A method and system for detecting defects in a smartphone

Abstract

The application relates to the technical field of smart phone surface defect detection, and discloses a smart phone defect detection method and system, which comprises the following steps: first, collecting defect detection image data through multi-dimensional optical imaging and dividing the image data into multiple detection sub-regions; then, extracting surface defect features based on a texture enhancement technology, combining pixel distribution and spatial correlation to extract defect difference indexes; subsequently, evaluating the regularity of defect modes through multi-scale decomposition analysis, evaluating the defect fluctuation amplitude through time series dynamic analysis, and comprehensively judging the stability of the defect difference indexes; finally, comparing the stable defect difference indexes with reference data to determine the detection result. Through multi-dimensional imaging, multi-scale analysis and dynamic evaluation, the method improves the recognition accuracy and reliability of complex surface defects, is suitable for smart phone surface detection of different materials and structures, and meets the high-quality detection requirements in large-scale production.

Description

Technical Field

[0001] The present invention relates to the technical field of smartphone surface defect detection, and in particular to a smartphone defect detection method and system. Background Art

[0002] With the rapid development of the smartphone industry, consumers are increasingly demanding product appearance quality. Surface defects (such as scratches, dents, color differences, and stains) have become a key factor affecting product qualification rates and user experience. Traditional smartphone defect detection relies primarily on manual visual inspection, which is inefficient, highly subjective, and easily affected by factors such as inspector experience and fatigue. This makes it difficult to meet the high-precision and high-consistency inspection requirements of large-scale production.

[0003] To improve inspection efficiency, some companies have introduced machine vision-based inspection methods, but existing technologies still have many limitations. For example, imaging methods using a single angle or light source cannot fully capture the complex defect characteristics of mobile phone surfaces. This is especially true for products with complex designs, such as curved screens and gradient back covers, which can easily lead to missed or false detections. During the defect feature extraction stage, traditional methods are unable to identify subtle texture anomalies, making it difficult to distinguish normal texture fluctuations from true defects. Furthermore, existing technologies often ignore the dynamic changes of defects over time, making it impossible to assess the stability of defects, resulting in unreliable inspection results.

[0004] Smartphone surfaces are made of a variety of materials (such as glass, metal, ceramic, and composite materials). These materials exhibit significant variations in their reflection and refraction properties, further complicating defect detection. Existing inspection systems are typically customized for specific materials or defect types, lacking versatility and adapting to the demands of rapid product line iterations. With the introduction of new technologies like 5G and foldable screens, the surface structure of mobile phones is becoming increasingly complex, placing higher demands on the accuracy, speed, and adaptability of defect detection. Developing a multi-dimensional, high-precision, stable, and reliable smartphone defect detection method has become an urgent challenge for the industry. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for detecting defects in a smart phone to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting defects in a smartphone, the method comprising:

[0007] Perform multi-dimensional optical imaging on the surface of the smartphone to obtain defect detection image data, and divide the defect detection image data into multiple detection sub-areas;

[0008] The surface defect features of the inspection sub-area are extracted based on texture enhancement technology to highlight abnormal texture information, and defect difference indicators are extracted based on pixel distribution and spatial correlation;

[0009] The complexity of the distribution characteristics of the defect difference index is evaluated through image analysis based on multi-scale decomposition to determine whether the regularity of the surface defect pattern in the local area meets the requirements;

[0010] Through dynamic analysis based on time series, the changing trend of defect difference indicators at different time points is evaluated to determine whether the fluctuation range of surface defects is within the allowable range;

[0011] Based on the regularity of the surface defect pattern in the local area and whether the fluctuation amplitude of the surface defects meets the requirements, the stability of the defect difference index is judged to be qualified;

[0012] When the stability of the defect difference index is qualified, the defect difference index is compared with the defect feature reference data. After the defect difference index meets the preset conditions, the smartphone defect detection result is determined.

[0013] Preferably, performing multi-dimensional optical imaging on the surface of the smartphone to obtain defect detection image data, and dividing the defect detection image data into multiple detection sub-areas, specifically includes:

[0014] Deploy multiple optical imaging devices around the surface of the smartphone and configure multiple sets of lighting sources. By adjusting the focal length and shooting parameters of the optical imaging devices, multi-dimensional defect detection image data can be recorded.

[0015] The multi-dimensional defect detection image data is stored and preliminarily processed to form a pre-processed image set, and the pre-processed image set is decomposed into multiple detection sub-areas according to pre-set division rules.

[0016] Preferably, the surface defect features of the detection sub-area are extracted based on texture enhancement technology to highlight abnormal texture information, and defect difference indicators are extracted based on pixel distribution and spatial correlation, specifically including:

[0017] Texture enhancement technology is applied to the defect detection image data within the detection sub-area. Abnormal texture features are enhanced by selecting structural units of different sizes. Structural units of different sizes are used to match abnormal textures of different thicknesses and directions.

[0018] Quantitative analysis of surface defect characteristics in the detection sub-area is performed based on pixel distribution parameters, which are obtained by calculating the grayscale value distribution characteristics of the local area;

[0019] Analyze the spatial correlation of surface defect features and extract the correlation feature parameters of surface defects based on the positional relationship between local pixels and their adjacent pixels;

[0020] The defect difference index is extracted by combining the pixel distribution parameters and the associated feature parameters. The defect difference index is used to represent the microscopic difference characteristics of surface defects.

[0021] Preferably, the complexity evaluation of the distribution characteristics of the defect difference index is performed by image analysis based on multi-scale decomposition to determine whether the regularity of the surface defect pattern in the local area meets the requirements, specifically including:

[0022] Generate a two-dimensional distribution matrix based on the spatial distribution of defect difference indicators calculated in the detection sub-area;

[0023] Decompose the two-dimensional distribution matrix into multiple hierarchical components based on the multi-scale decomposition model;

[0024] Calculate the analysis dimension corresponding to each level component to measure the complexity of surface defects at different scales in the two-dimensional distribution matrix;

[0025] According to the distribution law of analysis dimensions at different levels, the complexity characteristic parameters representing the regularity of surface defects are extracted.

[0026] Complexity characteristic parameters include decomposition range width and decomposition intensity level;

[0027] Compare the complexity characteristic parameters with the preset requirements to determine whether the regularity of the surface defect pattern in the local area meets the requirements.

