A mobile phone microcrystalline glass cover plate surface scratch detection method

Abstract

The application relates to the technical field of cover plate scratch detection, in particular to a mobile phone microcrystalline glass cover plate surface scratch detection method, which comprises the following steps: calculating the channel brightness ratio value of image data based on reflection brightness and scattering brightness, merging the channel brightness ratio value of the image data, a first characteristic index and a second characteristic index into a layer feature vector; constructing a coating characteristic difference judgment model, inputting the layer feature vector into the coating characteristic difference judgment model, and outputting a coating difference index according to the judgment of the coating characteristic difference judgment model; and triggering adaptive adjustment of the lighting condition according to the coating difference index. The application introduces the first characteristic index and the second characteristic index in the detection process, combines the coating difference index output by the coating characteristic difference judgment model, drives the adaptive adjustment of the lighting condition according to the coating difference index, makes the detection consistent with the condition with the change of the coating characteristics, and thus improves the stability and accuracy of the scratch detection.

Description

Technical Field

[0001] This application relates to the field of cover glass scratch detection technology, specifically to a method for detecting scratches on the surface of a mobile phone microcrystalline glass cover. Background Technology

[0002] Microcrystalline glass covers are an integral part of the smartphone's exterior components. They possess high hardness and impact resistance. A key feature of microcrystalline glass covers is the presence of an anti-fingerprint coating (AF) and an anti-reflective coating (AR). These coatings reduce fingerprint residue and glare. Currently, scratch detection for microcrystalline glass covers largely relies on machine vision-based optical imaging. This involves using dark-field or polarized light illumination to enhance the contrast between scratches and the background, and then employing image processing algorithms to identify and classify defects.

[0003] However, the following two problems still urgently need to be solved in the existing technology:

[0004] First, because the reflection and scattering characteristics of the coating on the surface of the microcrystalline glass cover plate vary with different batches, different machines and different lighting conditions, the visibility of the same scratch may fluctuate to varying degrees under different conditions, thus affecting the stability and accuracy of the detection.

[0005] Secondly, when switching batches or machines, existing testing methods lack the ability to adapt to the optical properties of coatings and cannot automatically adjust the light source intensity and illumination angle in a short time, which affects the production speed.

[0006] Therefore, a method for detecting scratches on the surface of mobile phone microcrystalline glass covers is proposed in order to solve the above problems. Summary of the Invention

[0007] This application provides a method for detecting scratches on the surface of a mobile phone microcrystalline glass cover, aiming to solve the problem that the visibility of the same scratch may fluctuate to varying degrees under different conditions in related technologies, and to avoid the problem that existing detection methods lack the ability to adapt to the optical properties of the coating.

[0008] To achieve the above objectives, this application provides a method for detecting scratches on the surface of a mobile phone microcrystalline glass cover, comprising the following steps:

[0009] Acquire image data and image condition information of the microcrystalline glass cover to be tested.

[0010] Extract the reflection and scattering brightness of non-defective areas from the image data, and extract the first characteristic index and the second characteristic index.

[0011] The channel luminance ratio of the image data is calculated based on the reflected luminance and the scattered luminance. The channel luminance ratio, the first characteristic index, and the second characteristic index of the image data are then combined into a layer feature vector.

[0012] A coating characteristic difference determination model is constructed. The layer feature vector is input into the coating characteristic difference determination model, and the coating characteristic difference determination model outputs the coating difference index based on the determination.

[0013] The system triggers adaptive adjustment of lighting conditions based on the coating difference index, generating and executing adaptive adjustment commands including light source intensity, lighting angle, and polarizer angle.

[0014] Based on the adaptively adjusted lighting conditions, image data of the microcrystalline glass cover to be tested is reacquired, and scratch detection is performed on the reacquired image data of the microcrystalline glass cover.

[0015] Preferably, acquiring image data and image condition information of the microcrystalline glass cover plate specifically includes:

[0016] The image data includes dark field illumination images, side illumination images, and polarized illumination images. The image data is used to represent the surface reflection and scattering characteristics of the microcrystalline glass cover under different illumination modes. The image condition information includes light source intensity parameters, illumination angle parameters, and polarizer angle parameters. The image condition information is used to represent the illumination state and optical configuration when acquiring the image.

[0017] Preferably, the reflection and scattering brightness of non-defective areas in the image data are extracted, specifically including:

[0018] Before extracting the reflection and scattering brightness of non-defective regions in the image data, normalization processing is performed on the image data, and region determination is performed on the normalized image data. The region determination is a process of identifying defective and non-defective regions in the normalized image, specifically as follows:

[0019] Pixel-level filtering and connected component analysis are performed on the normalized image according to preset brightness change thresholds, gradient continuity judgment criteria, and preset connected area thresholds:

[0020] When the grayscale difference between adjacent pixels exceeds the brightness change threshold and meets the gradient continuity judgment criterion in the main direction, the current pixel is added to the candidate defect set. Connectivity analysis is performed on the candidate defect set. Connectivity components with a candidate defect set area smaller than the preset connectivity area threshold are removed, while the remaining connectivity components are marked as defect regions. The remaining regions after removing the defect regions and boundary regions are marked as non-defect regions.

