Optical glass performance detection method and system

Abstract

The invention discloses an optical glass performance detection method and system, and relates to the technical field of optical detection.The method comprises the steps that according to application classification of optical glass, physical defect types are determined, a multi-channel bright and dark field is configured, and an optical path polarization test is conducted through incident light to obtain an optical detection coefficient. A sensing image set is collected through cooperative operation of a light controller and an image sensor, defect detection and position splicing are completed in combination with an image processing module, and a defect detection list is generated. And integrating the optical detection coefficient and the defect detection list to carry out weighting calculation, and finally obtaining a performance detection result of the detection target. The technical problem that the recognition precision of various physical defects in existing optical glass performance detection is insufficient is solved, and the technical effect of improving the recognition precision of various physical defects in optical glass detection is achieved.

Description

Technical Field

[0001] The present application relates to the field of optical detection technology, and in particular to a method and system for detecting the performance of optical glass. Background Art

[0002] With the widespread application of optical glass in precision instruments, electronic displays, aerospace and other fields, its performance requirements are increasing, especially the identification of surface and internal physical defects, and the accurate evaluation of overall optical performance, which have become the key to optical glass quality inspection. However, when faced with complex defect types, existing detection methods usually rely only on a single light source or simple manual observation, and cannot achieve high-precision detection of multiple physical defects. In addition, traditional methods lack comprehensive quantitative analysis of key indicators such as optical path polarization and optical consistency in optical performance evaluation, resulting in poor reliability and repeatability of detection results. To address this issue, there is an urgent need for an advanced detection method that can accurately identify multiple physical defects and combine quantitative analysis of optical performance to meet the strict requirements of high-end applications for optical glass quality.

[0003] In the current related technologies, there is a technical problem of insufficient accuracy in identifying various physical defects in optical glass performance testing. Summary of the invention

[0004] The present application provides an optical glass performance detection method and system, which determines the type of physical defects according to the application classification of optical glass, configures multi-channel bright and dark fields, and uses incident light to perform optical path polarization test to obtain optical detection coefficients. The sensor image set is collected through the coordinated operation of the light controller and the image sensor, and the defect detection and position stitching are completed in combination with the image processing module to generate a defect detection list. The comprehensive optical detection coefficient and the defect detection list are weighted and calculated, and the performance detection results of the detection target are finally obtained, achieving the technical effect of improving the recognition accuracy of various physical defects in optical glass detection.

[0005] The present application provides a method for detecting optical glass performance, comprising: According to the application classification of optical glass, the physical defect type of the target optical glass type is determined, and based on the defect contrast, a multi-channel light and dark field is configured; the incident light is introduced, and the optical path polarization test is performed on the detection target to determine the optical detection coefficient, wherein the polarization test includes a single-chip entity test and an array simulation test; based on the multi-channel light and dark field, the coordination of the light controller and the image sensor is established, and the image sensor performs light control tracking to obtain a sensor image set; the sensor image set is transmitted back, and the boundary segmentation is performed in combination with the image processing module, so as to determine the defect detection list based on the self-attention detection and position stitching of the physical defect type, and the defect detection list includes a sequence based on defect features-defect coefficients; based on the optical detection coefficient and the defect detection list, a weighted calculation is performed to determine the performance detection result of the detection target.

[0006] The present application also provides an optical glass performance detection system, comprising: A physical defect type determination module, the physical defect type determination module is used to determine the physical defect type of the target optical glass type according to the application classification of optical glass, and configure a multi-channel light and dark field based on the defect contrast; an optical path polarization test module, the optical path polarization test module is used to introduce incident light, perform an optical path polarization test on the detection target, and determine the optical detection coefficient, wherein the polarization test includes a single-chip entity test and an array simulation test; a sensor image set acquisition module, the sensor image set acquisition module is used to establish a collaboration between the light controller and the image sensor based on the multi-channel light and dark field, and the image sensor performs light control tracking to obtain a sensor image set; a boundary segmentation module, the boundary segmentation module is used to return the sensor image set, and perform boundary segmentation in combination with the image processing module, so as to determine a defect detection list based on self-attention detection and position splicing of the physical defect type, and the defect detection list includes a sequence based on defect features-defect coefficients; a weighted calculation module, the weighted calculation module is used to perform weighted calculation based on the optical detection coefficient and the defect detection list to determine the performance detection result of the detection target.

[0007] The optical glass performance detection method and system proposed in this application first determines the type of physical defects according to the application classification of optical glass, configures multi-channel light and dark fields, and uses incident light to perform optical path polarization test to obtain optical detection coefficients. The sensor image set is collected through the coordinated operation of the light controller and the image sensor, and the defect detection and position stitching are completed in combination with the image processing module to generate a defect detection list. The comprehensive optical detection coefficient and the defect detection list are weighted and calculated, and finally the performance detection result of the detection target is obtained, achieving the technical effect of improving the recognition accuracy of various physical defects in optical glass detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0009] Figure 1 A schematic flow chart of a method for detecting optical glass properties provided in an embodiment of the present application.

[0010] Figure 2 A schematic diagram of the structure of an optical glass performance detection system provided in an embodiment of the present application.

[0011] Explanation of the reference numerals: physical defect type determination module 10 , optical path polarization testing module 20 , sensor image set acquisition module 30 , boundary segmentation module 40 , weighted calculation module 50 . DETAILED DESCRIPTION

[0012] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0013] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0014] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0015] The present application embodiment provides a method for detecting optical glass performance, such as Figure 1 As shown, the method includes: Step S100, for the application classification of optical glass, determine the physical defect type of the target optical glass type, and configure multi-channel bright and dark fields based on the defect contrast. Specifically, for the application classification of optical glass, first study in detail the requirements of its application scenarios for optical performance, such as telescope lenses require high resolution, low chromatic aberration and good transmittance, microscope objective lenses require high-precision optical uniformity, and display screen protection glass focuses on hardness, wear resistance and no display interference, etc., and then identify the physical defect types that may appear in the target optical glass type, such as tiny bubbles, internal impurities, and uneven refractive index of telescope lens optical glass, and the glass used for microscope objective lenses also needs to pay attention to stress concentration, and the scratches and poor surface flatness of the display screen protection glass, etc., and organize these potential physical defect types into a list. Then, for each physical defect type, use professional equipment and methods to measure its defect contrast with the normal area, study the change law of this contrast under various lighting conditions, and finally determine the configuration parameters of the multi-channel bright and dark fields based on the analysis results. Each channel corresponds to a specific lighting condition, including intensity, angle, color and lighting mode, to ensure that different physical defects have high contrast in the corresponding channel, which is convenient for subsequent image acquisition and analysis.