[0028] Preferably, the changing trend of the defect difference index at different time points is evaluated by dynamic analysis based on time series to determine whether the fluctuation amplitude of the surface defects is within the allowable range, specifically including:

[0029] Arrange the defect difference indicators at each time point in sequence to form a defect difference indicator time series;

[0030] The time mapping method is used to transform the defect difference index time series into a multi-dimensional state vector set;

[0031] Dynamic analysis methods are applied to the multidimensional state vector set to extract fluctuation-related characteristic parameters, which are used to measure the variation pattern of surface defects in the time dimension.

[0032] Compare the fluctuation-related characteristic parameters with the preset range to determine whether the fluctuation amplitude of the surface defect is within the allowable range.

[0033] Preferably, a dynamic analysis method is applied to the multidimensional state vector set to extract fluctuation-related characteristic parameters, specifically:

[0034] The spatial distance between the state vectors corresponding to adjacent time points is calculated as the local fluctuation distance, and the spatial distance is obtained by taking the square root of the sum of the differences of the components in each dimension.

[0035] Preferably, based on the regularity of the surface defect pattern in the local area and whether the fluctuation amplitude of the surface defects meets the requirements, the stability of the defect difference index is judged to be qualified, specifically including:

[0036] When the regularity of the surface defect pattern in the local area meets the requirements and the fluctuation amplitude of the surface defects is within the allowable range, the stability of the defect difference index is judged to be qualified; otherwise, the stability of the defect difference index is judged to be unqualified.

[0037] Preferably, when the stability of the defect difference index is qualified, the defect difference index is compared with the defect feature reference data. After the defect difference index meets the preset conditions, the smartphone defect detection result is determined, specifically including:

[0038] Select corresponding reference data from a preset defect feature reference database according to the attributes of the detection sub-area;

[0039] Compare the defect difference indicators with the selected reference data item by item to determine whether the preset conditions are met;

[0040] When all comparison items between the defect difference index and the reference data meet the preset conditions, the smartphone defect detection results are generated.

[0041] Preferably, the present invention further includes a smartphone defect detection system for implementing the above-mentioned smartphone defect detection method, the system comprising a surface imaging acquisition module, a defect feature extraction module, a regularity complexity assessment module, a fluctuation amplitude analysis module, a stability comprehensive judgment module, and a defect result generation module;

[0042] Surface imaging acquisition module: This module performs multi-dimensional optical imaging on the smartphone surface to obtain defect detection image data, and divides the defect detection image data into multiple detection sub-areas;

[0043] Defect feature extraction module: This module extracts surface defect features of the detection sub-area based on texture enhancement technology to highlight abnormal texture information, and extracts defect difference indicators based on pixel distribution and spatial correlation;

[0044] Regularity Complexity Assessment Module: This module evaluates the complexity of the distribution characteristics of defect difference indicators through image analysis based on multi-scale decomposition to determine whether the regularity of the surface defect pattern in the local area meets the requirements;

[0045] fluctuation range analysis module: the fluctuation range of the surface defects is judged whether in the allowable range through the evaluation of the change trend of the defect difference index at different time points based on time series dynamic analysis;

[0046] stable comprehensive judgment module: based on the regularity of the surface defect mode in the local area and the evaluation result of whether the fluctuation range of the surface defects meets the requirements, the stability of the defect difference index is judged whether qualified;

[0047] defect result generation module: when the stability of the defect difference index is qualified, the defect difference index is compared with the defect feature reference data, and the smart phone defect detection result is determined when the defect difference index meets the preset condition.

[0048] Preferably, the surface imaging acquisition module is specifically used for: deploying multiple optical imaging devices around the surface of the smart phone and configuring multiple groups of illumination light sources, recording multi-dimensional defect detection image data by adjusting the focal length and shooting parameters of the optical imaging device; the multi-dimensional defect detection image data is stored and preliminarily processed to form a preprocessed image set, and the preprocessed image set is decomposed into multiple detection sub-regions according to a pre-set division rule.

[0049] Compared with the prior art, the beneficial effects of the present application are:

[0050] By combining multi-dimensional optical imaging acquisition and multi-scale analysis, the accuracy and reliability of defect detection are significantly improved, and the problems of missed detection, false detection and poor adaptability in traditional detection methods are effectively solved.

[0051] In the imaging stage, by deploying multiple optical imaging devices and multiple groups of illumination light sources, combined with dynamic adjustment of focal length and shooting parameters, the defect features of the surface of the smart phone under different light conditions can be fully captured, especially suitable for detection of curved surfaces, special-shaped structures and complex material surfaces, avoiding information loss caused by single imaging angle. The image data is divided into multiple detection sub-regions, realizing fine analysis of local areas and improving the recognition ability of subtle defects.

[0052] In the feature extraction link, texture enhancement technology is adopted combined with structure units of different sizes, which can highlight different abnormal textures in terms of thickness and direction, and through the fusion of pixel distribution parameters and spatial correlation features, the defect difference index constructed can accurately quantify the microscopic characteristics of surface defects, effectively distinguish normal texture fluctuation from real defects, and reduce the interference information in the feature extraction stage.

[0053] Through multi-scale decomposition of image analysis, the regularity of defect patterns at different spatial scales can be evaluated. Combined with the dynamic analysis of time series, the fluctuation trend of defects in the time dimension can be captured. The stability assessment mechanism composed of the two ensures the consistency and reliability of the detection results and avoids misjudgment due to instantaneous interference.

[0054] By comparing with a reference database of defect signatures, this method unifies and standardizes testing standards, adapting to the detection needs of different materials and defect types, and improving the system's versatility and scalability. While ensuring detection accuracy, this method can meet the needs of efficient testing in large-scale production, reduce labor costs, and improve product qualification rates, providing strong technical support for quality control in the smartphone manufacturing industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a diagram showing the working principle of the smartphone defect detection method according to the present invention;

[0056] Figure 2 Flowchart for surface defect feature extraction and difference index generation;

[0057] Figure 3 Flowchart for multi-scale decomposition and regularity assessment;

[0058] Figure 4 Flowchart for time series dynamic analysis and fluctuation range assessment;

[0059] Figure 5 This is the main flow chart of the smartphone defect detection system. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0061] See also Figure 1-Figure 5 The present invention relates to a method for detecting defects in a smart phone, and the specific implementation steps are as follows:

[0062] Multi-dimensional optical imaging is performed on the smartphone surface to acquire defect detection image data, which is then divided into multiple detection sub-areas. Multiple optical imaging devices are deployed around the smartphone surface, along with multiple sets of illumination light sources. Multi-dimensional defect detection image data is recorded by adjusting the focal length and shooting parameters of the optical imaging devices. The multi-dimensional defect detection image data is stored and preliminarily processed to form a pre-processed image set, which is then decomposed into multiple detection sub-areas according to pre-set division rules.