[0021] In the non-defect areas of the lateral illumination image and the polarized illumination image, the normalized gray mean is calculated separately, and the reflected brightness is obtained by summing according to the preset weight. When there is only a single lateral illumination angle, the normalized gray mean of the current angle is used as the reflected brightness. When there are multiple lateral illumination angles, the reflected brightness is calculated by weighting according to the brightness angle parameter, and the weighted result is used as the reflected brightness.

[0022] High-pass filtering is performed in the non-defective region of the dark-field illumination image, and the normalized mean of the high-frequency grayscale components is calculated. The normalized mean is used as the scattered brightness.

[0023] Preferably, the first characteristic index and the second characteristic index are extracted, specifically including:

[0024] The first characteristic index is the polarization retention index of the AF coating, which is used to represent the polarization retention capability of the AF coating. The first characteristic index is obtained by acquiring the co-polarized image and the cross-polarized image in the polarized illumination image, performing region determination on the co-polarized image and the cross-polarized image, and calculating the normalized grayscale mean of the co-polarized and cross-polarized images within the non-defect regions of the co-polarized and cross-polarized images respectively. The ratio of the difference to the sum of the normalized grayscale mean values ​​of the co-polarized and cross-polarized images is used as the first characteristic index. The calculation formula is as follows:

[0025]

[0026] In the formula, Indicates the first characteristic indicator. Representing an image Normalized grayscale mean in the non-defect region This represents the normalized grayscale mean of the co-polarized image within the non-defect region. Represents a co-polarized image. This represents the normalized grayscale mean of the cross-polarized image within the non-defect region. Represents an interpolarized image; To avoid extremely small positive constants with a denominator of zero.

[0027] The second characteristic index is the AR coating multi-angle reflection residual index, used to represent the multi-angle reflection residual of the AR coating; the second characteristic index is calculated by performing region determination on the side-illuminated image, statistically analyzing the normalized grayscale mean corresponding to each illumination angle within the non-defect area of ​​the side-illuminated image, and using the absolute value of the difference between the grayscale mean values ​​of different illumination angles as the second characteristic index. The calculation formula is as follows:

[0028]

[0029] In the formula, The second characteristic indicator; max represents the maximum value; Represents absolute value; Indicates the number of lateral lighting angles Indicates the first One side lighting angle; Indicates the lighting angle as Lateral illumination image; This represents the average grayscale value within the non-defect area. Indicates the lighting angle as The normalized gray mean of the lateral illumination image in the non-defect area.

[0030] Preferably, the channel luminance ratio of the image data is calculated based on the reflected luminance and the scattered luminance, and the channel luminance ratio, the first characteristic index, and the second characteristic index of the image data are combined into a layer feature vector, specifically including:

[0031] The formula for calculating the channel luminance ratio is:

[0032]

[0033] In the formula, This represents the channel luminance ratio of image data; Indicates the reflectance brightness of image data; Indicates the scattering brightness of image data; To avoid extremely small positive constants with a denominator of zero.

[0034] After performing dimensional consistency processing on the channel brightness ratio, the first characteristic index, and the second characteristic index of the image data, they are merged into a layer feature vector in the order of the first characteristic index, the second characteristic index, and the channel brightness ratio.

[0035] Preferably, a coating characteristic difference determination model is constructed. The layer feature vector is input into the coating characteristic difference determination model, and the coating characteristic difference determination model outputs a coating difference index based on the determination, specifically including:

[0036] The coating characteristic difference judgment model refers to a statistical judgment model based on a reference library, which is used to measure the degree of difference between the layer feature vector of the current batch and the reference batch. By collecting multiple batches of microcrystalline glass cover plate samples that have been judged to be qualified, the layer feature vectors are obtained under stable lighting and consistent algorithm caliber, and a reference library is established according to coating type and thickness specifications.

[0037] The layer feature vector is input into the coating characteristic difference determination model. The model first performs integrity verification on the input layer feature vector; then, it performs feature standardization on the integrity-verified layer feature vector; finally, it selects corresponding reference parameters from the reference library based on the coating type and thickness specifications in the layer feature vector, and performs difference calculation based on the selected reference parameters to obtain the difference calculation result; the coating characteristic difference determination model converts the difference calculation result into a coating difference index through a mapping function, and outputs the coating difference index according to a set threshold range.

[0038] Preferably, the illumination conditions are adaptively adjusted based on the coating difference index, generating and executing adaptive adjustment commands including light source intensity, illumination angle, and polarizer angle, specifically including:

[0039] When the coating difference index falls within different judgment ranges, different levels of adaptive adjustment strategies are triggered.

[0040] When the coating difference index is below the first threshold, it is determined that the current lighting conditions are in good agreement with the reference library, and no adjustment is made.

[0041] When the coating difference index is between the first threshold and the second threshold, it is determined that there is a slight deviation, triggering a fine-tuning of local parameters, including increasing or decreasing the light source intensity within a preset range, and simultaneously fine-tuning the polarizer angle.

[0042] When the coating difference index is higher than the second threshold, it is determined that there is a significant deviation, triggering a comprehensive adjustment strategy. Based on the fine-tuning, the lighting angle setting is further adjusted so that the lighting incident direction covers the entire angle range.

[0043] The relationship between the adaptive adjustment and the coating difference index is as follows: the higher the coating difference index, the greater the difference between the optical properties of the current batch of coatings and the reference batch, and the amplitude and range of the adaptive adjustment will increase accordingly.

