Mobile phone display screen defect detection method, device and equipment and storage medium

Abstract

The application relates to the technical field of display screen defect detection, and discloses a mobile phone display screen defect detection method, device, equipment and storage medium, the method comprising the following steps: collecting display images of a mobile phone display screen to be detected under multiple observation angles; analyzing the reflection intensity distribution of the display images by using the polarization optical characteristic, generating a de-reflection image after separating the surface reflection component; transforming the pixel brightness of each observation angle in the de-reflection image to a unified reference viewpoint to obtain a unified reference viewpoint image, and extracting a first defect feature vector of each pixel in the unified reference viewpoint image; fusing the first defect feature vectors extracted under multiple observation angles to obtain a second defect feature vector, comparing the second defect feature vector with a preset threshold, and outputting a defect detection report; the application utilizes the multi-angle information advantage, establishes a defect classification mechanism based on the RGB channel response difference, and can accurately distinguish multiple defect types such as bright spots, dark spots, color deviation and line defects.

Description

Technical Field

[0001] This application relates to the field of display screen defect detection technology, and in particular to a method, apparatus, device and storage medium for detecting defects in mobile phone displays. Background Technology

[0002] Traditional mobile phone display defect detection methods primarily rely on image acquisition and analysis from a fixed viewing angle. However, with the continuous improvement of display resolution and the refinement of display technology, the limitations of traditional detection methods are becoming increasingly apparent. Existing display defect detection technologies are significantly inadequate when dealing with changes in the viewing angle. The same defect may exhibit different visual characteristics under different viewing angles, leading to instability and unreliability in the detection results. Traditional methods typically select a single optimal viewing angle for detection or rely on high-precision angle calibration equipment and complex optical compensation systems. This not only increases equipment costs and system complexity but also makes it difficult to adapt to the variable detection conditions in actual production environments, resulting in low accuracy in existing defect detection methods. Summary of the Invention

[0003] This application provides a method, apparatus, device, and storage medium for detecting defects in mobile phone displays. This application utilizes the advantages of multi-angle information to establish a defect classification mechanism based on the response differences of RGB channels, which can accurately distinguish various defect types such as bright spots, dark spots, color shifts, and line defects.

[0004] The first aspect of this application provides a method for detecting defects in a mobile phone display screen, the method comprising:

[0005] Collect images of the mobile phone screen under test from multiple viewing angles;

[0006] The reflection intensity distribution of the displayed image is analyzed using polarization optics characteristics, and a dereflection image is generated after separating the surface reflection components.

[0007] The pixel brightness of each viewing angle in the dereflection image is transformed to a unified reference viewpoint to obtain a unified reference viewpoint image, and the first defect feature vector of each pixel in the unified reference viewpoint image is extracted.

[0008] The first defect feature vector extracted from multiple observation angles is fused to obtain the second defect feature vector, and the second defect feature vector is compared with a preset threshold to output a defect detection report.

[0009] In conjunction with the first aspect, in a first implementation of the first aspect of the present invention, the step of acquiring display images of the mobile phone screen to be detected at multiple viewing angles includes:

[0010] Determine multiple observation angles according to the preset angle acquisition range;

[0011] The CCD camera array is controlled to simultaneously acquire RGB three-channel images of the mobile phone display screen from various viewing angles, while recording the corresponding ambient light intensity parameters and display screen brightness parameters.

[0012] The RGB three-channel images acquired from each observation angle are preprocessed and filtered to generate standardized angle images;

[0013] Establish a database mapping relationship between the multiple observation angles and the corresponding standardized angle images to form display images under multiple observation angles.

[0014] In conjunction with the first aspect, in a second implementation of the first aspect of the present invention, the step of analyzing the reflection intensity distribution of the displayed image using polarization optical characteristics, separating the surface reflection components, and generating a dereflected image includes:

[0015] A first reflection intensity calculation model is established based on the polarization optical characteristics of the mobile phone display screen, including the s-polarization reflection component, the p-polarization reflection component, and the critical angle parameter.

[0016] Calculate the pixel intensity difference between adjacent viewing angles in the displayed image, extract the intensity change pattern caused by reflection, and correct the polarization component coefficient in the first reflection intensity calculation model based on the intensity change pattern to obtain the second reflection intensity calculation model;

[0017] The second reflection intensity calculation model is combined with the ambient light intensity parameter to calculate the predicted reflection intensity value of each pixel under different viewing angles.

[0018] The predicted reflection intensity value is subtracted from the original intensity value of the displayed image pixel by pixel to generate a dereflected image after separating the surface reflection components.

[0019] In conjunction with the first aspect, in a third implementation of the first aspect of the present invention, the step of transforming the pixel brightness of each viewing angle in the dereflection image to a unified reference viewpoint to obtain a unified reference viewpoint image, and extracting the first defect feature vector of each pixel in the unified reference viewpoint image, includes:

[0020] The brightness attenuation characteristics of normal pixels and defective pixels in the dereflection image under different viewing angles are analyzed, and a pixel brightness change model is established based on the brightness attenuation characteristics.

[0021] Extract the brightness gradient information of each pixel position in the dereflection image as the viewing angle changes, calculate the pixel viewing angle sensitivity coefficient, and use the pixel brightness change model to identify the angle response anomaly features of defective pixels.

[0022] A reference angle value is selected as a unified reference viewpoint. The brightness compensation coefficient from each observation angle to the reference angle value is calculated based on the pixel brightness change model and the pixel viewpoint sensitivity coefficient. The result is then assembled into a viewpoint compensation transformation matrix.

[0023] The viewpoint compensation transformation matrix is ​​applied to perform pixel-by-pixel brightness compensation on the de-reflection image to generate a unified reference viewpoint image;

[0024] The brightness differences of RGB sub-pixels in the unified reference viewpoint image are analyzed, and the first defect feature vector of each pixel is extracted.

[0025] In conjunction with the first aspect, in a fourth implementation of the first aspect of the present invention, the step of analyzing the RGB sub-pixel brightness differences of the unified reference viewpoint image and extracting the first defect feature vector of each pixel includes:

[0026] Separate the red, green and blue color channels of the unified reference viewpoint image to obtain red sub-pixel image, green sub-pixel image and blue sub-pixel image respectively;

[0027] Calculate the first viewpoint response difference parameter of the red sub-pixel image, the second viewpoint response difference parameter of the green sub-pixel image, and the third viewpoint response difference parameter of the blue sub-pixel image, respectively.

[0028] Construct an RGB response difference matrix based on the first viewpoint response difference parameter, the second viewpoint response difference parameter, and the third viewpoint response difference parameter;

[0029] For the Fresnel subpixel arrangement structure of a mobile phone display, the brightness consistency deviation between each pixel and its neighboring subpixels is analyzed using the RGB response difference matrix, and the subpixel consistency index is calculated.