[0016] In a possible implementation, for the application classification of optical glass, the physical defect class of the target optical glass type is determined, and based on the defect contrast, a multi-channel bright and dark field is configured, and step S100 further includes step S110, traversing the physical defect class, performing a clustering based on the defect contrast, and determining the first clustering result. Specifically, firstly, the data related to all physical defect classes of the target optical glass type previously determined are sorted and prepared to ensure their integrity and accuracy, and these data include the characteristic description of various physical defects, the position in the optical glass, the degree of influence on the optical performance, and other information, and then these physical defect classes are traversed and their characteristics are analyzed one by one. For each physical defect class being traversed, the defect contrast between it and the normal optical glass area is focused on, and the contrast value and change of each defect class with the normal area under different lighting conditions and different observation angles are accurately obtained through professional optical measurement equipment and analysis methods, such as measuring the contrast difference between the defect of uneven refractive index and the surrounding normal glass area in the light refraction effect under uniform illumination of different intensities, and analyzing the contrast between the scratch defect and the surrounding flat surface in terms of reflected light intensity when obliquely illuminated. Then, according to the relevant data of each physical defect class based on defect contrast, a clustering is performed using a suitable clustering algorithm (such as K-Means clustering algorithm, hierarchical clustering algorithm, etc.). Based on the similarity of defect contrast, physical defect classes with similar defect contrast characteristics are classified into the same cluster cluster. For example, physical defect classes with similar contrast change trends and similar contrast value ranges under specific lighting conditions will be classified into the same cluster cluster. Finally, the first clustering result based on defect contrast is obtained through clustering algorithm calculation. The result is presented as several cluster clusters. Each cluster cluster contains a group of physical defect classes with similar defect contrast characteristics. The division of these cluster clusters will provide an important basis for the subsequent optimization of multi-channel bright and dark field configurations, because defects in different cluster clusters may require different lighting conditions to achieve the best detection effect.

[0017] Step S120, traverse the physical defect classes, perform secondary clustering based on the lighting type, and determine the second clustering result. Specifically, after completing the first clustering based on defect contrast, it is necessary to traverse all physical defect classes again. This traversal must ensure that the relevant information of each physical defect class is comprehensively and carefully reviewed, covering all types of features analyzed previously and other factors that have not been fully considered, so as to be fully prepared for the next clustering based on lighting type. Next, for each physical defect type being traversed, we focus on analyzing it from the perspective of lighting type, considering the impact of different lighting types (such as backlight, ring light, coaxial light, etc.) on the presentation of various physical defects. For example, backlight lighting will cause certain internal impurity defects to appear in optical glass with specific shadows or bright area distribution characteristics. Ring light lighting can make surface scratch defects produce unique light reflection and refraction effects, highlighting the shape and position of the scratches more clearly. Coaxial light lighting can provide more accurate light propagation path information when detecting defects with uneven refractive index, which helps to observe the impact of defects on light refraction. By conducting a detailed analysis of the presentation effects of different physical defect types under various lighting types, we can obtain the characteristic performance data of each type of physical defect under different lighting types. Then, according to the relevant data of each physical defect class based on the lighting type, a suitable clustering algorithm is used for secondary clustering. This clustering is based on the similarity of the characteristic performance under the lighting type. The physical defect classes with similar presentation effects under different lighting types are classified into the same cluster cluster. For example, those physical defect classes that present similar light reflection and refraction characteristics when illuminated by ring light will be classified into the same cluster cluster. Finally, after the operation of the secondary clustering algorithm, the second clustering result based on the lighting type is obtained. The result is also presented as several clusters. Each cluster contains a group of physical defect classes with similar presentation effects under a specific lighting type. The division of these clusters further refines the classification of physical defect classes and provides a more targeted basis for more accurate configuration of multi-channel light and dark fields, because the defects of different clusters have different detection requirements under different lighting types.

[0018] Step S130, fuse the first clustering result and the second clustering result, and configure the multi-channel light and dark field, wherein the multi-channel light and dark field is determined based on the lighting type and the field brightness. Specifically, the first clustering result obtained by the primary clustering based on defect contrast and the second clustering result obtained by the secondary clustering based on the lighting type are fused. When fusion is performed, the physical defect class information contained in each cluster cluster in the two clustering results and the correlation between them should be comprehensively considered. For example, for a physical defect class that belongs to a certain cluster in the clustering based on defect contrast and another cluster in the clustering based on lighting type, it is necessary to analyze its characteristics and requirements from these two different clustering perspectives, so as to take full account of both when configuring the multi-channel light and dark field. According to the fused clustering results, the specific needs of configuring multi-channel light and dark fields are analyzed in depth. Since the multi-channel light and dark fields are determined by the lighting type and field brightness, the settings of each channel should be determined in combination with the requirements of the lighting type and field brightness of the physical defect types in different clustering clusters. For example, for the physical defect type in a cluster in the fused clustering results, if the ring light lighting effect is best when clustering based on the lighting type and a medium brightness field is required to highlight the defects when clustering based on the defect contrast, when configuring the multi-channel light and dark field, a channel with ring light lighting and medium brightness should be set to meet the detection needs of the physical defect type in this cluster. According to the configuration requirements obtained through analysis, multi-channel light and dark fields are actually configured to ensure the completeness of defect collection. Each channel is set by adjusting the brightness of the field of view and selecting the appropriate lighting type (such as different brightness under backlight, ring light, coaxial light, etc.). For example, channels are set separately for various types of physical defects in different clusters in the fusion clustering results. One channel is set to backlight illumination with a lower field of view brightness to detect physical defects in a cluster that can be better presented under such lighting conditions. Another channel is set to coaxial light illumination with a higher field of view brightness to detect physical defects in another cluster that are more suitable for detection under such lighting conditions. Through configuration, each channel provides the best detection conditions for a specific type of physical defect, thereby comprehensively and accurately collecting relevant information on different defects on the optical glass to ensure the completeness of defect collection.