[0044] Preferably, based on the adaptively adjusted lighting conditions, image data of the microcrystalline glass cover to be tested is re-acquired, and scratch detection is performed on the re-acquired image data of the microcrystalline glass cover, specifically including:

[0045] Under adaptively adjusted lighting conditions, the image data of the re-acquired microcrystalline glass cover plate is used to separate defective and non-defective areas according to the region determination rules.

[0046] Within the defect area, the brightness variation of the image data within the defect area is examined. In areas where the image data exhibits brightness variations, edge continuity and directional features are identified. For the identified scratch areas, the length, width, direction, and brightness contrast parameters of the scratch are further extracted. Based on the length, width, direction, and brightness contrast parameters, the scratch location distribution and scratch geometric feature parameters are determined. The scratch detection results are output based on the scratch location distribution and scratch geometric feature parameters.

[0047] Preferably, a method for detecting scratches on the surface of a mobile phone microcrystalline glass cover further includes: calculating the perceived risk level of the scratches in the image data of the microcrystalline glass cover under the user's expected observation conditions based on the visual feature parameters of the scratches in the image data and a preset visibility evaluation standard; combining a preset rejection threshold with the perceived risk level result; and judging and outputting the current evaluation result of the microcrystalline glass cover. Specifically, this includes:

[0048] The visual feature parameters include the brightness contrast, length, width, and direction of the scratch in the detected image; the brightness contrast is compared with the normalized grayscale difference of the background area, and the length and width parameters are combined to map the visibility probability of the scratch under the user's expected observation conditions.

[0049] Based on the preset visibility assessment criteria, the perceived risk classification results are obtained, which include the first level interval, the second level interval, and the third level interval.

[0050] The perceived risk classification results are combined with the rejection threshold. When the scratch length or width exceeds the rejection threshold and the perceived risk classification exceeds the third level, the rejection result is directly output. When the scratch size is lower than the rejection threshold but the perceived risk classification is within the second level range, a re-inspection suggestion is output. When the scratch size is lower than the rejection threshold and the perceived risk classification is within the first level range, a qualified result is output.

[0051] The final evaluation results include scratch detection results, the probability of scratches being visible under the user's expected observation conditions, perceived risk classification results, and the determination of the rejection threshold.

[0052] The beneficial effects of this application are as follows:

[0053] 1. This application extracts the reflection and scattering brightness of non-defective areas from image data, introduces a first characteristic index and a second characteristic index, merges the four data to form a layer feature vector, inputs it into a coating characteristic difference judgment model for analysis, quantifies the coating differences of different batches, different machines and different lighting conditions; and compensates and corrects the differences in coating optical properties before scratch detection, so that the visibility of the same scratch remains stable under different conditions, thereby improving the consistency and accuracy of the detection results.

[0054] 2. This application triggers adaptive adjustment of lighting conditions through coating difference index, automatically generates and executes adjustment instructions for light source intensity, lighting angle and polarizer angle according to the degree of difference, realizes the adaptation of the inspection station when switching batches or machines; saves the time of manual repeated adjustment of light source and lighting conditions, ensures the continuity of the inspection process, thereby improving production efficiency and avoiding the impact of inspection delay on production speed.

[0055] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0056] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a flowchart illustrating a method for detecting surface scratches on a mobile phone microcrystalline glass cover, provided in an embodiment of this application. Detailed Implementation

[0058] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0059] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for detecting surface scratches on a mobile phone microcrystalline glass cover, provided in an embodiment of this application.

[0060] In this embodiment, a method for detecting scratches on the surface of a mobile phone microcrystalline glass cover includes steps S10, S20, S30, S40, S50, and S60.

[0061] Step S10: Obtain image data and image condition information of the microcrystalline glass cover to be tested, specifically including:

[0062] The image data includes dark field illumination images, side illumination images, and polarized illumination images. The image data is used to represent the surface reflection and scattering characteristics of the microcrystalline glass cover under different illumination modes. The image condition information includes light source intensity parameters, illumination angle parameters, and polarizer angle parameters. The image condition information is used to represent the illumination state and optical configuration when the image is acquired.

[0063] It should be noted that the image data is acquired through a multi-channel optical imaging device installed at the inspection station. This device includes a dark-field illumination unit, a side-illumination unit, and a polarized illumination unit, with a CCD camera positioned at each location as the image acquisition device. By switching or combining different illumination units, the same inspection station can acquire corresponding image data under dark-field, side-illumination, and polarized illumination conditions. This multi-channel optical imaging device is a common configuration in the field, specifically employing multiple ring-shaped LED light sources combined with polarizers to form an illumination module, which is then used in conjunction with an industrial camera for synchronous acquisition.

[0064] It should be noted that the image condition information is acquired in real time through the light source control component and polarization angle control component at the detection station. Specifically, the light source control component and polarization angle control component are programmable power supply modules used to adjust and record the output intensity and illumination angle of the light source; the polarization angle control component is a motorized rotating polarizer mechanism used to change the installation angle of the polarizer and output the current polarizer angle information in real time. Both the light source control component and the polarization angle control component are existing, publicly available standardized modules that can be directly obtained and selected by those skilled in the art.

[0065] Step S20: Extract the reflection and scattering brightness of non-defective areas from the image data, and extract the first characteristic index and the second characteristic index, specifically including:

[0066] Before extracting the reflection and scattering brightness of non-defect areas in the image data, normalization processing is performed on the image data, and region determination is performed on the normalized image data. Region determination is a process of identifying defective and non-defective regions in the normalized image, specifically as follows:

[0067] Pixel-level filtering and connected component analysis are performed on the normalized image according to the preset brightness change threshold, gradient continuity judgment criteria, and preset connected area threshold.