[0063] The surface defect features of the detection sub-area are extracted based on texture enhancement technology to highlight abnormal texture information, and defect difference indicators are extracted based on pixel distribution and spatial correlation. Texture enhancement technology is applied to the defect detection image data within the detection sub-area. Abnormal texture features are enhanced by selecting structural units of different sizes. Structural units of different sizes are used to match abnormal textures of different thicknesses and directions. Surface defect features in the detection sub-area are quantitatively analyzed based on pixel distribution parameters, which are obtained by calculating the grayscale value distribution characteristics of the local area. The spatial correlation of surface defect features is analyzed, and the associated feature parameters of the surface defects are extracted based on the positional relationship between the local pixel and its adjacent pixels. The defect difference index is extracted by combining the pixel distribution parameters and the associated feature parameters. The defect difference index is used to represent the microscopic difference characteristics of the surface defects.

[0064] Through image analysis based on multi-scale decomposition, the complexity of the distribution characteristics of defect difference indicators is evaluated to determine whether the regularity of the surface defect pattern in the local area meets the requirements. The defect difference indicators calculated in the detection sub-area are distributed according to the spatial position to generate a two-dimensional distribution matrix; based on the multi-scale decomposition model, the two-dimensional distribution matrix is ​​decomposed into multiple hierarchical components; the analysis dimension corresponding to each hierarchical component is calculated to measure the complexity of surface defects at different scales in the two-dimensional distribution matrix; based on the distribution pattern of the analysis dimension at different levels, the complexity characteristic parameters representing the regularity of surface defects are extracted, and the complexity characteristic parameters include the decomposition range width and the decomposition intensity level; the complexity characteristic parameters are compared with the preset requirements to determine whether the regularity of the surface defect pattern in the local area meets the requirements.

[0065] Through dynamic analysis based on time series, the changing trends of defect difference indicators at different time points are evaluated to determine whether the fluctuation amplitude of surface defects is within the allowable range. The defect difference indicators at each time point are arranged in sequence to form a defect difference indicator time series; the defect difference indicator time series is converted into a multidimensional state vector set using a time mapping method; the dynamic analysis method is applied to the multidimensional state vector set to extract fluctuation-related characteristic parameters, which are used to measure the changing patterns of surface defects in the time dimension; the fluctuation-related characteristic parameters are compared with a preset range to determine whether the fluctuation amplitude of surface defects is within the allowable range. Specifically, the dynamic analysis method is applied to the multidimensional state vector set to extract fluctuation-related characteristic parameters. Specifically, the spatial distance between the corresponding state vectors at adjacent time points is calculated as the local fluctuation distance. The spatial distance is calculated by taking the square root of the sum of the squares of the differences of the components in each dimension.

[0066] Based on the evaluation results of whether the regularity of the surface defect pattern in the local area and the fluctuation amplitude of the surface defects meet the requirements, the stability of the defect difference index is judged to be qualified. When the regularity of the surface defect pattern in the local area meets the requirements and the fluctuation amplitude of the surface defects is within the allowable range, the stability of the defect difference index is judged to be qualified; otherwise, the stability of the defect difference index is judged to be unqualified.

[0067] When the stability of the defect difference index is qualified, the defect difference index is compared with the defect feature reference data. If the defect difference index meets the preset conditions, the smartphone defect detection result is determined. Based on the attributes of the detection sub-area, corresponding reference data is selected from the preset defect feature reference database. The defect difference index is compared with the selected reference data item by item to determine whether the preset conditions are met. When all comparison items between the defect difference index and the reference data meet the preset conditions, the smartphone defect detection result is generated.

[0068] Example 1: In this example, in the process of performing multi-dimensional optical imaging on the surface of a smartphone to obtain defect detection image data and dividing the defect detection image data into multiple detection sub-areas, the specific operations need to be carried out from multiple links such as equipment deployment, parameter adjustment, and data processing.

[0069] Multiple optical imaging devices are deployed around the smartphone surface, covering different aspects of the phone. For example, these devices should be mounted directly above, to the left, to the right, diagonally above, and diagonally below the phone. This allows for imaging of the phone's surface from multiple angles, ensuring that no areas with potential defects are missed. Imaging devices at different angles can capture the characteristics of the phone's surface from different perspectives. For directional defects such as scratches and wear, capturing them from multiple angles provides a more comprehensive picture.

[0070] Configuring multiple sets of lighting sources is also a crucial part of this step. These sets of lighting sources require different lighting conditions, including varying intensities, angles, and spectral characteristics. For example, one set of light sources can illuminate the phone surface directly with white light. This lighting condition is suitable for inspecting the overall flatness and color uniformity of the phone surface. Another set of light sources can illuminate with sidelight, which better highlights surface defects such as small scratches and pits, as sidelight casts shadows on the defects, making them more noticeable. Alternatively, a set of light sources can be configured that illuminate with monochromatic light of different wavelengths. Different wavelengths of light have different reflection and absorption characteristics for different materials on the phone surface, which helps detect defects on specific surfaces. By using different lighting conditions, various possible defects on the phone surface can be highlighted from multiple perspectives, improving the defect detection rate.

[0071] After deploying the optical imaging equipment and lighting source, the focal length of the optical imaging equipment needs to be adjusted. The purpose of adjusting the focal length is to enable the equipment to clearly capture different areas of the phone's surface. Since the phone's surface may have varying heights, such as the interface between the screen and the frame and the raised area of ​​the camera module, adjusting the focal length ensures that these areas are clearly captured in the image. Shooting parameters, including exposure time, aperture, and ISO sensitivity, also need to be adjusted. Exposure time affects image brightness: a long exposure time can result in an image that is too bright, while a short exposure time can result in an image that is too dark. Aperture size affects depth of field and the amount of light entering the image. A larger aperture creates a smaller depth of field, highlighting the subject but potentially blurring other areas, while a smaller aperture sharpens the entire image. ISO sensitivity affects image noise: a high ISO value can increase image noise, affecting image quality. Properly adjusting these shooting parameters ensures the acquisition of high-quality defect detection image data, providing a reliable foundation for subsequent defect detection.