[0030] The channel features of the RGB response difference matrix are fused with the sub-pixel consistency index to generate a first defect feature vector.

[0031] In conjunction with the first aspect, in a fifth implementation of the first aspect of the present invention, the step of fusing the first defect feature vectors extracted from multiple observation angles to obtain a second defect feature vector, and comparing the second defect feature vector with a preset threshold to output a defect detection report, includes:

[0032] The optimal detection angle is selected as the basis for weight calculation, and the angular distance weight of each observation angle relative to the optimal detection angle is calculated using the exponential decay function.

[0033] Analyze the statistical distribution characteristics of the first defect feature vector under different observation angles, calculate the standard deviation and mean, and calculate the ratio based on the standard deviation and the mean to obtain the feature quality assessment factor;

[0034] The reliability weights for each observation angle are formed by combining the angular distance weights with the corresponding feature quality evaluation factors.

[0035] The reliability weights are used to perform a weighted fusion operation on the first defect feature vectors from each observation angle to output a second defect feature vector.

[0036] The second defect feature vector is compared with a preset threshold to determine the defect pixel location and classify the defect type, and a defect detection report is output.

[0037] In conjunction with the first aspect, in the sixth implementation of the first aspect of the present invention, the step of comparing the second defect feature vector with a preset threshold, determining the defect pixel location and classifying the defect type, and outputting a defect detection report includes:

[0038] The second defect feature vector is input into a support vector machine classifier for binary classification, and the probability of defect presence at each pixel position is output.

[0039] Pixel locations with a probability of defect presence exceeding a preset threshold are selected, and defect regions are identified and defect pixel coordinates are determined based on the pixel locations. At the same time, the spatial geometric features of each defect region are recorded.

[0040] Extract the RGB channel response difference information from the second defect feature vector corresponding to the defect pixel coordinates and perform defect classification to obtain the defect type;

[0041] The severity score is calculated by combining the probability of the defect's existence with the spatial geometric features of the corresponding defect area. The defect pixel coordinates, defect type determination results, and severity score are then integrated to output a defect detection report.

[0042] A second aspect of this application provides a defect detection device for a mobile phone display screen, the defect detection device for the mobile phone display screen comprising:

[0043] The acquisition module is used to acquire images of the mobile phone screen under inspection from multiple viewing angles;

[0044] The analysis module is used to analyze the reflection intensity distribution of the displayed image using polarization optical characteristics, and generate a dereflected image after separating the surface reflection components;

[0045] The transformation module is used to transform the pixel brightness of each viewing angle in the dereflection image to a unified reference viewpoint to obtain a unified reference viewpoint image, and extract the first defect feature vector of each pixel in the unified reference viewpoint image.

[0046] The output module is used to fuse the first defect feature vectors extracted from multiple observation angles to obtain the second defect feature vector, compare the second defect feature vector with a preset threshold, and output a defect detection report.

[0047] A third aspect of this application provides an electronic device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the electronic device to perform the aforementioned method for detecting defects in a mobile phone display screen.

[0048] A fourth aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned method for detecting defects in a mobile phone display screen.

[0049] Compared with existing technologies, this application has the following advantages: By establishing a viewpoint compensation transformation matrix, a unified transformation of pixel brightness under different viewing angles is achieved, significantly enhancing the adaptability of the detection system to viewpoint changes and eliminating the problem of inconsistent detection results caused by different viewing angles in traditional methods. A polarization optics model is used to accurately separate the surface reflection component and the internal light-emitting component of the display screen, effectively eliminating ambient light reflection interference and enabling the detection system to maintain stable performance under complex lighting conditions. For the Fresnel sub-pixel arrangement structure of mobile phone displays, fine sub-pixel-level detection is achieved through RGB response difference matrix and sub-pixel consistency analysis, with detection accuracy far exceeding that of traditional pixel-level methods. By calculating the reliability weights of each viewing angle and performing adaptive weighted fusion, the advantages of multi-angle information are fully utilized to establish a defect classification mechanism based on RGB channel response differences, which can accurately distinguish various defect types such as bright spots, dark spots, color shifts, and line defects. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] The structures, proportions, sizes, etc., shown in the accompanying drawings of this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0052] Figure 1 This is a flowchart illustrating the defect detection method for a mobile phone display screen provided in an embodiment of the present invention;

[0053] Figure 2 This is a schematic block diagram of the structure of the defect detection device for a mobile phone display screen provided in an embodiment of the present invention;

[0054] Figure 3 This is a schematic block diagram of the structure of the electronic device provided in the embodiment of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0057] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0058] It should also be further understood that the term "and / or" as used in this application specification and the appended claims refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations. See also Figure 1 One embodiment of the defect detection method for mobile phone displays in this application includes:

[0059] Step 100: Acquire display images of the mobile phone screen to be tested from multiple viewing angles;

[0060] It is understood that the executing entity of this application can be a defect detection device for a mobile phone display screen, or it can be a terminal or a server; no specific limitation is made here. This application's embodiment uses a server as an example for illustration.

[0061] Specifically, a preset angle sequence containing actual viewing angles is set, with the angle range from -45° to +45°. The entire angle space is discretized using fixed-interval angle increments (e.g., 5°) to ensure that the acquisition covers typical observation positions from looking up to looking down. During image acquisition, the CCD camera array is controlled to mechanically rotate or switch arrays according to the aforementioned angle sequence, so that each observation angle θ corresponds to a camera viewpoint position. A synchronous triggering mechanism is used to acquire RGB three-channel images of the same display content at the same time, obtaining a color image dataset in (x,y) space. The acquired image data undergoes a unified image preprocessing workflow, including noise reduction filtering, color equalization, and brightness normalization, to eliminate systematic errors and random interference during imaging, generating a standardized angle image set. An index system for the angle images is constructed, associating each acquired angle with its corresponding standardized image for storage. A mapping structure is established in the database to achieve efficient retrieval and calling from any acquired angle to its standardized image, forming display images from multiple observation angles.

