threshold image through the characteristics of the reflective glass beads themselves;

[0135] Contact state discriminator, analyzing the positional relationship between two mark points in the reflection image to determine the contact state between two adjacent reflective glass beads;

[0136] Quality discriminant and correlation analyzer, retrieving and matching according to the threshold image characteristics, discriminating the quality of the reflective glass beads according to the matching result, and obtaining the internal correlation between the contact state between two adjacent reflective glass beads and the quality of the reflective glass beads;

[0137] Time series correlator, used to associate the matching result with the corresponding time series and obtain the time points when abnormal conditions occur in the production process of the reflective glass bead production line.

[0138] It should be added that in actual use of the present invention, the LED spectral lamp is installed above the production line of reflective glass beads to ensure that the spectral light source can evenly irradiate each reflective glass bead. The vision machine is installed on one side of the LED spectral lamp. The core components such as the positioning mark module, contact state discriminator, quality discriminant and correlation analyzer, and time series correlator are integrated in an intelligent control unit. The intelligent control unit is installed inside the vision machine. During actual operation, the intelligent control unit can quickly process and analyze the collected image data.

[0139] In addition, it should be noted that during actual operation, when the reflective glass beads pass through the production line, the LED spectral lamp emits a stable spectral source to irradiate the glass beads. The vision machine then captures the reflected image and the threshold image. The positioning mark module will identify the characteristic mark points of each reflective glass bead. The contact state discriminator analyzes the positional relationship of these mark points to determine the contact state between adjacent glass beads. Then, the quality discriminant and correlation analyzer retrieves the matching results according to the threshold image characteristics to discriminate the quality of the reflective glass beads and deeply analyzes the internal correlation between the contact state and the quality. Finally, the time series correlator correlates these analysis results with the corresponding time series. Once an abnormal situation on the production line is detected, the specific time point can be immediately located, facilitating timely processing and adjustment by the staff.

[0140] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended embodiments and their equivalents.

Claims

1. A rapid online detection method for the production of reflective glass beads, characterized in that: include: A threshold image of reflective glass beads is obtained using spectral source illumination; Based on the characteristics of the reflective glass beads, the marking points are determined in the threshold image; The reflection images formed by the spectrum of two adjacent reflective glass beads in the acquisition time series are different in each time series. Based on the positions of two marking points in the reflection image, the contact state of two adjacent reflective glass beads is determined; Among multiple reflective image features, search and match based on threshold image features, determine the quality of reflective glass beads based on the matching results, and obtain the correlation between the contact state between two adjacent reflective glass beads and the quality of the reflective glass beads; The matching results are associated with the corresponding time series to obtain the abnormal time points of the reflective glass beads production line.

2. A rapid online detection method for reflective glass beads production according to claim 1, characterized in that: The method for obtaining the threshold image includes: Collecting a primary spectral image, where the primary spectral image is formed when qualified reflective glass beads are illuminated by the maximum imaging angle of the spectral source; Collecting secondary spectral images, where the secondary spectral images are formed when qualified reflective glass beads are illuminated by the minimum imaging angle of the spectral source; Gray-scaling the primary spectral image and the secondary spectral image, and obtaining a primary grayscale histogram and a secondary grayscale histogram respectively; Stack the primary grayscale histogram and the secondary grayscale histogram, and obtain the threshold grayscale histogram; The threshold grayscale histogram determines the threshold image based on the binarization process.

3. The rapid online detection method for the production of reflective glass beads according to claim 1, characterized in that: Methods for determining marking points include: Identify the property characteristics of reflective glass beads based on visual machines; Collect spectral images produced by irradiating reflective glass beads with spectral sources at different angles; According to the classification of attribute features, the corresponding matching points are located in the spectral image; Select two spectral images with the largest difference in spectral source illumination angles, and retrieve the positional relationship between corresponding matching points in the two spectral images; Based on the evaluation method, a landmark point is determined among the multiple matching points.