[0072] After acquiring multi-dimensional defect inspection image data, it needs to be stored. This can be done on a computer hard drive, solid-state drive, or other specialized storage device to ensure secure and stable data storage for subsequent processing and analysis. The data needs to be categorized and named during storage to facilitate subsequent search and retrieval.

[0073] Perform preliminary processing on the stored data, including image denoising, brightness adjustment, and contrast enhancement. Image denoising aims to remove noise generated during the capture process due to environmental interference, equipment characteristics, and other factors, thereby enhancing image clarity. Common denoising methods include median filtering and Gaussian filtering. Brightness adjustment and contrast enhancement can make image defects more visible, facilitating subsequent feature extraction. For example, for images with uneven brightness, brightness adjustment can make the brightness of the entire image more uniform; contrast enhancement can make the difference between defects and normal areas more pronounced.

[0074] After processing is complete, a preprocessed image set is formed, which is then decomposed into multiple detection sub-areas according to pre-set division rules. The division rules can be set according to the structural characteristics of the phone surface. For example, according to the different functional areas of the phone surface, the screen area, frame area, back area, camera module area, etc. are divided into different detection sub-areas. Each detection sub-area corresponds to a specific part of the phone surface, which allows for more detailed defect detection based on the characteristics of different areas. For example, the screen area mainly detects defects such as scratches, bright spots, and dark spots, the frame area mainly detects defects such as wear and deformation, and the back area mainly detects defects such as stains and paint peeling. By decomposing the image into multiple detection sub-areas, the accuracy and efficiency of defect detection can be improved, making the detection process more targeted.

[0075] Throughout the entire operation, each step must strictly adhere to established procedures and parameters to ensure accurate and reliable data capture and effective, rational, and effective inspection sub-areas. The placement of optical imaging equipment and the configuration of lighting sources require multiple tests and adjustments to find the most suitable combination. Shooting parameters must be optimized based on factors such as the phone's material and color. Data storage and processing must adhere to specific specifications to ensure data integrity and traceability. The rules for demarcating inspection sub-areas must fully consider the phone's surface structural characteristics and defect distribution to ensure comprehensive and detailed inspection of each sub-area.

[0076] Example 2: This example extracts surface defect features of the detection sub-area based on texture enhancement technology to highlight abnormal texture information, and extracts defect difference indicators based on pixel distribution and spatial correlation. Specific operations need to be carried out from multiple technical links.

[0077] When applying texture enhancement technology to defect detection image data within the detection sub-area, the selection of structural units is crucial. In actual operation, structural units of different sizes need to be selected to match abnormal textures of varying thickness on the surface of the mobile phone. For example, a small-sized structural unit, whose side length may be set to 3×3 pixels or 5×5 pixels, is suitable for capturing fine abnormal textures such as small scratches and point-like impurities on the mobile phone screen glass. Because the small-sized structural unit has a smaller range of action, it can accurately locate subtle texture changes. Large-sized structural units, such as 9×9 pixels or 11×11 pixels, are used to match coarser abnormal textures such as deeper scratches and larger casting defects on the metal frame of the fuselage. The larger range of action can cover the overall shape of such defects.

[0078] The orientation of the structural units also needs to consider the direction of different abnormal textures. For example, horizontal structural units can be used to enhance the horizontal scratch features on the mobile phone screen, vertical structural units are suitable for highlighting the vertical wear marks on the frame, and structural units with a 45-degree or 135-degree orientation can better capture obliquely distributed texture anomalies. By combining structural units of different sizes and orientations, morphological operations such as dilation, erosion, opening or closing operations are performed on the image of the detection sub-area, thereby achieving directional enhancement of various abnormal texture features, making the defective area more prominent in the image.

[0079] After texture enhancement, the surface defect characteristics within the inspection subregion need to be quantitatively analyzed based on pixel distribution parameters. These parameters are obtained by calculating the grayscale value distribution characteristics of the local region. Specifically, several local windows are defined within the inspection subregion. The size of each window can be set based on the subregion size and defect type. For example, for a screen subregion, the local window can be set to 16×16 pixels or 32×32 pixels. Then, for each local window, the average grayscale value is calculated. This value reflects the overall brightness level of the pixels within the window. If a defect is present, the average grayscale value may differ from that of a normal area. The variance of the grayscale values ​​is calculated. The variance measures the dispersion of the pixel grayscale values ​​relative to the average. The grayscale distribution of pixels in defective areas is typically more dispersed and has a larger variance. The standard deviation, similar to the variance, further quantifies the range of grayscale value fluctuations. Furthermore, the grayscale histogram distribution of the local window is statistically analyzed. The histogram can visually display the distribution of the number of pixels of different grayscale values. The histogram of a normal area may exhibit a relatively concentrated distribution, while the histogram of a defective area may exhibit peak shifts or multiple peaks.

[0080] By calculating these pixel distribution parameters, the surface defect characteristics can be converted into specific numerical indicators. For example, if the grayscale average value of a local window is lower than the normal area threshold, or the variance exceeds the set range, it can be preliminarily determined that there may be defects in the area.

[0081] The spatial correlation of surface defect features is analyzed. This process extracts correlation feature parameters based on the positional relationship between local pixels and their adjacent pixels. In an image, each pixel has a positional correlation with its eight surrounding neighbors (i.e., pixels in the top, bottom, left, right, and four diagonal directions). In actual analysis, the grayscale differences between the local pixel and its adjacent pixels in the horizontal, vertical, and diagonal directions need to be calculated. For example, for a central pixel, the grayscale difference between it and the adjacent pixel to the right, the grayscale difference with the adjacent pixel above, and the grayscale difference with the adjacent pixel on the upper right diagonal are calculated. These grayscale differences can reflect the spatial variation trend of the pixels. The grayscale differences between adjacent pixels in normal areas are usually small and have a certain regularity, while the grayscale differences between adjacent pixels in defective areas may be large and irregular.