[0062] Step 200: Analyze the reflection intensity distribution of the displayed image using polarization optical characteristics, and generate a dereflected image after separating the surface reflection components;

[0063] Specifically, based on the optical polarization characteristics between the glass layer on the surface of the mobile phone display and the backlight structure, a first reflection intensity calculation model is established, including s-polarized reflection components, p-polarized reflection components, and critical angle parameters. This first reflection intensity calculation model describes the reflection behavior of incident light under different polarization states, forming a theoretical estimation framework for pixel-level reflection intensity. Based on multi-angle image acquisition, pixel intensity difference calculation is performed on images between any adjacent angles to extract the light reflection interference change pattern caused by angle changes. Using this light reflection interference change pattern as a basis, the s- and p-polarization component coefficients in the first model are corrected term by term to obtain a second reflection intensity calculation model that more closely approximates the actual optical response behavior. The updated second model is then combined with the ambient light intensity parameters recorded synchronously during image acquisition to consider the actual contribution of external light to the reflection intensity. This allows for the prediction of the expected reflection intensity value for each pixel (x, y) at each observation angle θ, forming a multi-angle predicted reflection image sequence. The predicted reflection value at each angle is subtracted from the brightness value of the original image pixel by pixel, thereby effectively separating the surface specular reflection component in the displayed image and retaining the pure internal light-emitting component, forming de-reflection image data containing the real display content. Step 300: The pixel brightness of each observation angle in the de-reflection image is transformed to a unified reference viewpoint to obtain a unified reference viewpoint image, and the first defect feature vector of each pixel in the unified reference viewpoint image is extracted;

[0064] Specifically, pixel-by-pixel brightness response analysis is performed on multi-angle image sequences after reflection interference has been removed. By comparing the brightness attenuation curves of normal pixels and potentially defective pixels at various viewing angles, the light intensity attenuation law as a function of angle is extracted. A pixel brightness change model including reference brightness, normal angle, and defect attenuation factor is established to characterize the angular response behavior of pixels at different viewpoints. The brightness gradient information of each pixel in the dereflected image as a function of the viewing angle θ is extracted. The viewpoint sensitivity coefficient of the pixel is obtained by calculating the gradient derivative to measure the degree of pixel brightness response to viewpoint shift. The viewpoint sensitivity coefficient is substituted into the pixel brightness change model for fitting analysis to identify pixels with abnormal deviations in angular response behavior. These pixels have skewed or abrupt angle response curves due to abnormal light-emitting structures or thin film stack defects, which serve as criteria for potential defect areas. After establishing the viewpoint response model, a physically meaningful viewing angle (e.g., positive viewing angle θ=0°) is selected as a unified reference viewpoint. Based on the response model and corresponding sensitivity coefficient of each pixel at different viewing angles, the brightness compensation coefficient required from the current angle to the reference viewpoint is calculated, and a viewpoint compensation transformation matrix for full-angle normalization is constructed. A viewpoint compensation transformation matrix is ​​applied to perform pixel-by-pixel brightness correction on all dereflection images, uniformly mapping the image brightness from different angles to the reference viewpoint to generate a unified reference viewpoint image. Sub-pixel separation is performed on the RGB three channels of the unified reference viewpoint image. By analyzing the brightness distribution differences of each sub-pixel within a local spatial range, multi-dimensional sub-pixel defect description information, including average brightness, standard deviation, inter-channel inconsistency, and local contrast, is extracted and combined to form a first defect feature vector, characterizing the display stability and local anomaly features of each pixel under the standard viewpoint.

[0065] Step 400: Merge the first defect feature vectors extracted from multiple observation angles to obtain the second defect feature vector, compare the second defect feature vector with a preset threshold, and output a defect detection report.

[0066] Specifically, the angle image with the strongest defect response capability is selected as the optimal detection angle, corresponding to the observation angle with the highest defect contrast and strongest feature stability, and used as the benchmark for weight calculation. An angle distance weight model is constructed using an exponential decay function. The angle difference between any observation angle and the optimal detection angle is calculated using an exponential function, causing the weight assigned to image features farther from the optimal angle to decrease exponentially with distance, ensuring that images closer to the optimal angle have higher weight influence. Based on the established angle distance weight, the first defect feature vector extracted at each angle is statistically analyzed, calculating its mean and standard deviation. A feature quality assessment factor is constructed based on the ratio between these two factors to measure the consistency and stability of the angle image in defect representation, reflecting the reliability of the features. The angle distance weight and the feature quality assessment factor are multiplied to obtain the comprehensive reliability weight for each angle. Based on this, the first defect feature vectors from each angle are weighted, superimposed, and fused to form the second defect feature vector. The second defect feature vector is compared with a preset defect identification threshold. The defect probability components and spatial consistency index are analyzed point-by-point along the pixel dimension. When the second defect feature vector exceeds the threshold, the pixel is determined to be a defective pixel. Combining channel heterogeneity, local contrast abrupt changes, and directional linear structure from the fused features, the defect type is identified as a bright spot, dark spot, color cast, or linear defect, etc. The detected defect pixel coordinates, defect type, severity score, and overall detection stability and angle consistency index are integrated and processed to output a standardized defect detection report.

[0067] In one specific embodiment, the process of performing step 100 may specifically include the following steps:

[0068] Determine multiple observation angles according to the preset angle acquisition range;

[0069] The CCD camera array is controlled to simultaneously acquire RGB three-channel images of the mobile phone display screen from various viewing angles, while recording the corresponding ambient light intensity parameters and display screen brightness parameters.

[0070] The RGB three-channel images acquired from various observation angles are preprocessed and filtered to generate standardized angle images;

[0071] Establish a database mapping relationship between multiple observation angles and corresponding standardized angle images to form display images under multiple observation angles.

[0072] Specifically, based on common human eye viewing angles and the screen's visible range, a comprehensive angle acquisition range is set from −45° to +45°. Within this range, multiple discrete angle points are divided at equal intervals. For example, with a 5° increment, 19 effective acquisition angles are obtained. This ensures that the set angle sequence can fully cover the brightness changes and reflection phenomena that occur when users observe the display screen from different directions during actual use. The multi-channel CCD camera array is controlled to mechanically rotate or electronically switch channels according to the preset angle acquisition sequence, so that each angle point θ1, θ2, ..., θ n Each CCD sensing channel or optical axis position corresponds to an independent CCD sensing channel, enabling precise positioning of observation positions from multiple angles. A synchronous acquisition controller triggers all CCD cameras at all angles in parallel or serial mode, acquiring RGB three-channel image data of the mobile phone display under test within a unified time window. Each image corresponds to a unique observation angle. Simultaneously, during the same acquisition time period, an ambient light sensor measures the current ambient illuminance parameters in real time and reads the brightness setpoint from the display control module to obtain the illumination boundary conditions. Preprocessing and filtering are applied to the RGB three-channel images acquired from each observation angle, including image smoothing, edge noise removal, color balancing, and brightness normalization. Median filters or bilateral filters are used to remove isolated pixel anomalies caused by sensors or noise. Color space transformation, such as RGB to Lab color normalization, improves image color consistency. Brightness histogram equalization eliminates brightness deviations between angles caused by directional screen emission, outputting a standardized image with consistent brightness and color distribution. After standardization processing for all angles, a database mapping structure is established, mapping each angle θ... i Its corresponding standardized image I i The image is bound to (x, y, c), and indexed fields including angle number, image content, ambient light parameters, brightness setting parameters, and image timestamp are constructed and stored in the image database in a structured data format such as JSON, XML, or relational database tables.