[0019] Step S200, introduce incident light, perform optical path polarization test on the detection target, and determine the optical detection coefficient, wherein the polarization test includes single-piece entity test and array simulation test. Specifically, first select a suitable incident light source, determine its wavelength, intensity, polarization direction and coherence and other parameters according to the characteristics of the optical glass and the requirements of the detection target, and then accurately guide the light to the detection target. Then perform a single-piece entity test, accurately and stably place the single-piece optical glass sample to be tested in the test light path, align it with the optical axis of the test equipment, and use a polarization detection device (such as a polarizer, a wave plate and a photodetector combination) to measure the change in polarization state of the incident light after passing through the single-piece optical glass. By rotating the polarizer and using Malus's law to calculate the change in polarization direction and polarization degree, use a wave plate to analyze the phase difference in the birefringence phenomenon, accurately record and analyze the change data of the polarization state of light at different positions and angles, and preliminarily understand the polarization characteristics of the single-piece optical glass. After that, array simulation tests were carried out, and computer simulation software was used to build an array model that can truly reflect the arrangement and optical properties of optical glass in actual applications. Incident light with the same characteristics as the single-piece entity test was introduced into the model to simulate the propagation process of light in the array. The boundary conditions of light at different glass interfaces, the interference effect caused by multiple reflections and refractions, and the changes in macroscopic polarization characteristics caused by the overall structure of the glass array were considered. The changes in the polarization state of the simulated light after propagation in the array were analyzed in detail, with a focus on the overall polarization characteristics under the interaction of multiple optical glasses. Finally, the optical detection coefficient was determined by combining the data of the single-piece entity test and the array simulation test. The coefficient can be obtained by weighted calculation of the polarization degree change parameters in the single-piece test and the overall polarization state change related parameters in the array simulation, or by establishing a specific mathematical model. This coefficient is used as the basis for evaluating the quality and performance of optical glass and judging whether it meets the requirements of a specific optical system.

[0020] In one possible implementation, incident light is introduced, and a light path polarization test is performed on the detection target to determine the optical detection coefficient, wherein the polarization test includes a single-chip entity test and an array simulation test, and step S200 further includes step S210, determining a light path polarization threshold based on the optical qualification. Specifically, the optical qualification is first determined based on the specific application fields of optical glass and the corresponding industry standards and technical specifications. Different application scenarios (such as camera lenses, microscope lenses, telescope lenses, etc.) have very different requirements. For example, the optical glass used for high-end camera lenses has extremely high optical qualification requirements, and the allowable deviation range of optical properties such as light refraction, reflection, and polarization is very small. The optical qualification requirements of ordinary optical instrument protective glass are relatively low. These standards are jointly determined by a large amount of experimental data, actual use feedback, and optical theoretical calculations. Then, on the basis of clarifying the optical qualification standards, the light path polarization threshold is determined by combining theoretical calculations and experimental verification. From a theoretical level, the maximum allowable value of the light path polarization that meets the optical qualification conditions is calculated based on the optical glass material properties, design purposes, and light propagation theory. At the same time, optical glass samples of different quality grades are used to measure the polarization in the standard test light path. Combined with the performance in actual use, a reasonable threshold is finally determined. If the optical glass is used in a high-precision optical imaging system, its light path polarization threshold may be set to a very low value to ensure imaging clarity and accuracy.

[0021] Step S220, identify the test light path, calculate the polarization angle and determine whether the light path polarization threshold is met. Specifically, when performing polarization testing, the test light path must first be accurately identified, including determining the source, direction, wavelength and other parameters of the incident light and the propagation path of the light in the optical system, that is, the optical elements (such as lenses, reflectors, etc.) and their arrangement. The entire test light path is modeled and analyzed through precise optical instruments and measurement methods to ensure a clear understanding of each link of the light path. For example, in the test of a complex optical lens group, the refraction, reflection and possible polarization change process of the light from the light source through each lens must be recorded in detail, and the geometric parameters of the light on the surface of each lens, such as the incident angle and refraction angle, must be clarified. These parameters are crucial for the subsequent calculation of the polarization angle. After identifying the test optical path, the polarization angle is calculated using professional polarization measurement equipment and technology such as polarizers, wave plates, and photodetectors. The polarization angle is calculated based on the change in the direction of the electric field vector after the light passes through the optical glass, by measuring the light intensity in different directions, and using optical laws such as Malus's law. During the calculation process, factors such as the birefringence of the optical glass and the polarization change caused by surface reflection that the light is subjected to during the propagation of the entire test optical path must be considered. For example, when the light passes through the optical glass with birefringence characteristics, ordinary light and extraordinary light with different polarization directions will be generated, and the synthesized polarization angle must be accurately measured and calculated to determine. The calculated polarization angle is then compared with the previously determined optical path polarization threshold. If the polarization angle is less than or equal to the optical path polarization threshold, it means that the optical glass preliminarily meets the optical qualification requirements in terms of polarization characteristics. If the polarization angle is greater than the optical path polarization threshold, it means that its polarization characteristics exceed the allowable range, which may have a negative impact on the performance of the optical system, such as causing blurred imaging, uneven light propagation, and other problems. This comparison and judgment is a key step in evaluating the quality of optical glass and directly determines the subsequent treatment of the optical glass.

[0022] Step S230, if the polarization angle is greater than the light path polarization threshold, a defective mark is generated. Specifically, when the polarization angle is greater than the light path polarization threshold, the optical glass is marked as defective according to a pre-set quality assessment rule, because for most optical applications, the light path polarization exceeds the threshold, indicating that its optical performance is not up to standard. In a high-precision optical system, even a small polarization deviation may cause a serious performance degradation. For example, in the optical system of a microscope or telescope, excessive polarization will cause image distortion and reduced resolution, affecting the observation effect. Subsequently, a clear defective mark is generated for the optical glass through the software or identification device of the test system. This mark can be a digital code, a specific symbol, or a piece of defective information recorded in a database, so as to quickly and accurately identify optical glass that does not meet quality requirements in the subsequent production process or quality control process, prevent them from entering the subsequent assembly or use link, and ensure the quality of the entire optical product.

[0023] In one possible implementation, incident light is introduced, and a light path polarization test is performed on the detection target to determine the optical detection coefficient, wherein the polarization test includes a single-chip entity test and an array simulation test, and step S200 further includes step S240, if the polarization angle is less than the light path polarization threshold, the detection target is subjected to physical twinning and array light path detection in combination with the simulation platform to determine the total polarization angle, wherein the parallel arrangement of multiple targets is a matrix distribution. Specifically, when it is detected that the polarization angle is less than the light path polarization threshold, because direct testing may have insufficient accuracy for high-precision optical glass, it is necessary to use an advanced simulation platform to create a physical twin model for the detection target, which is a high-precision digital mapping of the actual optical glass and its light path environment, including information such as the geometric shape and material properties of the optical glass, and can also accurately simulate the surrounding light path environment, including other optical components and their relative positional relationships. In the simulation platform, multiple detection targets are arranged in parallel as a matrix distribution to simulate actual applications such as optical lens arrays or multi-lens combinations in complex optical instruments. For the array optical path, each detection target and its interactions are analyzed in detail, and the propagation path of light in the array, the refraction, reflection and polarization changes on each detection target are accurately simulated and calculated. In the process, the distance and angle between multiple targets as well as the multiple propagations and mutual influence of light are considered, so as to comprehensively detect the polarization characteristics of each detection target in the array optical path. Through the simulation calculation of the physical twin model and array optical path detection, the polarization changes of each detection target in the entire array optical path are comprehensively considered to determine the total polarization angle. It reflects the overall polarization state change of light after passing through all detection targets in a complex multi-target and array optical path environment, and is an important intermediate parameter for the subsequent precise evaluation of optical glass performance.