[0082] Correlation parameters between adjacent pixels, such as covariance or correlation coefficient, are also calculated to measure the degree of correlation between their grayscale values. Regions with high correlation indicate relatively gradual grayscale changes, potentially reflecting normal surface texture, while regions with low correlation may contain abrupt texture changes, indicating defects. By comprehensively analyzing these positional relationships and grayscale differences, characteristic parameters associated with surface defects, such as average grayscale difference, maximum grayscale difference, and correlation coefficient matrix, can be extracted.

[0083] Defect difference indicators are extracted by combining pixel distribution parameters with associated feature parameters. Specifically, a multi-dimensional feature fusion model is established to weight pixel distribution parameters (such as grayscale mean, variance, standard deviation, and histogram features) and associated feature parameters (such as mean grayscale difference and correlation coefficient). Weightings are determined based on the characteristics of different defect types and detection requirements. For example, for scratches, the weight of associated feature parameters can be increased due to their strong spatial continuity. For point defects, the weight of pixel distribution parameters can be increased due to their more prominent grayscale characteristics.

[0084] The fused defect difference index can fully reflect the microscopic differences in surface defects. For example, the defect difference index for a particular inspection subregion may show a grayscale variance higher than normal, while the average grayscale difference between adjacent pixels also exceeds the threshold. This indicates that not only is the pixel grayscale distribution in this region discrete, but the spatial texture also exhibits abnormal variations, indicating a high probability of defects. This index can be used to convert qualitative descriptions of surface defects into quantitative numerical expressions, providing a precise characteristic basis for subsequent defect pattern regularity assessment and fluctuation amplitude analysis.

[0085] Example 3: This example uses image analysis based on multi-scale decomposition to perform complexity evaluation on the distribution characteristics of defect difference indicators to determine whether the regularity of the surface defect pattern in the local area meets the requirements. The specific implementation needs to start from the data matrix construction, multi-scale decomposition, dimension calculation, feature extraction and comparison judgment.

[0086] After calculating the defect difference indicators within the detection sub-region, it is necessary to generate a two-dimensional distribution matrix based on the distribution of these indicators in space. In practice, the pixel coordinates of the detection sub-region are used as the reference, and each pixel corresponds to an element in the matrix, with the element value being the defect difference indicator value at that location. For example, if the detection sub-region is a 100×100 pixel image area, a 100×100 two-dimensional matrix is ​​generated, and the value of each position (i, j) in the matrix corresponds to the defect difference indicator of the pixel in the i-th row and j-th column in the image. This intuitively presents the spatial distribution of the defect difference indicators, allowing subsequent analysis to be performed based on quantized matrix data.

[0087] The two-dimensional distribution matrix is ​​decomposed based on the multi-scale decomposition model. The multi-scale decomposition model can adopt a wavelet decomposition model, which can decompose the matrix into hierarchical components at different scales. In the specific implementation, the first layer of wavelet decomposition is first performed on the two-dimensional matrix to obtain approximate components and detail components. The approximate components reflect the low-frequency characteristics of the matrix, that is, the overall distribution trend, and the detail components reflect the high-frequency characteristics, that is, the local detail changes; then the approximate components continue to be decomposed to the next layer, and so on until the preset number of decomposition layers is reached. For example, if the number of decomposition layers is set to 3 layers, the two-dimensional matrix will be decomposed into 3 levels of approximate components and corresponding detail components. Each level of components corresponds to a different spatial scale, such as the first layer of decomposition corresponds to a larger scale, and the third layer of decomposition corresponds to a smaller scale. Components of different scales can capture the distribution characteristics of defects of different scales. Large-scale components are suitable for analyzing the defect distribution trend of large areas, while small-scale components are used to analyze subtle local changes in defects.

[0088] After completing the multiscale decomposition, it is necessary to calculate the analysis dimension corresponding to each hierarchical component to measure the complexity of surface defects at different scales in the two-dimensional distribution matrix. The analysis dimension can be calculated using the fractal dimension method, which can quantify the spatial filling degree of complex geometric shapes. For each hierarchical component, it is treated as a two-dimensional data set, and its fractal dimension is calculated using the grid covering method or the box dimension method. Specifically, the component data is covered with grids of different sizes, the number of data points contained in each grid is counted, and the fractal dimension is then calculated based on the relationship between grid size and number of data points. The larger the fractal dimension value, the more complex the surface defect distribution at that scale and the lower the regularity; conversely, the smaller the fractal dimension, the more regular the distribution and the higher the regularity. For example, a hierarchical component with a fractal dimension of 2.5 corresponds to a more complex surface defect distribution than a hierarchical component with a fractal dimension of 1.8.

[0089] Based on the distribution patterns of the analysis dimensions at different levels, the complexity characteristic parameters representing the regularity of surface defects are extracted. These parameters mainly include the decomposition range width and the decomposition intensity level. The decomposition range width refers to the range of variation of the analysis dimensions at different levels, which is obtained by calculating the difference between the maximum and minimum values ​​of the fractal dimension at each level. For example, if the fractal dimensions of the three levels are 1.5, 2.0, and 2.3, respectively, the decomposition range width is 2.3-1.5=0.8, which reflects the difference in the complexity of surface defects at different scales. The decomposition intensity level is determined based on the overall distribution of the analysis dimensions at each level. It can be divided into levels by calculating the average value of the fractal dimension at each level and comparing it with the preset intensity level threshold. For example, an average value of 1.8 may correspond to a lower decomposition intensity level, indicating that the regularity of the surface defect distribution is strong; an average value of 2.2 may correspond to a higher decomposition intensity level, indicating that the regularity is weak.

[0090] After the complexity feature parameters are extracted, they need to be compared with the preset requirements to determine whether the regularity of the surface defect pattern in the local area meets the requirements. The preset requirements clearly stipulate the qualified range of the complexity feature parameters, which is obtained through a large number of experimental statistics based on the normal texture characteristics and common defect patterns on the surface of smartphones. For example, the preset decomposition range width should not exceed 1.0, and the decomposition intensity level should not be higher than a certain level. During the comparison process, if the decomposition range width is within the preset qualified range and the decomposition intensity level also meets the preset requirements, it is determined that the regularity of the surface defect pattern in the local area meets the requirements; if any of the parameters exceeds the preset range, it is determined that the regularity does not meet the requirements.