[0073] In one specific embodiment, the process of performing the steps of analyzing the reflection intensity distribution of the displayed image using polarization optical characteristics, separating the surface reflection components, and generating a dereflected image can specifically include the following steps:

[0074] A first reflection intensity calculation model is established based on the polarization optical characteristics of the mobile phone display screen, including the s-polarization reflection component, the p-polarization reflection component, and the critical angle parameter.

[0075] The pixel intensity difference between adjacent viewing angles in the displayed image is calculated, the intensity change pattern caused by reflection is extracted, and the polarization component coefficient in the first reflection intensity calculation model is corrected based on the intensity change pattern to obtain the second reflection intensity calculation model.

[0076] The second reflection intensity calculation model is combined with the ambient light intensity parameter to calculate the predicted reflection intensity value of each pixel under different viewing angles.

[0077] The corresponding predicted reflection intensity value is subtracted from the original intensity value of the displayed image pixel by pixel, and the dereflection image is generated after the separation of surface reflection components is completed.

[0078] Specifically, the surface reflection characteristics of the display screen are theoretically defined from the perspective of optical modeling. Since the surface of a mobile phone display screen is made of glass or a composite medium coated with a transparent conductive film, it exhibits significant polarization-dependent reflection behavior under incident light. Therefore, a reflection model with the incident angle as the independent variable is used to describe it, constructing an initial first reflection intensity calculation model. This model introduces intensity expressions for the s-polarization and p-polarization components, and determines the reflection ratio under the two polarization directions through the geometric relationship between the angle parameter θ and the critical angle θ0. The s-polarization term mainly reflects the reflection component when the electric field is perpendicular to the incident plane, while the p-polarization term reflects the reflection response when the electric field is parallel to the incident plane. These two terms constitute the decomposition of the total reflection intensity in different polarization directions. Simultaneously, the critical angle is introduced as a structural parameter of the control function to achieve unified modeling of the physical properties of the material interface. After the model is established, from the display image sequence acquired from multiple angles, for each pixel coordinate (x, y), adjacent angles θ are calculated. i With θ i+1 The brightness difference between pixels is calculated to obtain the response difference of pixel intensity as the angle changes. The response difference is mainly caused by the fluctuation of reflection intensity due to the change of incident angle, and is therefore regarded as a brightness disturbance signal caused by reflection. By aggregating and statistically analyzing the brightness disturbance signal in the spatial range, and combining it with the previously established polarization reflection model, the reflection increment behavior of the actual pixel during the change of observation angle is backfitted. The coefficients of the s-polarization component and p-polarization component in the first model are dynamically corrected using the fitting error feedback mechanism to generate a second reflection intensity calculation model. The second reflection intensity calculation model has angle response adaptability and can automatically adjust the proportion of the reflection term in the high-angle or low-angle region to adapt to different pixel material structures. The corrected second model is fused with the ambient light intensity parameters recorded synchronously during the acquisition process. The ambient light is substituted into the model calculation formula as the external light source intensity factor, so as to predict the reflection intensity of each pixel position (x,y,θ) at each angle, and obtain a pixel-level reflection prediction image, which reflects the brightness information contributed by surface specular reflection under specific lighting conditions and angle parameters. The corresponding predicted reflection intensity value is subtracted from the original intensity value of the displayed image pixel by pixel to eliminate brightness anomalies caused by interface reflection, resulting in image data after reflection component removal.

[0079] The first reflection intensity calculation model, based on the polarization optical characteristics of a mobile phone display screen, includes s-polarization reflection components, p-polarization reflection components, and critical angle parameters. This model involves: measuring the optical refractive index and dielectric constant of the surface material of the mobile phone display screen; calculating the s-polarization theoretical reflection coefficient and p-polarization theoretical reflection coefficient at various viewing angles according to Fresnel's law of reflection; analyzing the brightness distribution characteristics of the displayed image at different viewing angles; identifying the angular relationship between the display screen normal vector and the viewing direction; determining the critical angle parameters affecting the reflection intensity calculation; and establishing a mathematical relationship between the reflection intensity and the incident angle based on the s-polarization theoretical reflection coefficient and the p-polarization theoretical reflection coefficient. The model is constructed by using a sine square function to describe the angle dependence of the s-polarization component and a cosine square function to describe the angle dependence of the p-polarization component. Angle corrections are applied to the mathematical relationship model using critical angle parameters to construct a corrected reflection intensity model that considers the geometric characteristics of the display screen surface. This model is used to predict the theoretical reflected light intensity at different viewing angles for each pixel position. Ambient light intensity parameters are substituted into the corrected reflection intensity model to calculate the predicted surface reflection intensity values ​​for each viewing angle and pixel position. The consistency between the predicted surface reflection intensity values ​​and the actual measured reflected light intensity is verified. The model parameters are iteratively optimized and adjusted to form the first reflection intensity calculation model.

[0080] In one specific embodiment, the process of performing step 200 may specifically include the following steps:

[0081] The brightness attenuation characteristics of normal and defective pixels in dereflection images under different viewing angles are analyzed, and a pixel brightness change model is established based on the brightness attenuation characteristics.

[0082] Extract the brightness gradient information of each pixel position in the dereflection image as the viewing angle changes, calculate the pixel viewing angle sensitivity coefficient, and use the pixel brightness change model to identify the abnormal angular response characteristics of defective pixels.

[0083] A reference angle value is selected as a unified reference viewpoint. The brightness compensation coefficient from each observation angle to the reference angle value is calculated based on the pixel brightness change model and the pixel viewpoint sensitivity coefficient. The result is then assembled into a viewpoint compensation transformation matrix.

[0084] Apply the viewpoint compensation transformation matrix to perform pixel-by-pixel brightness compensation on the dereflection image to generate a unified reference viewpoint image;

[0085] Analyze the RGB sub-pixel brightness differences of the unified reference viewpoint image and extract the first defect feature vector of each pixel.

