An automated testing method and system for mobile phone cases

Abstract

This invention relates to the field of automated inspection technology, and discloses an automated inspection method and system for mobile phone cases. The method includes: acquiring raw multidimensional inspection data of the mobile phone case to be inspected and performing multiple data preprocessing steps to obtain multimodal standardized data; then, through multidimensional inspection feature extraction and feature association calculation at the structural level of the mobile phone case, obtaining a multimodal quality feature set of the mobile phone case, and performing multidimensional weight allocation and feature weighted fusion on the comprehensive feature representation space to obtain quality fusion features; generating hierarchical quality evaluation results through multi-level quality evaluation calculation and importance ranking; determining the quality level and performing confidence cross-validation based on the usage scenario of the mobile phone case to obtain the quality level status of the mobile phone case; and generating automated inspection results of the mobile phone case to be inspected through anomaly pattern recognition and early warning level adjustment associated with production batch information. This application improves the accuracy of mobile phone case quality inspection and the efficiency of multimodal data fusion.

Description

Technical Field

[0001] This invention relates to the field of automated testing technology, and in particular to an automated testing method and system for mobile phone cases. Background Technology

[0002] In the field of mobile phone case manufacturing technology, quality inspection is a crucial link in product development and quality control, and real-time detection of multimodal features and defect identification are the most critical and challenging stages in quality inspection. Accurately detecting the quality status of mobile phone cases is essential for manufacturers, R&D institutions, and testing centers, directly impacting product market competitiveness, user satisfaction, and brand reputation. Therefore, an effective automated inspection method for monitoring the appearance quality, material safety, geometric accuracy, and functional performance of mobile phone cases is crucial for ensuring the quality stability and safety of mobile phone case products.

[0003] Currently, machine vision technology is used to analyze image data and build defect recognition models to assess appearance quality, or deep learning technology is being applied to material composition testing to better identify harmful substances and assess material safety. Meanwhile, laser scanning and coordinate measuring machine (CMM) technologies are widely used for geometric accuracy testing, while mechanical testing and electromagnetic compatibility testing are used to evaluate the functional performance of phone cases. However, these methods still face challenges in integrating multi-dimensional quality characteristics, handling nonlinear correlations, and adapting to diverse product specifications. Furthermore, these testing methods often overlook some unique characteristics of phone cases, such as the impact of opening precision on signal transmission, the impact of material thickness on wireless charging efficiency, and the compatibility of ruggedness and water resistance with usage scenarios. These factors can significantly affect quality judgment and rating. In other words, existing phone case quality testing methods neglect multi-dimensional quality characteristics and some unique features of phone cases, resulting in lower accuracy of the final quality testing results. Summary of the Invention

[0004] The main objective of this invention is to solve the problem that existing mobile phone case quality inspection methods neglect multi-dimensional quality characteristics and some unique features of mobile phone cases, resulting in low accuracy of the final quality inspection results.

[0005] The first aspect of this invention provides an automated inspection method for mobile phone cases. The automated inspection method includes: collecting raw multidimensional inspection data of the mobile phone case to be inspected, and performing multiple data preprocessing on the raw multidimensional inspection data to obtain multimodal standardized data; extracting multidimensional inspection features from the multimodal standardized data to obtain a multimodal quality feature set of the mobile phone case, and performing feature association calculation and cross-domain mapping transformation on the multimodal quality feature set of the mobile phone case according to a preset mobile phone case structural hierarchy to obtain a comprehensive feature representation space; performing multidimensional weight allocation and feature weighted fusion on the comprehensive feature representation space to obtain quality fusion features, and performing multi-level quality evaluation calculation and importance ranking on the quality fusion features based on a preset quality hierarchy structure to obtain hierarchical quality evaluation results; determining the quality level of the mobile phone case based on its usage scenario and performing confidence cross-validation on the hierarchical quality evaluation results to obtain the quality level status of the mobile phone case, and performing abnormal pattern recognition and early warning level adjustment based on production batch information association on the quality level status of the mobile phone case to generate automated inspection results for the mobile phone case to be inspected.