[0068] It should be noted that the preset brightness change threshold is set by statistically analyzing the brightness distribution of qualified samples, and the value is between 5 and 15 pixels grayscale difference.

[0069] It should be noted that the criterion for judging gradient continuity is: when the gradient change along the main direction of the image remains monotonic within a range of more than 3 consecutive pixels, the current region is judged to satisfy gradient continuity.

[0070] It should be noted that the preset connected area threshold is set by estimating the minimum distribution of the actual scratch width and length, and the value ranges from 20 to 50 pixels of connected area.

[0071] When the grayscale difference between adjacent pixels exceeds the brightness change threshold and meets the gradient continuity judgment criterion in the main direction, the current pixel is added to the candidate defect set. Connectivity analysis is performed on the candidate defect set. Connectivity components with a candidate defect set area smaller than the preset connectivity area threshold are removed, while the remaining connectivity components are marked as defect regions. The remaining regions after removing the defect regions and boundary regions are marked as non-defect regions.

[0072] It should be noted that the normalization process first performs layer optical characteristic analysis on the image data and image condition information. Based on the dark field illumination image, the side illumination image, and the polarized illumination image, and then combined with the light source intensity parameter, illumination angle parameter, and polarizer angle parameter, the image data is initially corrected. Then, the image data after the initial correction is weighted and corrected according to the light source intensity parameter, illumination angle parameter, and polarizer angle parameter, and the normalized image is obtained based on the weighted correction.

[0073] In the non-defect areas of the lateral illumination image and the polarized illumination image, the normalized gray mean is calculated separately, and the reflected brightness is obtained by summing according to the preset weight. When there is only a single lateral illumination angle, the normalized gray mean of the current angle is used as the reflected brightness. When there are multiple lateral illumination angles, the reflected brightness is calculated by weighting according to the brightness angle parameter, and the weighted result is used as the reflected brightness.

[0074] It should be noted that the normalized grayscale mean refers to the average value of all pixel grayscale values ​​in the non-defect area after normalization.

[0075] It should be noted that the preset weights range from 0.2 to 0.5, and the sum of the weights is 1.

[0076] It should be noted that the weighted calculation is as follows: the normalized grayscale mean under different lighting angles is multiplied by the corresponding angle weight coefficient, and the sum is used as the final reflectance brightness.

[0077] High-pass filtering is performed in the non-defective region of the dark-field illumination image, and the normalized mean of the high-frequency grayscale components is calculated. The normalized mean is used as the scattered brightness.

[0078] It should be noted that the normalized mean of the high-frequency grayscale component is obtained by performing Fourier high-pass filtering on the dark-field illumination image, retaining the spatial frequency portion above the preset threshold, and statistically averaging it in the non-defect area; the preset threshold of spatial frequency is set by statistical analysis of the spectral distribution of qualified samples and samples with minor scratches, specifically by selecting the boundary point in the two-dimensional spectrum that can effectively distinguish the high-frequency features of scratches from the background noise frequency; the preset threshold of spatial frequency is set in the range of 10% to 20% of the maximum resolvable frequency of the image. When the image resolution is 2048×2048 pixels, the corresponding spatial frequency threshold is set to 200 to 400 periods / image width.

[0079] The first characteristic index is the polarization retention index of the AF coating, which represents the polarization retention capability of the AF coating. This index is determined by acquiring co-polarized and cross-polarized images from the polarized illumination image, performing region determination on both images, and statistically calculating the normalized grayscale mean of the co-polarized and cross-polarized images within the non-defect regions of both images. The ratio of the difference to the sum of the normalized grayscale mean values ​​of the co-polarized and cross-polarized images is used as the first characteristic index. The calculation formula is as follows:

[0080]

[0081] In the formula, Indicates the first characteristic indicator. Representing an image Normalized grayscale mean in the non-defect region This represents the normalized grayscale mean of the co-polarized image within the non-defect region. Represents a co-polarized image. This represents the normalized grayscale mean of the cross-polarized image within the non-defect region. Represents an interpolarized image; To avoid extremely small positive constants with a denominator of zero, the range of values ​​is [range missing]. .

[0082] The second characteristic index is the AR coating multi-angle reflection residual index, which is used to represent the multi-angle reflection residual of the AR coating. The second characteristic index is determined by performing region determination on the side-illuminated image, statistically analyzing the normalized grayscale mean corresponding to each illumination angle within the non-defective region of the side-illuminated image, and using the absolute value of the difference between the grayscale mean values ​​of different illumination angles as the second characteristic index. The calculation formula is as follows:

[0083]

[0084] In the formula, The second characteristic indicator; max represents the maximum value; Represents absolute value; Indicates the number of lateral lighting angles Indicates the first One side lighting angle; Indicates the lighting angle as Lateral illumination image; This represents the average grayscale value within the non-defect area. Indicates the lighting angle as The normalized gray mean of the lateral illumination image in the non-defect area.