[0086] Specifically, based on dereflection images, the brightness change trajectory of each pixel in the angular dimension is extracted from clean display images viewed from multiple angles. By statistically summarizing the brightness curves of a large number of normal area pixels, it is found that their brightness decay behavior with angle θ has consistent characteristics, that is, it satisfies the decay form based on Lambert's cosine law. The maximum brightness is located near the normal viewing direction (e.g., θ=0°) and gradually decreases with angle deviation, showing a symmetrical and smooth response curve. In contrast, defective pixels, due to damage to their light-emitting structure or abnormal pixel control, will cause discontinuities, abrupt changes, or asymmetries in the brightness response under different viewing angles. Therefore, by fitting the angular brightness curve and comparing the deviation between the angular brightness curve and the theoretical model, a brightness change model describing the consistency of pixel light emission is constructed. The brightness change model uses the reference brightness value, normal angle, and local attenuation factor as core parameters to reflect the brightness response function of the pixel under multiple angles. For each pixel, a first-order difference or fitted derivative operation is performed along the angular direction on the brightness value at all acquisition angles to obtain the sensitivity of brightness to angle changes, i.e., extracting the brightness gradient information of each pixel. Based on the range and continuity of the brightness gradient value, a viewing angle sensitivity coefficient is constructed for each pixel, representing the intensity of the effect of angular perturbation on the pixel response. When the sensitivity index of some pixels deviates significantly from its neighborhood average, it is inferred that they belong to potential defective pixel regions, especially pixels exhibiting reverse brightening or sharp attenuation at high angles, which are micro-bright spots or dark spots. θ=0° or other optimal detection response angles are selected as a unified reference viewpoint. The pixel brightness change model is used to inversely deduce the brightness compensation coefficients corresponding to the reference viewpoint for each angle. The brightness compensation coefficient represents the intensity gain or attenuation ratio required for a pixel to be mapped from the current viewpoint to the reference angle. The set of compensation coefficients at all angles is combined into a spatially consistent viewpoint compensation transformation matrix. The viewpoint compensation transformation matrix is ​​applied to perform pixel-by-pixel brightness compensation on all dereflection images, adjusting the brightness of each pixel at θ... i The actual brightness value is multiplied by the corresponding compensation coefficient to ensure that the image at all angles achieves a consistent brightness state at the reference viewpoint after compensation, eliminating pixel intensity fluctuations caused by viewpoint differences and generating a unified reference viewpoint image. In the unified viewpoint image, the image is decomposed into sub-pixel levels according to the RGB channels. The brightness values ​​of each pixel in the R, G, and B channels are analyzed to determine whether there are significant imbalances, abnormal increases or decreases, or inconsistencies between adjacent pixels. The first defect feature vector is constructed by combining multiple sub-features such as brightness variance, color shift, and gradient continuity.

[0087] In this embodiment, the brightness attenuation characteristics of normal pixels and defective pixels in the dereflection image under different viewing angles are analyzed, and a pixel brightness change model is established based on the brightness attenuation characteristics. This includes: collecting multi-angle reference brightness data of the mobile phone display screen under standard testing conditions; establishing a Lambertian luminous characteristic reference curve for normal pixels under different viewing angles; determining the theoretical attenuation law of pixel brightness with angle variation; identifying the positions of known defective pixels in the dereflection image as training samples; measuring the actual brightness response of bright spot defects, dark spot defects, and color shift defects at various viewing angles; establishing an angle response feature library for different defect types; calculating the deviation between the reference curve of normal pixels and the angle response curves of various defective pixels; extracting the unique brightness attenuation anomaly patterns of defects; and defining... The defect attenuation coefficient is defined as a key parameter to distinguish between normal pixels and defective pixels. Based on the physical mechanism of display light emission, a four-parameter mathematical model of pixel brightness variation is established, including reference brightness, viewing angle, normal angle, and defect attenuation coefficient. The reference brightness reflects the intrinsic luminous intensity of the pixel, and the normal angle reflects the geometric characteristics of the display screen. The coefficients of each parameter in the four-parameter pixel brightness variation mathematical model are optimized using the least squares fitting algorithm with training sample data, and the mapping relationship between model parameters and defect types is established to form a pixel brightness variation model suitable for viewpoint compensation. The prediction accuracy of the pixel brightness variation model at unknown pixel positions is verified, and the generalization ability of the model is evaluated through cross-validation to ensure the stability of the model in different display batches and testing environments.

[0088] In one specific embodiment, the process of performing the step of analyzing the RGB sub-pixel brightness differences of the unified reference viewpoint image and extracting the first defect feature vector of each pixel can specifically include the following steps:

[0089] Separate the red, green and blue color channels of the unified reference viewpoint image to obtain red sub-pixel images, green sub-pixel images and blue sub-pixel images respectively;

[0090] Calculate the first-viewpoint response difference parameters of the red sub-pixel image, the second-viewpoint response difference parameters of the green sub-pixel image, and the third-viewpoint response difference parameters of the blue sub-pixel image, respectively.

[0091] Construct an RGB response difference matrix based on the first-view response difference parameters, the second-view response difference parameters, and the third-view response difference parameters;

[0092] For the Fresnel subpixel arrangement structure of mobile phone displays, the brightness consistency deviation between each pixel and its neighboring subpixels is analyzed using the RGB response difference matrix, and the subpixel consistency index is calculated.

[0093] The first defect feature vector is generated by fusing the channel features of the RGB response difference matrix with the sub-pixel consistency index.

[0094] Specifically, the red, green, and blue color channels of the unified reference viewpoint image are separated, with each channel representing the intensity distribution of the corresponding color channel across all pixel locations. Angle response difference parameters are calculated for each of the three channels, tracing the pixel's response trajectory across multiple angles to assess the stability of its color channels with angular variations. When analyzing the red channel image, the root mean square of the brightness difference at each angle is calculated based on the curve formed by the red channel brightness changing with angle in the original angle image sequence, forming the first viewpoint response difference parameter to measure the stability of the red sub-pixel's angle response. Similarly, the same difference analysis is performed on the green and blue sub-pixel images to obtain the second and third viewpoint response difference parameters. These three parameters are then unified into an RGB response difference matrix, where each element corresponds to a pixel's response fluctuation measure in a specific channel. The three columns of the matrix correspond to the R, G, and B channel difference information, respectively, forming a multi-dimensional vector space describing the pixel's angle response characteristics. After obtaining the RGB response difference matrix, considering that the mobile phone display uses a Fresnel-based subpixel arrangement, where each pixel is not a single light-emitting unit but consists of three subpixels (R, G, and B) in a fixed order, the brightness consistency analysis compares the brightness anomaly of a single subpixel with that of its neighboring subpixels in the same physical pixel structure. A sliding window is constructed to extract the RGB values ​​of neighboring pixels within the area surrounding each subpixel, calculating the brightness consistency deviation of the current pixel and its neighbors in each channel. This deviation is expressed as channel absolute difference or Euclidean distance, yielding a subpixel consistency index for each pixel. A higher subpixel consistency index indicates a more unstable local brightness structure and a higher likelihood of a defect. The difference features of the three channels in the RGB response difference matrix are then fused with the subpixel consistency index. Linear combination, weighted averaging, or principal component compression are used to combine the four components into a unified first defect feature vector.