[0024] Step S250, based on the number of arrays, average the total polarization angle to determine the target polarization angle. Specifically, after the total polarization angle is obtained, average calculation is performed according to the number of detection targets in the array, because in the array light path, each detection target contributes to the overall polarization situation, and calculating the average can more accurately reflect the average polarization characteristics of a single detection target in such a complex environment. For example, if there are detection targets, and the polarization angle corresponding to each target is , then the total polarization angle is the sum of all these polarization angles, and the target polarization angle is this sum divided by ,Through the mean calculation, a target polarization angle is obtained that is more representative of the polarization characteristics of the detection target in the case of multi-target and array optical path. The target polarization angle eliminates the influence of local polarization changes that may be caused by different positions and interactions in the array, providing a more stable and representative parameter for the subsequent quality evaluation.

[0025] Step S260, match the quality grade of the target polarization angle and determine the optical detection coefficient. Specifically, first, a complete set of quality grade classification standards is established based on the application requirements and industry standards of optical glass, and the value range of the target polarization angle is divided into different intervals, each interval corresponds to a specific quality grade. For example, for high-precision imaging optical glass, the interval with extremely small target polarization angle corresponds to the highest quality grade, which means that the optical glass has excellent performance in polarization characteristics, and the quality grade gradually decreases as the target polarization angle increases. Then, the calculated target polarization angle is matched with the quality grade classification standard, and the corresponding quality grade is determined according to the interval in which it is located. This matching process is a key step in accurately evaluating the quality of optical glass, and can convert the target polarization angle, a quantitative indicator of polarization characteristics, into grade information with practical quality significance. Finally, the optical detection coefficient is determined according to the quality level of the target polarization angle match. The optical detection coefficient is an indicator that comprehensively reflects the quality of the polarization characteristics of the optical glass and is related to the quality level. A high quality level corresponds to a higher optical detection coefficient, indicating that the optical glass has excellent polarization performance. A low quality level corresponds to a lower optical detection coefficient, indicating that the optical glass may have quality problems in polarization. The optical detection coefficient can be used to conveniently compare, screen and evaluate the quality of optical glass to meet the requirements of different optical application scenarios.

[0026] Step S300, based on the multi-channel light and dark field, establish the coordination between the light controller and the image sensor, the image sensor performs light control tracking, and obtains a set of sensor images. Specifically, firstly, the configured multi-channel light and dark field is deeply analyzed to understand the illumination characteristics of each channel composed of different lighting types and field brightness, including intensity, angle, color and the resulting light and dark effects. On this basis, the light controller and the image sensor are connected through a suitable hardware interface and a communication protocol is established to specify the data transmission format, rate and control signal encoding method. At the same time, a parameter synchronization mechanism is established so that the illumination parameters of the light controller and the parameters of the image sensor such as exposure time, gain, resolution, frame rate, etc. can be synchronously adjusted according to the multi-channel light and dark field and the detection target characteristics. When the light controller changes the lighting conditions according to the multi-channel light and dark field settings, the image sensor follows the real-time light control, automatically adjusts the exposure and gain according to the changes in light intensity, adjusts the focusing parameters when the lighting angle changes, and collects images of the optical glass in all directions, scanning from different angles and positions. The mobile image sensor or multi-view imaging technology is used to ensure that the entire inspection area is covered, and the image is kept stable and continuous. Finally, all images collected under different lighting conditions are combined to construct a sensor image set, which serves as an important data source for subsequent defect detection and analysis.

[0027] In one possible implementation, based on the multi-channel light and dark field, a light controller and an image sensor are coordinated, the image sensor performs light control tracking, and a sensor image set is obtained. Step S300 further includes step S310, for the detection target, determining the azimuth trajectory of the image sensor for full-field detection of the target. Specifically, firstly, the geometric shape and size of the detection target, i.e., optical glass, are analyzed in detail. Its shape may be circular, rectangular, polygonal, etc. and of different sizes. Understanding these geometric features is crucial to determining the detection azimuth trajectory of the image sensor. For example, a circular optical glass lens needs to consider the angle of radiation from the center of the circle and the angle change in the circumferential direction. A rectangular optical glass needs to determine the scanning angle range along the long side and the short side. Then, according to the detection accuracy requirements and the principle of full coverage of optical glass, the appropriate azimuth step and angle range are determined. If high-precision detection is required, the azimuth step should be set smaller to ensure that every tiny area on the surface and inside of the optical glass can be detected. For coverage requirements, the azimuth trajectory must cover the entire optical glass including the edge and center area. For example, the azimuth step of high-precision optical lens detection may be set to 1 degree or even smaller, and the angle range covers 360 degrees in the circumferential direction and a certain angle range perpendicular to the lens plane to detect the internal situation of the lens. Finally, the azimuth trajectory of the image sensor for full-domain detection of the target is planned based on the geometric characteristics, detection accuracy and coverage requirements. The trajectory can be represented by a mathematical model or a coordinate system to clarify the path of the angular position change of the image sensor relative to the detection target during the detection process. For example, for circular optical glass, a series of spiral trajectories centered on the center of the circle, starting from different starting angles and gradually changing with a specific step can be planned. At the same time, a three-dimensional azimuth trajectory is formed by combining the angle change in the vertical direction to ensure comprehensive detection of the entire circular optical glass.