[0091] Throughout the implementation process, the construction of the two-dimensional distribution matrix must ensure accurate spatial alignment of the defect difference indicators. The number of layers and model selection for the multi-scale decomposition must be optimized based on the size of the detection sub-area and the defect type. The calculation method for the analysis dimension must ensure rigorous mathematical logic. The extraction rules for the complexity feature parameters must form a clear mapping relationship with the judgment criteria of the preset requirements. For example, for the detection sub-area of ​​the mobile phone screen glass, due to its relatively uniform normal texture, the preset requirements may set a smaller decomposition range width and a lower decomposition intensity level. However, for the detection sub-area of ​​the mobile phone metal frame, due to the possibility of certain processing lines in its normal texture, the preset requirements can be appropriately relaxed.

[0092] Example 4: This example evaluates the changing trend of the defect difference index at different time points through dynamic analysis based on time series to determine whether the fluctuation amplitude of the surface defects is within the allowable range. The specific implementation needs to be carried out from the aspects of time series construction, state vector conversion, fluctuation feature extraction and comparison judgment in combination with specific examples. For example, in the inspection of a certain model of smartphone screen glass, it is set to collect the defect difference index of the inspection sub-area every 5 minutes, and continue to collect data for 1 hour, and obtain data at a total of 12 time points. These data are arranged in chronological order to form a defect difference index time series: [x1, x2, x3, ..., x12], where x1 corresponds to the index value of the first time point, and x12 corresponds to the index value of the last time point. The construction of this time series needs to ensure the consistency of the time interval and the continuity of data collection to accurately reflect the changes in the defect difference index in the time dimension.

[0093] After obtaining the time series, it is converted into a multidimensional set of state vectors using a time mapping method. Taking the phase space reconstruction method as an example, assuming an embedding dimension of 3 and a time delay of 1, then for the time series [x1, x2, x3, …, x12], the first state vector is (x1, x2, x3), the second state vector is (x2, x3, x4), and so on, until the last state vector is (x10, x11, x12). Ultimately, 10 three-dimensional state vectors are obtained, forming a multidimensional set of state vectors. The choice of embedding dimension and time delay depends on the characteristics of the time series. For example, for a series with relatively gentle fluctuations, a larger embedding dimension and time delay can be selected to capture longer-term trends; for a series with more volatile fluctuations, smaller parameters can be selected to focus on short-term changes.

[0094] The dynamic analysis method is applied to the multi-dimensional state vector set to extract the fluctuation-related characteristic parameters, specifically calculating the spatial distance between the state vectors corresponding to adjacent time points as the local fluctuation distance. and the second state vector For example, the spatial distance is calculated by taking the square root of the sum of the differences of the components of each dimension: First calculate and The difference, and The difference, and The difference between the three differences is squared and added together to get , and then take the square root of the sum, and the result is the local fluctuation distance between the two adjacent state vectors. According to this method, the spatial distance of each pair of adjacent state vectors is calculated in turn, thereby obtaining 9 local fluctuation distance values, forming a fluctuation distance sequence: ,in is the distance between the first and second state vectors, is the distance between the ninth and tenth state vectors.

[0095] After extracting the fluctuation distance sequence, it is necessary to perform statistical analysis on these data to obtain characteristic parameters that can measure the changing patterns of surface defects in the time dimension. For example, calculate the average value of the fluctuation distance, which reflects the average fluctuation amplitude of the defect difference index in the time series; calculate the maximum and minimum values ​​of the fluctuation distance, the maximum value represents the largest single fluctuation amplitude, and the minimum value represents the smallest single fluctuation amplitude; calculate the standard deviation of the fluctuation distance, which is used to measure the degree of dispersion of the fluctuation amplitude. The larger the standard deviation, the more severe the fluctuation, and vice versa. Assuming that in the above example, the 9 local fluctuation distance values ​​are 0.2, 0.3, 0.1, 0.4, 0.3, 0.2, 0.5, 0.3, and 0.2, respectively, then the average value is , the maximum value is 0.5, the minimum value is 0.1, and the standard deviation can be calculated by the corresponding formula. These parameters together describe the fluctuation characteristics of the defect difference index of the detection sub-area.

[0096] The extracted fluctuation-related characteristic parameters are compared with preset ranges to determine whether the fluctuation amplitude of the surface defect is within the allowable range. The preset range is determined based on the characteristics of the smartphone surface material and production process standards and may vary for different detection sub-areas. For example, for the phone screen glass detection sub-area, the preset average fluctuation distance should not exceed 0.35, the maximum fluctuation distance should not exceed 0.6, and the standard deviation should not exceed 0.15. For the phone metal frame detection sub-area, due to the material characteristics and texture fluctuations that may occur during normal processing, the preset ranges may be relaxed appropriately, such as the average fluctuation distance not exceeding 0.45, the maximum fluctuation distance not exceeding 0.7, and the standard deviation not exceeding 0.2. In this example, the average fluctuation distance of 0.28 is less than 0.35, the maximum fluctuation distance of 0.5 is less than 0.6, and the standard deviation is calculated to be 0.12 less than 0.15. Therefore, if all three characteristic parameters are within the preset range, the fluctuation amplitude of the surface defect in this detection sub-area can be determined to be within the allowable range. If any of these parameters exceeds the preset range, such as the maximum fluctuation distance of 0.7, the fluctuation amplitude is determined to be outside the allowable range.

[0097] Throughout the implementation process, the frequency of time series acquisition should be determined based on the actual needs of defect detection. For scenarios requiring rapid detection, the interval can be shortened, such as every 1 minute; for stability testing, the interval can be extended. The parameters for phase space reconstruction should be optimized through preliminary experiments to ensure that the converted state vector accurately reflects the dynamic characteristics of the time series. The calculation of local fluctuation distances must strictly adhere to the definition of spatial distance to ensure numerical accuracy. The setting of preset ranges requires reference to a large amount of inspection data from both normal and defective products, and a reasonable threshold value determined through statistical analysis to ensure the reliability of the judgment results. For example, when setting the preset range for the screen glass inspection sub-area, the same time dynamic analysis can be performed on 1,000 normal screen samples to calculate the distribution of their fluctuation-related characteristic parameters. The parameter range of 95% of normal samples can be used as the preset allowable range. This ensures accurate judgment of normal products while effectively identifying defective products.