[0095] In this embodiment, for the Fresnel subpixel arrangement structure of a mobile phone display screen, the brightness consistency deviation between each pixel and its neighboring subpixels is analyzed using the RGB response difference matrix, and the subpixel consistency index is calculated. This includes: obtaining the Fresnel subpixel arrangement parameters of the mobile phone display screen, including the spatial coordinates, geometric dimensions, and relative arrangement angles of the red, green, and blue subpixels, and establishing a subpixel spatial topology model; based on the subpixel spatial topology model, determining the neighborhood range of neighboring subpixels for each pixel position; establishing a first neighborhood template containing surrounding green and blue subpixels for red subpixels, a second neighborhood template containing surrounding red and blue subpixels for green subpixels, and a third neighborhood template containing surrounding red and green subpixels for blue subpixels; and using the red channel difference parameter in the RGB response difference matrix, combined with the first neighborhood... The template calculates the brightness difference between the red subpixel and other subpixels in its neighborhood, and obtains the consistency deviation of the red subpixel through statistical analysis. Using the green and blue channel difference parameters in the RGB response difference matrix, combined with the second and third neighborhood templates respectively, the consistency deviation of the green and blue subpixels is calculated. Combining the consistency deviations of the red, green, and blue subpixels, a weighted average algorithm is used to calculate the overall consistency deviation of each pixel position, where the weighting coefficients are set according to the sensitivity of the human eye to different colors. Based on the overall consistency deviation, a subpixel consistency evaluation function is established, and a consistency threshold range is set. When the overall consistency deviation exceeds the normal range, it is marked as a potential defect area, forming a quantified subpixel consistency index.

[0096] In one specific embodiment, the process of performing step 300 may specifically include the following steps:

[0097] The optimal detection angle is selected as the basis for weight calculation, and the angular distance weight of each observation angle relative to the optimal detection angle is calculated using the exponential decay function.

[0098] The statistical distribution characteristics of the first defect feature vector under different observation angles are analyzed, the standard deviation and mean are calculated, and the feature quality assessment factor is obtained by calculating the ratio based on the standard deviation and mean.

[0099] By combining the angular distance weights with the corresponding feature quality assessment factors, a reliability weight for each observation angle is formed.

[0100] The reliability weights are used to perform a weighted fusion operation on the first defect feature vectors from each observation angle, and the second defect feature vector is output.

[0101] Compare the second defect feature vector with the preset threshold to determine the defect pixel location and classify the defect type, and output a defect detection report.

[0102] Specifically, the optimal detection angle is selected as the basis for weight calculation. The selection of the optimal angle is based on the contrast analysis results of defect responses in images from multiple angles. The angle with the most obvious defect features, the highest signal-to-noise ratio, or the strongest color separation is prioritized as the optimal detection angle to ensure the best defect identifiability. To avoid unnecessary interference from other angles to the final fused features, an exponential decay function is introduced to construct an angle distance weight model. This exponential angle distance function ensures that the angle distance weight of images at greater distances is smaller, thereby enhancing the information value of images near the optimal detection angle and suppressing interference from edge views. Statistical analysis is performed on the first defect feature vector extracted at each angle, extracting the mean and standard deviation as two distribution characteristic parameters. The ratio between the standard deviation and the mean is used as the quality judgment criterion to construct a feature quality evaluation factor. A smaller feature quality evaluation factor indicates that the image features are concentrated and stable with little fluctuation, indicating high reliability and suitability for defect fusion processing. A larger ratio indicates that the feature response at the angle is severely affected by noise or the defect is not obvious, and its weight should be reduced. Combining the angle distance weight with the corresponding feature quality evaluation factor forms a comprehensive reliability weight. Based on reliability weights, the first defect feature vectors from each angle are weighted and fused pixel by pixel to output a second defect feature vector. This second defect feature vector is then compared with a preset defect judgment threshold, with pixel-by-pixel difference evaluation. When the fused feature of a pixel exceeds the defect judgment threshold, it is directly identified as a defective pixel. Defect types are identified and classified based on indicators such as brightness extremes, color shifts, edge strength, and channel inconsistencies of the fused features, categorizing defects into different manifestation types such as bright spots, dark spots, color casts, and linear defects. The coordinates of all detected defective pixels, their corresponding defect type labels, feature response intensity values, and the feature evaluation coefficients from each angle on which the detection relies are structured and organized. Combined with angle consistency indicators and fused feature stability indices, a defect detection report is output.

[0103] In this embodiment, the reliability weights for each observation angle are formed by combining the angle distance weights with the corresponding feature quality assessment factors. This includes: establishing a multi-dimensional index system for angle reliability assessment, including angle stability index, feature consistency index, and environmental interference resistance index. Angle stability is measured by the feature change rate between consecutive angles, and feature consistency is measured by the feature correlation of the same pixel at different angles. The temporal stability of the first defect feature vector at each observation angle is calculated. By analyzing the fluctuations of feature vectors in consecutive frames, unstable angles caused by factors such as changes in ambient light and fluctuations in display brightness are identified. A quality assessment algorithm based on feature vector distribution entropy is established to calculate the information entropy value of the feature vector at each observation angle. The smaller the entropy value, the more stable the feature vector. The more concentrated the feature distribution, the higher the quality; the larger the entropy value, the more dispersed the features and the lower the reliability. A mutual verification mechanism between angles is introduced. By comparing the similarity of feature vectors extracted from adjacent observation angles, abnormal angles are identified and their weight contributions are reduced, establishing a mutual verification and compensation mechanism between angles. Combining angle distance weight, feature quality evaluation factor, temporal stability score, and information entropy quality score, a multi-objective optimization algorithm is used to calculate the comprehensive reliability weight of each observation angle, ensuring that the weight allocation considers both geometric distance factors and feature quality. A dynamic update mechanism for reliability weights is established, adjusting the weight parameters based on real-time feedback information during the detection process. When the detection environment of a certain angle changes, its reliability weight is automatically recalculated, ensuring the adaptability of the fusion algorithm.

[0104] In one specific embodiment, the process of comparing the second defect feature vector with a preset threshold, determining the defect pixel location and classifying the defect type, and outputting a defect detection report may specifically include the following steps:

[0105] The second defect feature vector is input into the support vector machine classifier for binary classification, and the probability of defect presence at each pixel position is output.

[0106] Pixel locations with a probability of defects exceeding a preset threshold are selected, and defect regions are identified and defect pixel coordinates are determined based on pixel locations. At the same time, the spatial geometric features of each defect region are recorded.