[0028] Step S320, based on the multi-channel light and dark field, multiple groups of light control parameters are determined. Specifically, the characteristics of each channel in the multi-channel light and dark field are studied in depth, such as the type of illumination (bright field, dark field, oblique illumination at different angles, etc.), illumination intensity, illumination color, and the switching logic between channels. Since each channel is designed to highlight specific physical defects, its illumination characteristics directly affect the determination of light control parameters. For example, the dark field channel used to detect tiny bubbles has low illumination intensity and special light angles, which can make bubbles appear as bright points against a dark background, while the illumination intensity and uniformity requirements of the bright field channel for detecting surface scratches are different from those of the dark field channel. Then, the corresponding light control parameters are determined according to the characteristics of each channel, including the output power of the light controller, illumination duration, illumination frequency (when pulsed illumination), and the mixing ratio of different color lights (when colored illumination). These parameter values ​​are different for different channels. For example, the output power of the light controller for the dark field channel should be set to a lower level to achieve lower background brightness, and the output power of the high-brightness bright field channel should be increased accordingly. At the same time, the illumination duration is determined according to the optimal imaging time of the detection target under different illumination to ensure that the image sensor captures a clear image. Finally, by analyzing and determining the parameters of each channel of the multi-channel bright and dark field, multiple sets of light control parameters are generated. Each set of parameters corresponds to a specific channel, which is used to guide the light controller to accurately control the illumination conditions to meet the needs of different physical defect detection.

[0029] Step S330, according to the multiple sets of light control parameters, control the light controller to adjust the illumination light, and the image sensor follows to collect images of the detection target to determine the sensor image set. Specifically, according to the generated multiple sets of light control parameters, the light controller accurately controls the lighting conditions, adjusts the lighting output of each channel according to the parameter settings, and involves operations such as turning on and off specific channel lighting, adjusting lighting intensity, and changing lighting color. For example, when a dark field channel is required for detection, the light controller will reduce the output power and adjust the light angle according to the light control parameters corresponding to the channel to create a dark field lighting environment that meets the requirements, and switch different channels in a preset order and time interval during the entire detection process to achieve switching of multiple lighting conditions. At the same time, when the light controller is performing lighting regulation, the image sensor follows to collect images of the detection target. Every time the light controller adjusts the lighting conditions, the image sensor performs a full-range image acquisition. During the acquisition process, the image sensor will automatically adjust its own parameters according to the current lighting conditions, such as exposure time, gain, white balance, etc., to adapt to the changes in different light intensities and colors, and ensure the acquisition of high-quality images. For example, when the light intensity drops suddenly, the exposure time is automatically increased to ensure that the image is not too dark, and when the light color changes, the white balance is adjusted to make the image color accurate. In addition, the image sensor scans the detection target according to the previously determined azimuth trajectory, and collects images in all directions to ensure that the entire detection area is covered. In the process of image sensor tracking image acquisition, all the collected images are combined to construct a sensor image set. The sensor image set covers the images of optical glass collected under different lighting conditions determined by multi-channel bright and dark fields and light control parameters, and at different azimuths determined by the azimuth trajectory. These images contain rich information and provide a comprehensive data basis for subsequent image analysis and defect detection. By processing the sensor image set, various physical defects in optical glass can be further discovered.

[0030] Step S400, the sensor image set is transmitted back, and the boundary segmentation is performed in combination with the image processing module, so as to determine the defect detection list based on the self-attention detection and position splicing of the physical defect class, and the defect detection list contains a sequence based on defect feature-defect coefficient. Specifically, the sensor image set is transmitted back to the detection system processing center through a communication link coordinated with the image sensor and the light controller to ensure stable and complete data transmission, and a high-speed data transmission protocol and error correction mechanism can be adopted. After the image set arrives at the image processing module, it is first pre-processed, including adjusting the brightness and contrast, removing noise, such as using a filtering algorithm to reduce salt and pepper or Gaussian noise, improving image clarity, and highlighting the physical defect characteristics. Then, a suitable boundary segmentation algorithm (such as a method based on threshold, edge detection or region growth) is used to process the pre-processed image. Taking the edge detection method as an example, the Sobel operator, Canny operator, etc. are used to detect the sudden change of pixel gray value to determine the object boundary, which helps to analyze the shape, size and position of physical defects. Then, a self-attention detection mechanism based on physical defects is introduced. The model assigns different attention weights to different types of physical defects (such as bubbles, scratches, impurities, etc.) by learning sample data. For example, when detecting bubbles, it focuses on circular or elliptical areas with different grayscale values, and when detecting scratches, it focuses on linear grayscale change areas. At the same time, the detected physical defects are spliced, and the position information of the same physical defect under different viewing angles and lighting conditions is integrated to form a comprehensive description. Finally, a defect detection list based on the defect feature-defect coefficient sequence is generated. The defect features cover the type, shape, size, and position of physical defects. The defect coefficient is determined based on the severity of the defect and the degree of impact on optical performance. For example, the defect coefficient of large-area scratches is high, and the defect coefficient of small bubbles is low. The defect detection list is used as the basis for evaluating the quality of optical glass and provides data support for subsequent decision-making.

[0031] In a possible implementation, the sensor image set is transmitted back, and boundary segmentation is performed in combination with the image processing module, so as to determine a defect detection list based on self-attention detection and position splicing of physical defect types, and the defect detection list includes a sequence based on defect features-defect coefficients. Step S400 further includes step S410, performing boundary segmentation on the sensor image set to determine a valid image set, wherein the boundary between the detection target and the background is used as the basis for segmentation. Specifically, a boundary segmentation operation is first performed on the sensor image set, because the segmentation is based on the boundary between the detection target, i.e., the optical glass, and the background, and a suitable image segmentation algorithm needs to be selected. A threshold-based segmentation method can be used to determine a suitable threshold by analyzing the distribution of pixel grayscale values ​​in the image to distinguish the detection target from the background. For example, if the grayscale value of the optical glass in the image is relatively high and the background grayscale value is low, a grayscale value between the two is selected as the threshold, so that pixels greater than the threshold are classified as the optical glass part, and pixels less than the threshold are classified as the background part. However, in actual operation, the boundary between the detection target and the background may be uneven due to uneven lighting, optical The image is blurred due to factors such as reflection on the glass surface. At this time, a more complex segmentation algorithm is required, such as the method based on edge detection. The boundary is determined by detecting the area in the image where the grayscale value changes dramatically. For example, the Canny edge detection algorithm can accurately detect the edge in the image while suppressing noise, find the optical glass boundary more accurately, and reduce the segmentation error when the boundary is blurred. After boundary segmentation, the image area belonging to the detection target is extracted to form a valid image set. Each image in the valid image set only contains information related to the optical glass, and irrelevant information such as the background is removed, providing a purer data basis for subsequent defect detection.