[0098] Example 5: This example determines whether the stability of the defect difference index is qualified based on the evaluation results of the regularity of the surface defect pattern in the local area and whether the fluctuation amplitude of the surface defects meets the requirements. When the stability of the defect difference index is qualified, the defect difference index is compared with the defect feature reference data to determine the smartphone defect detection result. The specific implementation needs to be carried out in combination with specific examples from two aspects: stability judgment and result generation.

[0099] Taking the detection sub-region of the back glass of a smartphone as an example, after completing the complexity evaluation of multi-scale decomposition and the dynamic analysis of the time series, the regularity evaluation results of the surface defect pattern in the local area and the fluctuation amplitude evaluation results of the surface defects are obtained. Assuming that through the multi-scale decomposition analysis of Example 3, the complexity characteristic parameters of the detection sub-region are obtained as the decomposition range width of 0.7 and the decomposition intensity level of level 2, while the preset requirements for the decomposition range width qualification upper limit is 1.0 and the decomposition intensity level qualification upper limit is level 3, then the decomposition range width of 0.7 is less than 1.0, and the decomposition intensity level of level 2 is lower than level 3, which indicates that the regularity of the surface defect pattern in the local area meets the requirements. At the same time, through the dynamic analysis of Example 4, the fluctuation-related characteristic parameters of the detection sub-region are obtained as the average fluctuation distance of 0.25, the maximum fluctuation distance of 0.45, and the standard deviation of 0.1. In the preset allowable range, the average fluctuation distance does not exceed 0.3, the maximum fluctuation distance does not exceed 0.5, and the standard deviation does not exceed 0.12. These three parameters are all within the preset range, indicating that the fluctuation amplitude of the surface defects is within the allowable range. According to the results of these two evaluations, the detection sub-area simultaneously meets the conditions that the regularity of the surface defect pattern meets the requirements and the fluctuation amplitude is within the allowable range. Therefore, the stability of the defect difference index is judged to be qualified.

[0100] If the decomposition range width of the complexity characteristic parameter of the detection sub-area is 1.2, which exceeds the preset 1.0, or the maximum fluctuation distance of the fluctuation amplitude is 0.6, which exceeds the preset 0.5, as long as any one of the conditions is not met, the stability of the defect difference index is judged to be unqualified.

[0101] When the stability of the defect difference index is qualified, it is necessary to compare the defect difference index with the defect feature reference data to determine the detection result. For example, the detection sub-area is the back glass area of ​​a smartphone, and its attributes are glass material and smooth surface. Based on this attribute, the corresponding reference data is selected from the preset defect feature reference database. The reference database stores the defect feature data of normal samples with different materials and different surface features. For the back glass area, the reference data may include indicator ranges in multiple dimensions, such as the grayscale average range [180,220], variance range [10,30], and average grayscale difference range [5,15].

[0102] The defect difference indicators extracted from the inspection sub-region are compared item by item with the selected reference data. Assuming that the grayscale average value of the defect difference indicator is 200, within the range [180, 220]; the variance is 25, within the range [10, 30]; and the average grayscale difference is 10, within the range [5, 15], and all comparison items meet the preset conditions, a smartphone defect detection result is generated, indicating that the back glass inspection sub-region has no defects or the defects are within the allowable range.

[0103] If any of the defect difference indicators does not meet the preset conditions, such as the grayscale average value is 170, which is lower than the lower limit of the reference data range of 180, or the variance is 35, which exceeds the upper limit of the reference data range of 30, as long as any comparison item does not meet the preset conditions, the corresponding defect detection result will be generated, indicating that there is a defect in the detection sub-area.

[0104] In practice, the construction of a reference database requires extensive test data from various areas of a normal smartphone. For example, for the back glass area, data from 1,000 normal smartphones is collected from the back glass inspection sub-areas. The mean and standard deviation of each defect difference indicator are calculated, and the range of the mean ± 2 times the standard deviation is used as the appropriate reference data interval. The database also requires regular maintenance and updates. As production processes improve and new materials are introduced, the reference data is adjusted to ensure its accuracy and applicability.

[0105] During the comparison process, the weights of various indicators can vary depending on the degree of impact of the defect on product quality. For example, the grayscale average value, which directly affects color uniformity, is weighted at 0.3; the variance reflects the uniformity of surface texture, with a weight of 0.3; and the average grayscale difference affects surface flatness, with a weight of 0.4. A weighted composite score is calculated, and if the composite score reaches the preset pass threshold, the product is considered qualified; otherwise, it is considered unqualified.

[0106] Throughout the implementation process, the basis for stability judgment must be based on the evaluation results of the previous embodiments to ensure logical coherence. The selection of reference data must accurately correspond to the attributes of the detection sub-area to avoid deviations in the detection results due to attribute matching errors. The comparison process must be carried out strictly in accordance with the preset conditions to ensure the objectivity and consistency of the detection results. For different models of smartphones, due to differences in design and process, the data in the reference database needs to be set separately to meet the detection requirements of different products. For example, the back of a certain model of mobile phone is made of frosted glass, and its reference data is different from the reference data of smooth glass material. When selecting, it is necessary to accurately correspond to the model and material attributes.