[0107] Extract the RGB channel response difference information from the second defect feature vector corresponding to the defect pixel coordinates and classify the defects to obtain the defect type;

[0108] The severity score is calculated by combining the probability of defect existence with the spatial geometric features of the corresponding defect area. The defect pixel coordinates, defect type determination results and severity score are then integrated to output a defect detection report.

[0109] Specifically, the second defect feature vector is input pixel by pixel into the support vector machine classifier for binary classification. As a discriminative model with good generalization ability, the support vector machine can construct the optimal margin hyperplane in the high-dimensional feature space to distinguish the feature distribution of normal pixels and defective pixels. After the classification model is constructed by introducing the labeled sample set during the training phase, the second defect feature vector corresponding to each pixel is used as the input sample to the classifier during the actual detection phase, and a continuous probability value is output. The probability value represents the probability or confidence that the current pixel belongs to the defect category. The closer the value is to 1, the higher the probability of a defect, while the lower the value, the greater the probability that the pixel is a normal structure. After outputting the defect probability of all image pixels, a threshold judgment process is performed on all output results. A preset defect probability threshold is used as the screening criterion. The positions of all pixels with defect probabilities exceeding the threshold are extracted to form a preliminary set of suspected defect pixels. Connectivity clustering analysis is then performed on the suspected defect pixels in the spatial range. Continuous defect regions are identified through region growing, eight-neighbor merging, or boundary tracking algorithms based on structural connectivity. Each defect region is assigned a unique identifier, and the set of defect pixel coordinates it contains is recorded. At the same time, its corresponding spatial geometric feature information is extracted, including geometric descriptors such as region area, boundary length, principal axis direction, aspect ratio, centroid coordinates, and directional moments, which are used to describe the structural attributes of the defect morphology. After identifying the defective areas, to classify the defect types, the corresponding second defect feature vector is traced back, and RGB channel response difference information is extracted. RGB channel response difference information manifests as an imbalance between the three channel brightness values. For example, a bright spot defect is characterized by a sudden increase in one channel, while a color shift defect is characterized by an abnormally amplified difference between channels. A dark spot is characterized by an overall attenuation of all three channels, but the relative difference remains consistent. Linear defects exhibit spatial directionality and abnormal channel synchronization. Therefore, by comparing the response distribution patterns of the R, G, and B channels with their neighborhood consistency index, a structured identification and classification output of defect types is achieved, categorizing each defective area into categories such as bright spot, dark spot, color shift, or linear defects. After determining the defect type, to reflect the overall impact of the defect on display performance, a severity score is calculated by combining the defect presence probability and spatial geometric features of each defective area. The defect presence probability reflects confidence, and the geometric features reflect the visual impact range. The two are fused using a weighted model to form a standardized severity score. A higher score indicates a greater impact of the defective area on display quality, requiring priority processing or scrapping. The coordinates of all identified defect pixels, their corresponding defect type determination results, severity scores, and spatial distribution information are integrated and summarized to form a defect detection report.

[0110] The above describes the defect detection method for a mobile phone display screen in the embodiments of this application. The following describes the defect detection device for a mobile phone display screen in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of the defect detection device for mobile phone displays in this application includes:

[0111] The acquisition module 21 is used to acquire the display images of the mobile phone screen under test from multiple viewing angles;

[0112] Analysis module 22 is used to analyze the reflection intensity distribution of the displayed image using polarization optical characteristics, and generate a dereflected image after separating the surface reflection components;

[0113] Transformation module 23 is used to transform the pixel brightness of each viewing angle in the dereflection image to a unified reference viewpoint to obtain a unified reference viewpoint image, and extract the first defect feature vector of each pixel in the unified reference viewpoint image.

[0114] Output module 24 is used to fuse the first defect feature vector extracted from multiple observation angles to obtain the second defect feature vector, compare the second defect feature vector with a preset threshold, and output a defect detection report.

[0115] Through the synergistic cooperation of the aforementioned components and the establishment of a viewpoint compensation transformation matrix, this invention can uniformly transform pixel brightness from different viewing angles to a reference viewpoint, eliminating the influence of viewing angle changes on defect detection results. This gives the detection system excellent viewpoint invariance and avoids the inconsistency in detection results caused by different viewing angles in traditional methods. A polarization optics model is used to establish a reflection intensity calculation model, which can accurately separate the surface reflection component and the internal light-emitting component of the display screen, effectively eliminating the interference of ambient light reflection on detection accuracy and enabling the detection system to maintain stable detection performance under complex lighting conditions. Specifically optimized for the Fresnel sub-pixel arrangement structure of mobile phone displays, and through RGB response difference matrix and sub-pixel consistency analysis, higher detection accuracy than traditional pixel-level detection is achieved, enabling the identification of minute display anomalies and color deviations. By calculating the reliability weights of each viewing angle and performing adaptive weighted fusion, the advantages of multi-angle acquisition are fully utilized, improving the reliability and robustness of the detection results and overcoming the limitations of insufficient information from a single viewing angle. A defect classification mechanism based on RGB channel response differences and brightness deviation patterns was established, which can accurately distinguish various defect types such as bright spots, dark spots, color shifts, and line defects, providing detailed defect information for quality control. It does not rely on additional high-precision angle calibration equipment or complex optical compensation systems, reducing system complexity and deployment costs. Viewpoint change compensation technology ensures consistency of detection results under different operating conditions, reducing the impact of human factors and environmental changes on detection quality, and improving the stability and repeatability of the detection system.

[0116] Please see Figure 3 , Figure 3 The present invention provides a schematic block diagram of the structure of an electronic device 300. The electronic device 300 includes a processor 301 and a memory 302, which are connected via a device bus 303. The memory 302 may include a non-volatile storage medium and internal memory.

[0117] The non-volatile storage medium can store a computer program. The computer program includes program instructions, which, when executed by the processor 301, cause the processor 301 to perform any of the aforementioned methods for detecting defects in a mobile phone display screen.

[0118] The processor 301 provides computing and control capabilities to support the operation of the entire electronic device 300.

[0119] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor 301, the processor 301 can execute any of the above-mentioned methods for detecting defects in mobile phone displays.

[0120] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device 300 involved in the present application. The specific electronic device 300 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0121] It should be understood that processor 301 can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0122] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the electronic device 300 described above can be referred to the corresponding process of the aforementioned defect detection method for mobile phone displays, and will not be repeated here.

[0123] This application also provides a computer-readable storage medium storing a computer program that, when executed by one or more processors, causes the one or more processors to implement the mobile phone display defect detection method provided in this application.

[0124] The computer-readable storage medium can be an internal storage unit of the electronic device 300 described in the foregoing embodiments, such as a hard disk or memory of the electronic device 300. The computer-readable storage medium can also be an external storage device of the electronic device 300, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., provided by the electronic device 300.