[0032] Step S420, traverse the valid image set to determine the defect detection set based on the self-attention detection of the corresponding physical defect class. Specifically, after obtaining the valid image set, start traversing each image therein, and start the self-attention detection mechanism based on a large amount of physical defect sample data trained based on the corresponding physical defect class for each image. It can automatically learn and assign different attention weights according to the characteristics of different physical defect classes (such as bubbles, scratches, impurities, etc.). For example, for bubble defects, the model focuses on the circular or elliptical areas in the image with relatively low grayscale values ​​(assuming that bubbles appear as dark areas in the image) and relatively smooth edges. For scratch defects, it pays more attention to the linear areas with obvious changes in grayscale values. In the process of traversing the valid image set, the self-attention detection mechanism is used to analyze each image. When areas matching the physical defect class features are detected, these areas are marked as potential defect areas. For example, in a valid image, if the self-attention detection mechanism finds a circular dark area that meets the bubble features, this area is recorded as a possible bubble defect. By traversing the entire valid image set, all detected potential defect areas are combined to form a defect detection set. In order to improve the accuracy of defect detection, some post-processing methods can be combined in the process of determining the defect detection set, such as morphological processing of the detected potential defect areas, removing small false defect areas caused by noise or image segmentation errors, and further verifying and screening the detected defect areas based on some prior knowledge of physical defects to ensure the accuracy of the defect detection set.

[0033] Step S430, according to the relative spatial position of the detection target, the defect detection set is distributed and spliced, and the defect detection list is determined by integration. Specifically, after obtaining the defect detection set, the relative spatial position information of the detection target (optical glass) needs to be considered, because the position of the physical defects in different images is relative to the optical glass, so these spatial position information should be analyzed. For example, each effective image may only cover a part of the optical glass, and the position of the physical defects is known. Through analysis, the relative position of the physical defects in different images on the entire optical glass can be determined. Then, according to the relative spatial position of the detection target, the defect information in the defect detection set is distributed and spliced, that is, the physical defects detected in different images are combined according to their actual position relationship on the optical glass. For example, if a scratch is detected in the upper left corner of the optical glass in one image, and a bubble is detected in the upper right corner adjacent to it in another image, these two defects can be correctly spliced ​​in space through spatial position analysis and distribution splicing to form a more complete description of the distribution of physical defects on the optical glass. Finally, all defect information is integrated through distributed splicing to determine the defect detection list, which contains detailed information of all physical defects detected on the optical glass, such as defect type (bubbles, scratches, etc.), location, size, shape, etc. This information will provide important basis for optical glass quality assessment, defect analysis and repair decisions. At the same time, the format of the defect detection list can be designed according to actual needs for easy storage, query and further processing.

[0034] In one possible implementation, the valid image set is traversed to determine a defect detection set based on self-attention detection of corresponding physical defect classes. Step S420 further includes step S421, configuring a self-attention mechanism based on the physical defect class and defect identification factors. Specifically, firstly, a comprehensive analysis of the physical defect class is performed. The physical defects of optical glass include different types such as bubbles, scratches, impurities, cracks, etc. Unique defect identification factors are determined for each type. For example, bubbles are often circular or elliptical low-grayscale areas with smooth edges, scratches are linear grayscale change areas, and impurities in the image present grayscale values ​​and shapes that are different from the material of the optical glass itself. These factors are the key basis for configuring the self-attention mechanism. At the same time, the complexity of multi-lighting conditions must be considered. When there is detection under multi-lighting, although the acquisition under one lighting condition is mainly for the detection of certain defect types, other types of defect images may be collected incidentally. These defects are in the current The clarity and other conditions may be poor under certain lighting conditions. For example, under a certain oblique lighting method, scratches are mainly detected but bubbles may be captured and the bubble image is not clear enough and the contrast is not high. The self-attention mechanism is configured according to the identification factors of different physical defect types and the complex situation of multiple lighting. The model can focus on the part of the image related to the defect to be identified. For example, when training the self-attention mechanism, a higher attention weight is given to the defect type (such as scratches) that is mainly detected under specific lighting, and the weights of other defects (such as bubbles) that may be detected incidentally are adjusted according to their characteristics under the lighting conditions. This allows the model to accurately focus on the defect information to be detected under various lighting conditions and ignore the interference caused by poor lighting conditions.

[0035] Step S422, according to the light and dark field sources of each group of images in the effective image set, a mapping is established with the self-attention mechanism to determine the self-attention mechanism-image group. Specifically, each group of images in the effective image set is analyzed to obtain its light and dark field source information. Different light and dark field sources (such as bright field, dark field, lighting of different angles and intensities, etc.) will cause the physical defects of the optical glass to present different characteristics in the image. For example, bright field lighting is more conducive to detecting scratches and impurities on the surface, because it can make these defects form a more obvious contrast with the surrounding under a bright background, while dark field lighting is more effective for detecting internal bubbles or tiny cracks. These defects will appear as bright areas under a dark background due to light scattering and other reasons. Then, based on the characteristics of the light and dark field sources of each group of images and the configured self-attention mechanism, a mapping relationship between the two is established. For the image group under specific light and dark field source conditions, it is associated with a self-attention mechanism that can effectively detect physical defects under such conditions. For example, a group of images from dark field illumination will be associated with a self-attention mechanism optimized for bubble and crack detection, because these defects are easier to find and analyze in dark field images, and the image group of bright field illumination is mapped with a self-attention mechanism for scratch and impurity detection. Finally, the self-attention mechanism-image group is determined through this mapping relationship. Each such combination targets a specific light and dark field source and physical defect type, so that subsequent defect detection can more accurately use the appropriate self-attention mechanism to carry out analysis in the corresponding image group.

[0036] Step S423, based on the self-attention mechanism-image group, defect-oriented detection is performed to determine the defect detection set. Specifically, for each self-attention mechanism-image group, the self-attention mechanism is used to perform defect-oriented detection on the images in the image group. During the detection, the self-attention mechanism focuses on the areas and features related to the target physical defects in the image according to the parameters specified by the previous configuration and mapping. For example, in the self-attention mechanism-image group for the dark field image group and related to bubble detection, the model will search for those circular or elliptical low-grayscale areas in the image that match the bubble features. With the effect of the self-attention mechanism, these areas can be identified more accurately even if there are some noise or other interference factors in the image. When the self-attention mechanism detects areas in the image that meet the characteristics of the target physical defects, these areas are marked as potential defect areas, and relevant information such as the location, size, and shape of the defects are collected. For example, if a bubble is detected, its coordinate position, diameter, and circular shape characteristics in the image are recorded. This detection and information collection operation is performed on all images in the entire self-attention mechanism-image group, and then all marked and collected potential defect information is aggregated and integrated to form a defect detection set. This defect detection set covers the information of all physical defects detected in the effective image set with the help of the self-attention mechanism-image group, providing a key data source for subsequent further analysis of the quality and defects of optical glass.