4. A rapid online detection method for reflective glass beads production according to claim 3, characterized in that: The evaluation method includes: A two-dimensional coordinate system is constructed to represent the specifications of the spectral image, wherein the horizontal and vertical coordinates of the two-dimensional coordinate system correspond to the size specifications of the spectral image respectively; Mapping the spectral image into the constructed two-dimensional coordinate system; Extract the coordinate position of each matching point in the spectral image in the two-dimensional coordinate system; Select two spectral images with the largest difference in spectral source illumination angle, and associate the corresponding matching points in the two spectral images; Analyze the relationship between two corresponding matching points based on the spacing algorithm; Based on the analysis results, the landmark points are identified among the numerous matching points.

5. The rapid online detection method for the production of reflective glass beads according to claim 3, characterized in that: The spacing algorithm includes: ,in Indicates the spacing, Represent two points in n-dimensional space, and Respectively indicate points and Point The coordinate value in the i-th dimension, Indicates the number of dimensions of the space, i represents the index of the dimension, from 1 to n, The parameter representing the Minkowski distance is a positive real number parameter, and The value range is any integer between 1 and 2. =1, is the Manhattan distance, when p=2, Degenerates to Euclidean distance.

6. The rapid online detection method for the production of reflective glass beads according to claim 3, characterized in that: The spacing algorithm includes: ,in represents the spacing, where Represent two points in n-dimensional space, The covariance matrix T representing the data is used to adjust the vector direction and ensure that the matrix multiplication dimensions match, thereby converting multidimensional differences into scalar values ​​and realizing covariance-weighted distance calculation.

7. A rapid online detection method for reflective glass beads production according to claim 1, characterized in that: The contact status determination methods include: According to the physical properties of reflective glass beads, the specification limits of the threshold image are set; Based on the positioning of the marked points within the specification limits, a contrast image composed of two threshold images is obtained; Extract the distance between two marked points from the comparison image as the discrimination distance range; The contact status of adjacent reflective glass beads is determined by comparing the identification distance range with the distance between the two marking points in the reflection image. If the distance between the two marking points in the reflection image is within the identification distance range, it is determined that the adjacent reflective glass beads are in contact; If the distance between the positions of two marking points in the reflected image exceeds the identification distance range, it is determined that the adjacent reflective glass beads are in a non-contact state.

8. The rapid online detection method for reflective glass beads production according to claim 1, characterized in that: Methods for judging the quality of reflective glass beads based on matching results include: Construct an image standard feature library based on threshold image features; Collect reflection image features in real time and match them with the image standard feature library, and use the matching algorithm to obtain the matching results; Formulate quality grading rules based on the production indicators of reflective glass beads; According to the corresponding relationship between the matching results and the quality grading rules, the quality grading of the reflective glass beads is determined; The matching algorithm is ; They are used to represent the average pixel values ​​of image x and image y, representing the brightness. They are used to represent the pixel variance of image x and image y, respectively, and characterize the contrast. Used to represent the covariance of image x and image y, characterizing structural similarity, is a stability constant used to avoid the denominator being 0, where , , where L is the pixel dynamic range, the value is 255, .

9. A rapid online detection device for the production of reflective glass beads, using a rapid online detection method for the production of reflective glass beads as claimed in any one of claims 1 to 8, characterized in that: include: LED spectrum light, responsible for emitting a stable spectrum source to illuminate the reflective glass beads; The visual machine is used to obtain the threshold image and the reflection image, and is responsible for collecting the reflection images formed by two adjacent reflective glass beads under spectral illumination in different time series, and capturing the difference of the reflection images in each time series; The positioning marking module accurately identifies the marking point in the acquired threshold image through the characteristics of the reflective glass beads themselves; A contact state discriminator analyzes the positional relationship between two marking points in the reflection image to determine the contact state between two adjacent reflective glass beads; The quality discrimination and correlation analyzer searches and matches based on the threshold image features, and discriminates the quality of the reflective glass beads based on the matching results, and obtains the intrinsic correlation between the contact state between two adjacent reflective glass beads and the quality of the reflective glass beads; The time series associator is used to associate the matching results with the corresponding time series and obtain the time points when the reflective glass beads production line has abnormal conditions during the production process.

Images