[0125] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0126] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0127] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for detecting defects in a mobile phone display screen, characterized in that, include: Collect images of the mobile phone screen under test from multiple viewing angles; The reflection intensity distribution of the displayed image is analyzed using polarization optics characteristics, and a dereflection image is generated after separating the surface reflection components. The pixel brightness of the dereflection image at each viewing angle is transformed to a unified reference viewpoint to obtain a unified reference viewpoint image, and the first defect feature vector of each pixel in the unified reference viewpoint image is extracted. Specifically, this includes: analyzing the brightness attenuation characteristics of normal and defective pixels in the dereflection image at different viewing angles, and establishing a pixel brightness change model based on the brightness attenuation characteristics; extracting the brightness gradient information of each pixel position in the dereflection image as a function of the viewing angle, calculating the pixel viewpoint sensitivity coefficient, and using the pixel brightness change model to identify the angular response anomaly characteristics of defective pixels; selecting a reference angle value as the unified reference viewpoint, calculating the brightness compensation coefficient from each viewing angle to the reference angle value according to the pixel brightness change model and the pixel viewpoint sensitivity coefficient, and assembling a viewpoint compensation transformation matrix; and applying the viewpoint compensation transformation matrix to the dereflection image. Perform pixel-by-pixel brightness compensation calculations to generate a unified reference viewpoint image; separate the red, green, and blue color channels of the unified reference viewpoint image to obtain red, green, and blue sub-pixel images respectively; calculate the first-viewpoint response difference parameter of the red sub-pixel image, the second-viewpoint response difference parameter of the green sub-pixel image, and the third-viewpoint response difference parameter of the blue sub-pixel image respectively; construct an RGB response difference matrix based on the first-viewpoint response difference parameter, the second-viewpoint response difference parameter, and the third-viewpoint response difference parameter; for the Fresnel sub-pixel arrangement structure of the mobile phone display screen, analyze the brightness consistency deviation between each pixel and its adjacent sub-pixels using the RGB response difference matrix, and calculate the sub-pixel consistency index; fuse the channel features of the RGB response difference matrix with the sub-pixel consistency index to generate a first defect feature vector; The first defect feature vector extracted from multiple observation angles is fused to obtain the second defect feature vector, and the second defect feature vector is compared with a preset threshold to output a defect detection report.

2. The defect detection method for mobile phone displays according to claim 1, characterized in that, The acquisition of images of the mobile phone screen under test from multiple viewing angles includes: Multiple observation angles are determined according to the preset angle acquisition range; The CCD camera array is controlled to simultaneously acquire RGB three-channel images of the mobile phone display screen from various viewing angles, while recording the corresponding ambient light intensity parameters and display screen brightness parameters. The RGB three-channel images acquired from each observation angle are preprocessed and filtered to generate standardized angle images; Establish a database mapping relationship between the multiple observation angles and the corresponding standardized angle images to form display images under multiple observation angles.

3. The defect detection method for mobile phone displays according to claim 2, characterized in that, The step of analyzing the reflection intensity distribution of the displayed image using polarization optical characteristics, separating the surface reflection components, and generating a dereflected image includes: A first reflection intensity calculation model is established based on the polarization optical characteristics of the mobile phone display screen, including the s-polarization reflection component, the p-polarization reflection component, and the critical angle parameter. Calculate the pixel intensity difference between adjacent viewing angles in the displayed image, extract the intensity change pattern caused by reflection, and correct the polarization component coefficient in the first reflection intensity calculation model based on the intensity change pattern to obtain the second reflection intensity calculation model; The second reflection intensity calculation model is combined with the ambient light intensity parameter to calculate the predicted reflection intensity value of each pixel under different viewing angles. The predicted reflection intensity value is subtracted from the original intensity value of the displayed image pixel by pixel to generate a dereflected image after separating the surface reflection components.

4. The defect detection method for mobile phone displays according to claim 1, characterized in that, The first defect feature vector extracted from multiple observation angles is fused to obtain a second defect feature vector. The second defect feature vector is then compared with a preset threshold, and a defect detection report is output, including: The optimal detection angle is selected as the basis for weight calculation, and the angular distance weight of each observation angle relative to the optimal detection angle is calculated using the exponential decay function. Analyze the statistical distribution characteristics of the first defect feature vector under different observation angles, calculate the standard deviation and mean, and calculate the ratio based on the standard deviation and the mean to obtain the feature quality assessment factor; The reliability weights for each observation angle are formed by combining the angular distance weights with the corresponding feature quality evaluation factors. The reliability weights are used to perform a weighted fusion operation on the first defect feature vectors from each observation angle to output a second defect feature vector. The second defect feature vector is compared with a preset threshold to determine the defect pixel location and classify the defect type, and a defect detection report is output.

5. The defect detection method for mobile phone displays according to claim 4, characterized in that, The process of comparing the second defect feature vector with a preset threshold, determining the defect pixel location and classifying the defect type, and outputting a defect detection report includes: The second defect feature vector is input into a support vector machine classifier for binary classification, and the probability of defect presence at each pixel position is output. Pixel locations with a probability of defect presence exceeding a preset threshold are selected, and defect regions are identified and defect pixel coordinates are determined based on the pixel locations. At the same time, the spatial geometric features of each defect region are recorded. Extract the RGB channel response difference information from the second defect feature vector corresponding to the defect pixel coordinates and perform defect classification to obtain the defect type; The severity score is calculated by combining the probability of the defect's existence with the spatial geometric features of the corresponding defect area. The defect pixel coordinates, defect type determination results, and severity score are then integrated to output a defect detection report.

6. A defect detection device for a mobile phone display screen, characterized in that, For performing the defect detection method for a mobile phone display screen as described in any one of claims 1-5, the defect detection device for the mobile phone display screen includes: The acquisition module is used to acquire images of the mobile phone screen under inspection from multiple viewing angles; The analysis module is used to analyze the reflection intensity distribution of the displayed image using polarization optical characteristics, and generate a dereflected image after separating the surface reflection components; The transformation module is used to transform the pixel brightness of each viewing angle in the dereflection image to a unified reference viewpoint to obtain a unified reference viewpoint image, and extract the first defect feature vector of each pixel in the unified reference viewpoint image. The output module is used to fuse the first defect feature vectors extracted from multiple observation angles to obtain the second defect feature vector, compare the second defect feature vector with a preset threshold, and output a defect detection report.

7. An electronic device, characterized in that, The electronic device includes: a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the electronic device to perform the defect detection method for a mobile phone display screen as described in any one of claims 1-5.

8. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the defect detection method for a mobile phone display screen as described in any one of claims 1-5.

Images