[0085] Step S30: Calculate the channel brightness ratio of the image data based on reflected brightness and scattered brightness, and merge the channel brightness ratio, the first characteristic index, and the second characteristic index of the image data into a layer feature vector, specifically including:

[0086] The formula for calculating the channel luminance ratio is:

[0087]

[0088] In the formula, This represents the channel luminance ratio of image data; Indicates the reflectance brightness of image data; Indicates the scattering brightness of image data; To avoid extremely small positive constants with a denominator of zero, the range of values ​​is [range missing]. .

[0089] After performing dimensional consistency processing on the channel brightness ratio, the first characteristic index, and the second characteristic index of the image data, they are merged into a layer feature vector in the order of the first characteristic index, the second characteristic index, and the channel brightness ratio.

[0090] It should be noted that the specific process for dimensional consistency processing is as follows: First, the channel luminance ratio, the first characteristic index, and the second characteristic index are respectively dimensionless, and the numerical ranges of the channel luminance ratio, the first characteristic index, and the second characteristic index are uniformly mapped to... The range; where the channel brightness ratio is processed using the maximum-minimum normalization method, and the first characteristic index and the second characteristic index are standardized according to the statistical mean and standard deviation in the reference library.

[0091] Step S40: Construct a coating characteristic difference determination model. Input the layer feature vector into the coating characteristic difference determination model. The coating characteristic difference determination model outputs a coating difference index based on the determination, specifically including:

[0092] The coating characteristic difference judgment model is a statistical judgment model based on a reference library, used to measure the degree of difference between the layer feature vector of the current batch and the reference batch. By collecting multiple batches of microcrystalline glass cover plate samples that have been judged to be qualified, the layer feature vectors are obtained under stable lighting and consistent algorithm caliber, and a reference library is established according to coating type and thickness specifications.

[0093] The layer feature vector is input into the coating characteristic difference determination model. The model first performs integrity verification on the input layer feature vector; then, it performs feature standardization on the integrity-verified layer feature vector; finally, it selects corresponding reference parameters from the reference library based on the coating type and thickness specifications in the layer feature vector, and performs difference calculation based on the selected reference parameters to obtain the difference calculation result; the coating characteristic difference determination model converts the difference calculation result into a coating difference index through a mapping function, and outputs the coating difference index according to a set threshold range.

[0094] It should be noted that the integrity check specifically involves checking whether the input layer feature vector contains three types of data: channel brightness ratio, first characteristic index, and second characteristic index. If any of the three data items is missing or has an abnormal value, the abnormal sample will be removed or a re-collection will be triggered.

[0095] It should be noted that the standardization process specifically involves applying zero-mean and unit variance to the input features based on the statistical mean and standard deviation of the corresponding parameters in the reference library, in order to ensure that different feature quantities are comparable under a uniform scale.

[0096] It should be noted that the selection of the corresponding reference parameters is achieved by reading the coating type and thickness specification information marked in the layer feature vector, and calling the average feature vector and variance of the same type and thickness specification from the reference library as a comparison benchmark.

[0097] It should be noted that the formula for calculating the difference is based on the sum of squares of the Euclidean distance, which will not be elaborated here.

[0098] It should be noted that the mapping function converts the difference calculation result into a coating difference index through a monotonically increasing logarithmic mapping function, ensuring that the index changes slowly when the difference is small and is significantly amplified when the difference is large.

[0099] It should be noted that the threshold range includes a first threshold and a second threshold. The first threshold is set by statistically analyzing the difference distribution of qualified samples in the reference library and taking the mean difference of qualified samples plus 1 to 2 times the standard deviation. The second threshold is set by selecting the upper limit value covered by more than 95% of qualified samples in the difference distribution and correcting it based on the significant fluctuations in the visibility of scratch detection in the test.

[0100] Step S50: Trigger adaptive adjustment of lighting conditions based on coating difference index, generate and execute adaptive adjustment commands including light source intensity, illumination angle, and polarizer angle, specifically including:

[0101] When the coating difference index falls within different judgment ranges, different levels of adaptive adjustment strategies are triggered.

[0102] When the coating difference index is below the first threshold, it is determined that the current lighting conditions are in good agreement with the reference library, and no adjustment is made.

[0103] When the coating difference index is between the first threshold and the second threshold, it is determined that there is a slight deviation, triggering a fine-tuning of local parameters, including increasing or decreasing the light source intensity within a preset range, and simultaneously fine-tuning the polarizer angle.

[0104] It should be noted that fine-tuning of local parameters specifically refers to making small adjustments to only two parameters, the light source intensity and the polarizer angle, while keeping the illumination angle unchanged, thereby correcting the optical response shift caused by coating differences.

[0105] It should be noted that adjusting the light source intensity within the preset range means adjusting the light source output linearly or in stages within the range of ±5% to ±15%, based on the degree of deviation of the coating difference index, under the condition that the light source's rated power and minimum effective illuminance do not exceed the range, thereby maintaining the stability of reflected and scattered brightness.

[0106] It should be noted that fine-tuning the polarizer angle refers to making a slight rotation within the range of ±2° to ±5° around the current angle, based on the reference set angle, in order to correct the differences in the polarization retention characteristics of the AF coating and make the grayscale ratio of the co-polarized and cross-polarized channels tend to be stable.

[0107] When the coating difference index is higher than the second threshold, it is determined that there is a significant deviation, triggering a comprehensive adjustment strategy. Based on the fine-tuning, the lighting angle setting is further adjusted so that the lighting incident direction covers the entire angle range.