[0037] Step S500, based on the optical detection coefficient and the defect detection list, a weighted calculation is performed to determine the performance detection result of the detection target. Specifically, the optical detection coefficient and the defect detection list are first analyzed. The optical detection coefficient is an indicator that reflects the quality of the polarization characteristics of optical glass. It is related to the quality level determined by matching the polarization angle quality level. The coefficient reflects the quality of polarization performance. The defect detection list contains a defect feature-defect coefficient sequence. The defect features include type, shape, size, position, etc. The defect coefficient is determined based on the severity of the defect and the degree of influence on the optical performance. The defect coefficient of large-area scratches is high, and the defect coefficient of small bubbles is low, which fully describes the physical defects of the optical glass. Then determine the weighting strategy, determine the relative importance of the optical detection coefficient and the defect detection list in the overall performance evaluation according to the application requirements and quality assessment standards of optical glass, and use expert experience, historical data statistical analysis or experimental verification methods to formulate a weighting scheme. Suppose the weight of the optical detection coefficient is , the weight of defect inspection order is and Then perform weighted calculation to obtain the optical detection coefficient value And the comprehensive value of all defect coefficients of the defect inspection list , according to the performance test results Calculation, such as , , , hour, Finally, the performance test results of the test target are determined according to Value interpretation and classification of optical glass performance, according to the preset performance level classification standards, such as For excellence, For good, To be qualified, In order to determine if the product is unqualified, the results and grades are compiled into a performance evaluation report, which explains the optical glass in two aspects and the comprehensive performance grade, providing a basis for quality control and usage decisions.

[0038] The embodiment of the present application determines the type of physical defects according to the application classification of optical glass, configures a multi-channel bright and dark field, and uses the incident light to perform a light path polarization test to obtain the optical detection coefficient. The sensor image set is collected through the coordinated operation of the light controller and the image sensor, and the defect detection and position stitching are completed in combination with the image processing module to generate a defect detection list. The comprehensive optical detection coefficient and the defect detection list are weighted and calculated, and the performance detection result of the detection target is finally obtained, achieving the technical effect of improving the recognition accuracy of various physical defects in optical glass detection.

[0039] In the above, refer to Figure 1 A method for detecting optical glass performance according to an embodiment of the present invention is described in detail. Figure 2 An optical glass performance detection system according to an embodiment of the present invention is described.

[0040] An optical glass performance detection system according to an embodiment of the present invention is used to solve the technical problem of insufficient recognition accuracy of various physical defects in existing optical glass performance detection, and achieve the technical effect of improving the recognition accuracy of various physical defects in optical glass detection. An optical glass performance detection system includes: a physical defect type determination module 10, an optical path polarization test module 20, a sensor image set acquisition module 30, a boundary segmentation module 40, and a weighted calculation module 50.

[0041] The physical defect type determination module 10 is used to determine the physical defect type of the target optical glass type according to the application classification of the optical glass, and configure the multi-channel bright and dark fields based on the defect contrast.

[0042] The optical path polarization test module 20 is used to introduce incident light, perform an optical path polarization test on the detection target, and determine the optical detection coefficient, wherein the polarization test includes a single chip entity test and an array simulation test.

[0043] The sensor image set acquisition module 30 is used to establish the cooperation between the light controller and the image sensor based on the multi-channel light and dark field, and the image sensor performs light control tracking to acquire the sensor image set.

[0044] The boundary segmentation module 40 is used to transmit back the sensor image set, and perform boundary segmentation in combination with the image processing module to determine a defect detection list based on self-attention detection and position splicing of the physical defect class, and the defect detection list includes a sequence based on defect feature-defect coefficient.

[0045] The weighted calculation module 50 is used to perform weighted calculation based on the optical detection coefficient and the defect detection list to determine the performance detection result of the detection target.

[0046] The specific configuration of the physical defect class determination module 10 will be described in detail below. As described above, for the application classification of optical glass, the physical defect class of the target optical glass type is determined, and based on the defect contrast, a multi-channel light and dark field is configured. The physical defect class determination module 10 further includes: a first clustering result determination unit, the first clustering result determination unit is used to traverse the physical defect class, perform a clustering based on the defect contrast, and determine a first clustering result; a second clustering result determination unit, the second clustering result determination unit is used to traverse the physical defect class, perform a secondary clustering based on the lighting type, and determine a second clustering result; a multi-channel light and dark field configuration unit, the multi-channel light and dark field configuration unit is used to fuse the first clustering result and the second clustering result, and configure the multi-channel light and dark field, wherein the multi-channel light and dark field is determined based on the lighting type and the field brightness.

[0047] The specific configuration of the optical path polarization test module 20 will be described in detail below. As described above, an incident light is introduced, and an optical path polarization test is performed on the detection target to determine the optical detection coefficient, wherein the polarization test includes a single-chip entity test and an array simulation test, and the optical path polarization test module 20 further includes: an optical path polarization threshold determination unit, the optical path polarization threshold determination unit is used to determine the optical path polarization threshold based on the optical qualification; a test optical path identification unit, the test optical path identification unit is used to identify the test optical path, calculate the polarization angle and determine whether the optical path polarization threshold is met; a defective identification generation unit, the defective identification generation unit is used to generate a defective identification if the polarization angle is greater than the optical path polarization threshold.

[0048] Among them, the optical path polarization test module 20 further includes: an assembly polarization angle determination unit, the assembly polarization angle determination unit is used to perform physical twinning and array optical path detection on the detection target in combination with a simulation platform if the polarization angle is less than the optical path polarization threshold, to determine the assembly polarization angle, wherein the parallel arrangement of multiple targets is a matrix distribution; a mean calculation unit, the mean calculation unit is used to perform mean calculation on the assembly polarization angle based on the number of arrays to determine the target polarization angle; an optical detection coefficient determination unit, the optical detection coefficient determination unit is used to perform quality level matching on the target polarization angle and determine the optical detection coefficient.