[0107] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0108] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A method for detecting defects in a smart phone, characterized in that: The steps include: Perform multi-dimensional optical imaging on the surface of the smartphone to obtain defect detection image data, and divide the defect detection image data into multiple detection sub-areas; The surface defect features of the inspection sub-area are extracted based on texture enhancement technology to highlight abnormal texture information, and defect difference indicators are extracted based on pixel distribution and spatial correlation; The complexity of the distribution characteristics of the defect difference index is evaluated through image analysis based on multi-scale decomposition to determine whether the regularity of the surface defect pattern in the local area meets the requirements; Through dynamic analysis based on time series, the changing trend of defect difference indicators at different time points is evaluated to determine whether the fluctuation range of surface defects is within the allowable range; Based on the regularity of the surface defect pattern in the local area and whether the fluctuation amplitude of the surface defects meets the requirements, the stability of the defect difference index is judged to be qualified; When the stability of the defect difference index is qualified, the defect difference index is compared with the defect feature reference data. After the defect difference index meets the preset conditions, the smartphone defect detection result is determined; Multi-dimensional optical imaging is performed on the surface of the smartphone to obtain defect detection image data, and the defect detection image data is divided into multiple detection sub-areas, including: Deploy multiple optical imaging devices around the surface of the smartphone and configure multiple sets of lighting sources. By adjusting the focal length and shooting parameters of the optical imaging devices, multi-dimensional defect detection image data can be recorded. Store and preliminarily process multi-dimensional defect detection image data to form a pre-processed image set, and decompose the pre-processed image set into multiple detection sub-areas according to pre-set division rules; The surface defect features of the inspection sub-area are extracted based on texture enhancement technology to highlight abnormal texture information, and defect difference indicators are extracted based on pixel distribution and spatial correlation, including: Texture enhancement technology is applied to the defect detection image data within the detection sub-area. Abnormal texture features are enhanced by selecting structural units of different sizes. Structural units of different sizes are used to match abnormal textures of different thicknesses and directions. Quantitative analysis of surface defect characteristics in the detection sub-area is performed based on pixel distribution parameters, which are obtained by calculating the grayscale value distribution characteristics of the local area; Analyze the spatial correlation of surface defect features and extract the correlation feature parameters of surface defects based on the positional relationship between local pixels and their adjacent pixels; The defect difference index is extracted by combining pixel distribution parameters and associated feature parameters. The defect difference index is used to represent the microscopic difference characteristics of surface defects. The complexity of the distribution characteristics of defect difference indicators is evaluated through image analysis based on multi-scale decomposition to determine whether the regularity of the surface defect pattern in the local area meets the requirements. Specifically, the following are included: Generate a two-dimensional distribution matrix based on the spatial distribution of defect difference indicators calculated in the detection sub-area; Decompose the two-dimensional distribution matrix into multiple hierarchical components based on the multi-scale decomposition model; Calculate the analysis dimension corresponding to each level component to measure the complexity of surface defects at different scales in the two-dimensional distribution matrix; According to the distribution law of analysis dimensions at different levels, the complexity characteristic parameters representing the regularity of surface defects are extracted. Complexity characteristic parameters include decomposition range width and decomposition intensity level; Compare the complexity characteristic parameters with the preset requirements to determine whether the regularity of the surface defect pattern in the local area meets the requirements; Through dynamic analysis based on time series, the changing trend of defect difference indicators at different time points is evaluated to determine whether the fluctuation range of surface defects is within the allowable range, including: Arrange the defect difference indicators at each time point in sequence to form a defect difference indicator time series; The time mapping method is used to transform the defect difference index time series into a multi-dimensional state vector set; Dynamic analysis methods are applied to the multidimensional state vector set to extract fluctuation-related characteristic parameters, which are used to measure the variation pattern of surface defects in the time dimension. Compare the fluctuation-related characteristic parameters with the preset range to determine whether the fluctuation amplitude of the surface defect is within the allowable range.

2. The method for detecting defects in a smart phone according to claim 1, wherein: Dynamic analysis methods are applied to the multidimensional state vector set to extract the fluctuation-related characteristic parameters, specifically: The spatial distance between the state vectors corresponding to adjacent time points is calculated as the local fluctuation distance, and the spatial distance is obtained by taking the square root of the sum of the differences of the components in each dimension.

3. The method for detecting defects in a smart phone according to claim 2, wherein: Based on the regularity of the surface defect pattern in the local area and whether the fluctuation amplitude of the surface defects meets the requirements, the stability of the defect difference index is judged to be qualified, including: When the regularity of the surface defect pattern in the local area meets the requirements and the fluctuation amplitude of the surface defects is within the allowable range, the stability of the defect difference index is judged to be qualified; otherwise, the stability of the defect difference index is judged to be unqualified.

4. The method for detecting defects in a smart phone according to claim 3, wherein: When the stability of the defect difference index is qualified, the defect difference index is compared with the defect feature reference data. After the defect difference index meets the preset conditions, the smartphone defect detection result is determined, specifically including: Select corresponding reference data from a preset defect feature reference database according to the attributes of the detection sub-area; Compare the defect difference indicators with the selected reference data item by item to determine whether the preset conditions are met; When all comparison items between the defect difference index and the reference data meet the preset conditions, the smartphone defect detection results are generated.

5. A smartphone defect detection system, used to implement the smartphone defect detection method according to any one of claims 1 to 4, characterized in that: It includes surface imaging acquisition module, defect feature extraction module, regularity complexity assessment module, fluctuation amplitude analysis module, stability comprehensive judgment module and defect result generation module; Surface imaging acquisition module: This module performs multi-dimensional optical imaging on the smartphone surface to obtain defect detection image data, and divides the defect detection image data into multiple detection sub-areas; Defect feature extraction module: This module extracts surface defect features of the detection sub-area based on texture enhancement technology to highlight abnormal texture information, and extracts defect difference indicators based on pixel distribution and spatial correlation; Regularity Complexity Assessment Module: This module evaluates the complexity of the distribution characteristics of defect difference indicators through image analysis based on multi-scale decomposition to determine whether the regularity of the surface defect pattern in the local area meets the requirements; Fluctuation amplitude analysis module: evaluates the changing trend of defect difference indicators at different time points through dynamic analysis based on time series to determine whether the fluctuation amplitude of surface defects is within the allowable range; Stability comprehensive judgment module: Based on the regularity of the surface defect pattern in the local area and whether the fluctuation amplitude of the surface defects meets the requirements, it judges whether the stability of the defect difference index is qualified; Defect result generation module: When the stability of the defect difference index is qualified, the defect difference index is compared with the defect feature reference data. After the defect difference index meets the preset conditions, the smartphone defect detection result is determined.

6. The smartphone defect detection system according to claim 5, characterized in that: The surface imaging acquisition module is specifically used to: deploy multiple optical imaging devices and configure multiple groups of lighting light sources around the surface of the smartphone, record multi-dimensional defect detection image data by adjusting the focal length and shooting parameters of the optical imaging devices; store and preliminarily process the multi-dimensional defect detection image data to form a pre-processed image set, and decompose the pre-processed image set into multiple detection sub-areas according to pre-set division rules.

Images