[0108] It should be noted that further adjusting the illumination angle setting means, while maintaining the light source intensity and fine-tuning the polarizer angle, expanding the illumination angle from a single or partial angle to cover the full-angle incident range from 0° to 60°, and ensuring that the multi-angle reflection characteristics of the AR coating are fully recorded and corrected by using multi-angle acquisition methods, thereby reducing the deviation of the detection results caused by changes in the incident angle.

[0109] The relationship between adaptive adjustment and coating difference index is as follows: the higher the coating difference index, the greater the difference between the optical properties of the current batch of coating and the reference batch, and the magnitude and range of adaptive adjustment will increase accordingly.

[0110] Step S60: Based on the adaptively adjusted lighting conditions, re-acquire image data of the microcrystalline glass cover to be tested, and perform scratch detection on the re-acquired image data of the microcrystalline glass cover, specifically including:

[0111] Under adaptively adjusted lighting conditions, the image data of the re-acquired microcrystalline glass cover plate is used to separate the defective and non-defective areas according to the region determination rules.

[0112] Within the defect area, the brightness variation of the image data within the defect area is examined. In areas where the image data exhibits brightness variations, edge continuity and directional features are identified. For the identified scratch areas, the length, width, direction, and brightness contrast parameters of the scratch are further extracted. Based on the length, width, direction, and brightness contrast parameters, the scratch location distribution and scratch geometric feature parameters are determined. The scratch detection results are output based on the scratch location distribution and scratch geometric feature parameters.

[0113] It should be noted that inspecting the brightness variation within the defect area of ​​the image data refers to performing local statistical analysis and gradient calculation on the pixel grayscale values ​​within the defect area. When the local grayscale difference exceeds a preset brightness fluctuation threshold, it is determined that a potential scratch exists at the current location. The brightness fluctuation threshold is a grayscale difference threshold used to distinguish between normal textures and potential scratches. It is obtained by calculating the statistical distribution of grayscale differences within the non-defective area of ​​a qualified sample and adding 1 to 2 times the standard deviation to the mean of the calculated grayscale differences.

[0114] It should be noted that identifying edge continuity and directionality features within areas of image data with brightness variations means applying an edge detection operator to extract edge pixels within a suspected scratch area, and then determining whether the edge remains continuous and oriented within a certain length range by fitting a principal direction vector. When the continuity and directionality conditions are met, the current area is confirmed as a scratch area.

[0115] It should be noted that determining the location distribution and geometric feature parameters of scratches refers to extracting the start and end points of the scratches according to pixel coordinates within the area that has been identified as a scratch, calculating the length and direction of the scratches; calculating the scratch width through the grayscale profile of the scratch cross-section; calculating the brightness contrast through the grayscale difference between the scratch area and the surrounding non-defect area; and finally forming a set of geometric feature parameters including the scratch location coordinates, length, width, direction, and brightness contrast.

[0116] After performing the above steps, a method for detecting surface scratches on a mobile phone microcrystalline glass cover further includes: calculating the perceived risk level of the scratches in the image data of the microcrystalline glass cover under the user's expected observation conditions based on the visual feature parameters of the scratches in the image data and a preset visibility evaluation standard; combining a preset rejection threshold with the perceived risk level result; and judging and outputting the current evaluation result of the microcrystalline glass cover. Specifically, this includes:

[0117] Visual feature parameters include the brightness contrast, length, width, and direction of the scratch in the detection image; the brightness contrast is compared with the normalized grayscale difference of the background area, and combined with the length and width parameters, it is mapped to the probability of the scratch being visible under the user's expected observation conditions.

[0118] Based on the preset visibility assessment criteria, the perceived risk classification results are obtained, which include the first level interval, the second level interval, and the third level interval.

[0119] The perceived risk classification results are combined with the rejection threshold. When the scratch length or width exceeds the rejection threshold and the perceived risk classification exceeds the third level, the rejection result is directly output. When the scratch size is lower than the rejection threshold but the perceived risk classification is within the second level range, a re-inspection suggestion is output. When the scratch size is lower than the rejection threshold and the perceived risk classification is within the first level range, a qualified result is output.

[0120] The final evaluation results include scratch detection results, the probability of scratches being visible under the user's expected observation conditions, perceived risk classification results, and the determination of the rejection threshold.

[0121] It should be noted that the perceived risk classification results are obtained by comparing the visual feature parameters of the scratch, such as brightness contrast, length and width, with the visibility criteria, and then dividing the visibility probability into corresponding level intervals according to the mapping rules.

[0122] It should be noted that the first, second, and third level intervals were set by statistical analysis of the observation results under different combinations of scratch size and brightness contrast. Specifically, the first level interval corresponds to the case where the visibility probability is less than 30%, the second level interval corresponds to the case where the visibility probability is between 30% and 70%, and the third level interval corresponds to the case where the visibility probability is greater than 70%.

[0123] It should be noted that the rejection threshold is set through production processes and customer acceptance standards, and is determined by a combination of historical quality inspection data and user experience feedback. The specific range of the rejection threshold is by default that the scratch length is greater than 2mm or the width is greater than 50μm.