[0049] The specific configuration of the sensor image set acquisition module 30 will be described in detail below. As described above, based on the multi-channel light and dark field, the light controller and the image sensor are coordinated, the image sensor performs light control tracking, and the sensor image set is acquired. The sensor image set acquisition module 30 further includes: an azimuth trajectory determination unit, the azimuth trajectory determination unit is used to determine the azimuth trajectory of the image sensor for full-domain detection of the target for detection; multiple sets of light control parameter determination units, the multiple sets of light control parameter determination units are used to determine multiple sets of light control parameters based on the multi-channel light and dark field; a light controller control unit, the light controller control unit is used to control the light controller according to the multiple sets of light control parameters, and perform irradiation light regulation, the image sensor follows the detection target to collect images and determine the sensor image set.

[0050] The specific configuration of the boundary segmentation module 40 will be described in detail below. As described above, the sensor image set is transmitted back, and the boundary segmentation is performed in combination with the image processing module to determine the defect detection list based on the self-attention detection and position splicing of the physical defect class, and the defect detection list contains a sequence based on defect features-defect coefficients. The boundary segmentation module 40 further includes: an effective image set determination unit, which is used to perform boundary segmentation on the sensor image set and determine the effective image set, wherein the boundary between the detection target and the background is used as the segmentation basis; a self-attention detection unit, which is used to traverse the effective image set to determine the defect detection set based on the self-attention detection of the corresponding physical defect class; a distribution splicing unit, which is used to distribute and splice the defect detection set according to the relative spatial position of the detection target, and integrate to determine the defect detection list.

[0051] Among them, the valid image set is traversed to determine the defect detection set based on the self-attention detection of the corresponding physical defect class, and the self-attention detection unit further includes: a self-attention mechanism configuration subunit, the self-attention mechanism configuration subunit is used to configure the self-attention mechanism based on the physical defect class and the defect identification factor; a mapping subunit is established, the mapping subunit is used to establish a mapping with the self-attention mechanism according to the light and dark field sources of each group of images in the valid image set, and determine the self-attention mechanism-image group; a defect orientation detection subunit, the defect orientation detection subunit is used to perform defect orientation detection based on the self-attention mechanism-image group to determine the defect detection set.

[0052] An optical glass performance detection system provided by an embodiment of the present invention can execute an optical glass performance detection method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0053] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0054] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for detecting optical glass performance, characterized in that: The method comprises: According to the application classification of optical glass, determine the physical defect type of the target optical glass type, and configure multi-channel bright and dark fields based on the defect contrast; Introduce incident light, perform optical path polarization test on the detection target, and determine the optical detection coefficient, wherein the polarization test includes single chip entity test and array simulation test; Based on the multi-channel bright and dark fields, a light controller and an image sensor are coordinated, and the image sensor performs light control tracking to obtain a sensor image set; The sensor image set is transmitted back, and the boundary segmentation is performed in combination with the image processing module to determine the defect detection list based on the self-attention detection and position splicing of the physical defect class, and the defect detection list includes a sequence based on defect feature-defect coefficient; Based on the optical detection coefficient and the defect detection list, a weighted calculation is performed to determine the performance detection result of the detection target.

2. The optical glass performance testing method according to claim 1, characterized in that: The multi-channel bright and dark field configuration includes: Traversing the physical defect classes, performing a clustering based on defect contrast, and determining a first clustering result; Traversing the physical defect classes, performing secondary clustering based on the lighting type, and determining a second clustering result; The first clustering result and the second clustering result are integrated to configure the multi-channel light and dark field, wherein the multi-channel light and dark field is determined based on the lighting type and the field brightness.

3. The optical glass performance testing method according to claim 1, characterized in that: The determining of the optical detection coefficient comprises: Based on the optical qualification, a polarization threshold of the optical path is determined; Identify the test light path, calculate the polarization angle and determine whether the light path polarization threshold is met; If the polarization angle is greater than the optical path polarization threshold, a defective product mark is generated.

4. The optical glass performance testing method according to claim 3, characterized in that: The method further comprises: If the polarization angle is less than the optical path polarization threshold, the detection target is subjected to physical twinning and array optical path detection in combination with the simulation platform to determine the total polarization angle, wherein the parallel arrangement of multiple targets is a matrix distribution; Based on the number of arrays, the total polarization angle is averaged to determine the target polarization angle; The target polarization angle is quality-grade matched to determine an optical detection coefficient.

5. The optical glass performance testing method according to claim 1, characterized in that: The image sensor performs light control following to obtain a sensor image set, including: For the detection target, determine the azimuth trajectory of the image sensor for full-area detection of the target; Based on the multi-channel bright and dark fields, determining multiple groups of light control parameters; According to the multiple groups of light control parameters, the light controller is controlled to adjust the irradiated light, and the image sensor follows to collect images of the detection target to determine the sensor image set.

6. The optical glass performance testing method according to claim 1, characterized in that: The step of determining the defect detection list includes: Perform boundary segmentation on the sensor image set to determine a valid image set, wherein the boundary between the detected target and the background is used as a basis for segmentation; traversing the valid image set to determine a defect detection set based on self-attention detection of corresponding physical defect classes; According to the relative spatial position of the detection target, the defect detection set distribution is spliced ​​and integrated to determine the defect detection list.

7. The optical glass performance testing method according to claim 6, characterized in that: The step of determining a defect detection set based on self-attention detection of corresponding physical defect classes includes: Based on the physical defect class, configure the self-attention mechanism according to the defect identification factors; According to the light and dark field sources of each group of images in the valid image set, a mapping is established with the self-attention mechanism to determine the self-attention mechanism-image group; Based on the self-attention mechanism-image group, defect-oriented detection is performed to determine the defect detection set.

8. An optical glass performance detection system, characterized in that: The system is used to implement an optical glass performance detection method according to any one of claims 1 to 7, and the system comprises: A physical defect type determination module, which is used to determine the physical defect type of the target optical glass type according to the application classification of the optical glass, and configure a multi-channel bright and dark field based on the defect contrast; An optical path polarization test module, which is used to introduce incident light, perform an optical path polarization test on a detection target, and determine an optical detection coefficient, wherein the polarization test includes a single chip entity test and an array simulation test; A sensor image set acquisition module, the sensor image set acquisition module is used to establish the cooperation between the light controller and the image sensor based on the multi-channel light and dark field, the image sensor performs light control follow-up, and acquires the sensor image set; A boundary segmentation module, which is used to transmit the sensor image set back and perform boundary segmentation in combination with the image processing module, so as to determine a defect detection list based on self-attention detection and position splicing of physical defect classes, and the defect detection list includes a sequence based on defect features-defect coefficients; A weighted calculation module is used to perform weighted calculation based on the optical detection coefficient and the defect detection list to determine the performance detection result of the detection target.

Images