[0124] Furthermore, it should be noted that while microcrystalline glass covers exhibit good visual appeal and durability after applying AF anti-fingerprint coating and AR anti-reflective coating, their surface optical properties fluctuate due to variations in coating type, batch differences, and processing techniques, resulting in inconsistencies in reflectivity and transmittance parameters. Simultaneously, in large-scale production, it is difficult to maintain consistent testing conditions across different machines and batches, easily leading to deviations in test results. Therefore, this application introduces a first characteristic index and a second characteristic index into the testing process, combining them with a coating characteristic difference judgment model to output a coating difference index. This index drives adaptive adjustments to lighting conditions, ensuring consistent testing conditions regardless of coating characteristics, thereby improving the stability and accuracy of scratch detection.

[0125] This completes a method for detecting scratches on the surface of a mobile phone microcrystalline glass cover.

[0126] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0127] Those skilled in the art will recognize that the algorithms or steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0128] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for detecting scratches on the surface of a mobile phone microcrystalline glass cover, characterized in that, Includes the following steps: Acquire image data and image condition information of the microcrystalline glass cover plate to be tested; Extract the reflectance and scattering brightness of non-defective regions from the image data, and extract the first characteristic index and the second characteristic index, specifically including: The first characteristic index is the polarization retention index of the AF coating, which is used to represent the polarization retention capability of the AF coating. The first characteristic index is obtained by acquiring the co-polarized image and the cross-polarized image in the polarized illumination image, performing region determination on the co-polarized image and the cross-polarized image, and calculating the normalized grayscale mean of the co-polarized and cross-polarized images within the non-defect regions of the co-polarized and cross-polarized images respectively. The ratio of the difference to the sum of the normalized grayscale mean values ​​of the co-polarized and cross-polarized images is used as the first characteristic index. The calculation formula is as follows: ； In the formula, Indicates the first characteristic indicator. Representing an image Normalized grayscale mean in the non-defect region This represents the normalized grayscale mean of the co-polarized image within the non-defect region. Represents a co-polarized image. This represents the normalized grayscale mean of the cross-polarized image within the non-defect region. Represents an interpolarized image; To avoid extremely small positive constants with a denominator of zero; The second characteristic index is the AR coating multi-angle reflection residual index, used to represent the multi-angle reflection residual of the AR coating; the second characteristic index is calculated by performing region determination on the side-illuminated image, statistically analyzing the normalized grayscale mean corresponding to each illumination angle within the non-defect area of ​​the side-illuminated image, and using the absolute value of the difference between the grayscale mean values ​​of different illumination angles as the second characteristic index. The calculation formula is as follows: ； In the formula, The second characteristic indicator; max represents the maximum value; Represents absolute value; Indicates the number of lateral lighting angles Indicates the first One side lighting angle; Indicates the lighting angle as Lateral illumination image; This represents the average gray value within the non-defect area. Indicates the lighting angle as The normalized gray mean of the lateral illumination image in the non-defect area; The channel brightness ratio of image data is calculated based on reflected brightness and scattered brightness, and the channel brightness ratio, first characteristic index and second characteristic index of image data are combined into a layer feature vector. Construct a coating characteristic difference determination model, input the layer feature vector into the coating characteristic difference determination model, and the coating characteristic difference determination model outputs the coating difference index based on the determination; The system triggers adaptive adjustment of lighting conditions based on the coating difference index, generating and executing adaptive adjustment commands including light source intensity, lighting angle, and polarizer angle. Based on the adaptively adjusted lighting conditions, image data of the microcrystalline glass cover to be tested is reacquired, and scratch detection is performed on the reacquired image data of the microcrystalline glass cover.

2. The method for detecting surface scratches on a mobile phone microcrystalline glass cover as described in claim 1, characterized in that, Acquiring image data and image condition information of the microcrystalline glass cover plate, specifically including: The image data includes dark field illumination images, side illumination images, and polarized illumination images. The image data is used to represent the surface reflection and scattering characteristics of the microcrystalline glass cover under different illumination modes. The image condition information includes light source intensity parameters, illumination angle parameters, and polarizer angle parameters. The image condition information is used to represent the illumination state and optical configuration when acquiring the image.

3. The method for detecting surface scratches on a mobile phone microcrystalline glass cover as described in claim 2, characterized in that, Extracting the reflectance and scatterance of non-defective regions from image data, specifically including: Before extracting the reflection and scattering brightness of non-defective regions in the image data, normalization processing is performed on the image data, and region determination is performed on the normalized image data. The region determination is a process of identifying defective and non-defective regions in the normalized image, specifically as follows: Pixel-level filtering and connected component analysis are performed on the normalized image according to preset brightness change thresholds, gradient continuity judgment criteria, and preset connected area thresholds: When the grayscale difference between adjacent pixels exceeds the brightness change threshold and meets the gradient continuity judgment criterion in the main direction, the current pixel is added to the candidate defect set; connected component analysis is performed on the candidate defect set, connected components whose area is less than the preset connected area threshold are removed, and the remaining connected components are marked as defect regions, and the remaining regions after removing the defect regions and boundary regions are marked as non-defect regions. In the non-defect areas of the lateral illumination image and the polarized illumination image, the normalized gray mean is calculated separately, and the reflected brightness is obtained by summing according to the preset weights. When there is only a single lateral illumination angle, the normalized gray mean of the current angle is used as the reflected brightness. When there are multiple lateral illumination angles, the reflected brightness is calculated by weighting according to the brightness angle parameter, and the weighted result is used as the reflected brightness. High-pass filtering is performed in the non-defective region of the dark-field illumination image, and the normalized mean of the high-frequency grayscale components is calculated. The normalized mean is used as the scattered brightness.

4. The method for detecting surface scratches on a mobile phone microcrystalline glass cover as described in claim 1, characterized in that, The channel luminance ratio of image data is calculated based on reflected luminance and scattered luminance. The channel luminance ratio, the first characteristic index, and the second characteristic index of the image data are then combined into a layer feature vector, specifically including: The formula for calculating the channel luminance ratio is: ； In the formula, This represents the channel luminance ratio of image data; Indicates the reflectance brightness of image data; Indicates the scattering brightness of image data; To avoid extremely small positive constants with a denominator of zero; After performing dimensional consistency processing on the channel brightness ratio, the first characteristic index, and the second characteristic index of the image data, they are merged into a layer feature vector in the order of the first characteristic index, the second characteristic index, and the channel brightness ratio.

5. The method for detecting surface scratches on a mobile phone microcrystalline glass cover as described in claim 1, characterized in that, A coating characteristic difference determination model is constructed. Layer feature vectors are input into the model, and the model outputs a coating difference index based on the determination. Specifically, this includes: The coating characteristic difference judgment model refers to a statistical judgment model based on a reference library, which is used to measure the degree of difference between the layer feature vector of the current batch and the reference batch. By collecting multiple batches of microcrystalline glass cover plate samples that have been judged to be qualified, the layer feature vectors are obtained under stable lighting and consistent algorithm caliber, and a reference library is established according to coating type and thickness specifications. The layer feature vector is input into the coating characteristic difference determination model. The model first performs integrity verification on the input layer feature vector; then, it performs feature standardization on the integrity-verified layer feature vector; finally, it selects corresponding reference parameters from the reference library based on the coating type and thickness specifications in the layer feature vector, and performs difference calculation based on the selected reference parameters to obtain the difference calculation result; the coating characteristic difference determination model converts the difference calculation result into a coating difference index through a mapping function, and outputs the coating difference index according to a set threshold range.

6. The method for detecting surface scratches on a mobile phone microcrystalline glass cover as described in claim 5, characterized in that, Based on the coating difference index, the lighting conditions are adaptively adjusted, generating and executing adaptive adjustment commands including light source intensity, illumination angle, and polarizer angle. Specifically, this includes: When the coating difference index falls within different judgment ranges, different levels of adaptive adjustment strategies are triggered. When the coating difference index is below the first threshold, it is determined that the current lighting conditions are in good consistency with the reference library, and no adjustment is made. When the coating difference index is between the first threshold and the second threshold, it is determined that there is a slight deviation, triggering a fine-tuning of local parameters, including increasing or decreasing the light source intensity within a preset range, and simultaneously fine-tuning the polarizer angle; When the coating difference index is higher than the second threshold, it is determined that there is a significant deviation, triggering a comprehensive adjustment strategy. Based on the fine-tuning, the lighting angle setting is further adjusted so that the lighting incident direction covers the entire angle range. The relationship between the adaptive adjustment and the coating difference index is as follows: the higher the coating difference index, the greater the difference between the optical properties of the current batch of coatings and the reference batch, and the amplitude and range of the adaptive adjustment will increase accordingly.

7. The method for detecting surface scratches on a mobile phone microcrystalline glass cover as described in claim 1, characterized in that, Based on the adaptively adjusted lighting conditions, image data of the microcrystalline glass cover to be tested is reacquired, and scratch detection is performed on the reacquired image data of the microcrystalline glass cover, specifically including: Under adaptively adjusted lighting conditions, the image data of the re-acquired microcrystalline glass cover plate is used to separate the defective and non-defective areas according to the region determination rules. Within the defect area, the brightness variation of the image data within the defect area is examined. In areas where the image data exhibits brightness variations, edge continuity and directional features are identified. For the identified scratch areas, the length, width, direction, and brightness contrast parameters of the scratch are further extracted. Based on the length, width, direction, and brightness contrast parameters, the scratch location distribution and scratch geometric feature parameters are determined. The scratch detection results are output based on the scratch location distribution and scratch geometric feature parameters.

8. The method for detecting surface scratches on a mobile phone microcrystalline glass cover as described in claim 1, characterized in that, Also includes: Based on the visual feature parameters of scratches in the image data of the microcrystalline glass cover to be tested and the preset visibility evaluation criteria, the perceived risk level of the scratches in the image data under the user's expected observation conditions is calculated. The preset rejection threshold is combined with the perceived risk level result to judge and output the current evaluation result of the microcrystalline glass cover, specifically including: The visual feature parameters include the brightness contrast, length, width, and direction of the scratch in the detection image; the brightness contrast is compared with the normalized grayscale difference of the background area, and the length and width parameters are combined to map the visibility probability of the scratch under the user's expected observation conditions. Based on the preset visibility assessment criteria, the perceived risk classification results are obtained, which include the first level interval, the second level interval, and the third level interval. The perceived risk classification results are combined with the rejection threshold. When the scratch length or width exceeds the rejection threshold and the perceived risk classification exceeds the third level, the rejection result is directly output. When the scratch size is lower than the rejection threshold but the perceived risk classification is within the second level range, a re-inspection suggestion is output. When the scratch size is lower than the rejection threshold and the perceived risk classification is within the first level range, a qualified result is output. The final evaluation results include scratch detection results, the probability of scratches being visible under the user's expected observation conditions, perceived risk classification results, and the determination of the rejection threshold.

Images