[0006] Optionally, in a first implementation of the first aspect of the present invention, the original multidimensional detection data includes visible light image raw data, near-infrared spectral raw data, X-ray fluorescence raw data, laser scanning raw data, mechanical test raw data, and electromagnetic performance raw data. The step of performing multiple data preprocessing on the original multidimensional detection data to obtain multimodal standardized data includes: calculating the intensity gradient and direction distribution statistics of the reflective area on the surface of the mobile phone case from the visible light image raw data to obtain a reflective interference distribution map; performing regional illumination compensation and local contrast enhancement based on the reflective interference distribution map to obtain a homogenized visible light image; and performing baseline drift detection and polynomial fitting correction on the near-infrared spectral raw data to obtain standardized spectral data; and performing characteristic peak identification and component quantification calculation on the standardized spectral data to obtain... Material composition feature data, and elemental composition data obtained by performing energy spectrum deconvolution and elemental quantitative calculation on the raw X-ray fluorescence data, and geometric structure point cloud data obtained by performing point cloud filtering and coordinate system registration on the raw laser scanning data, and mechanical property data obtained by performing temperature compensation and performance parameter standardization on the raw mechanical test data, and electromagnetic property data obtained by performing frequency domain transformation and feature extraction on the raw electromagnetic property data; edge detection and texture feature extraction of the mobile phone case contour are performed on the homogenized visible light image to obtain structured visual feature data, and timestamp alignment and data format standardization are performed on the structured visual feature data, the material composition feature data, the elemental composition data, the geometric structure point cloud data, the mechanical property data and the electromagnetic property data to generate multimodal standardized data.

[0007] Optionally, in a second implementation of the first aspect of the present invention, the step of extracting multi-dimensional detection features from the multimodal standardized data to obtain a multimodal quality feature set for the mobile phone case includes: performing dual-path convolutional defect feature extraction and multi-scale fusion on the structured visual feature data in the multimodal standardized data to obtain an appearance defect feature map, and performing defect region segmentation and defect level calculation on the appearance defect feature map to obtain an appearance quality feature vector; performing hazardous substance concentration calculation and material safety index calculation on the material composition feature data and elemental composition data in the multimodal standardized data to obtain a material safety evaluation result, and performing risk level numerical mapping and safety encoding on the material safety evaluation result to obtain a material safety feature vector. The process involves: measuring the opening accuracy and calculating the surface quality of the geometric point cloud data in the multimodal standardized data to obtain a geometric accuracy evaluation result; calculating the assembly matching degree and encoding the feature of the geometric accuracy evaluation result to obtain a geometric quality feature vector; quantifying the three-proof performance and calculating the wireless charging efficiency of the mechanical performance data and electromagnetic property data in the multimodal standardized data to obtain a functional performance evaluation result; standardizing the performance indicators and encoding the feature of the functional performance evaluation result to obtain a functional performance feature vector; and unifying and encoding the feature dimensions based on the appearance quality feature vector, the material safety feature vector, the geometric quality feature vector, and the functional performance feature vector to generate a multimodal quality feature set for the phone case.

[0008] Optionally, in the third implementation of the first aspect of the present invention, the step of performing feature association calculation and cross-domain mapping transformation on the multimodal quality feature set of the phone case with a preset phone case structure hierarchy to obtain a comprehensive feature representation space includes: performing feature domain division and association calculation on the multimodal quality feature set of the phone case with appearance and material association to obtain a cross-domain feature association matrix, and performing feature semantic alignment and dimension unification on geometric accuracy and functional performance based on the cross-domain feature association matrix to obtain multimodal features with unified quality attributes; performing feature interaction calculation and synergy effect calculation on the multimodal features with unified quality attributes to obtain a feature synergy effect representation, and performing weight adaptive calculation and weighted fusion on the feature synergy effect representation with phone case user demand features to obtain a fused feature representation; performing high-dimensional mapping and spatial projection on the fused feature representation to obtain a high-dimensional feature representation, and using the high-dimensional feature representation to construct a hierarchical feature representation structure to obtain a multidimensional quality subspace of the phone case; and performing unified integration and index construction on the multidimensional quality subspace of the phone case to generate a comprehensive feature representation space.

[0009] Optionally, in the fourth implementation of the first aspect of the present invention, the step of performing multi-dimensional weight allocation and feature weighted fusion on the comprehensive feature representation space to obtain quality fusion features includes: performing multi-head attention calculation on the comprehensive feature representation space to correlate the appearance, material, geometry, and function of the phone case, obtaining the attention weight distribution of the multi-dimensional features, and performing feature interaction intensity calculation and synergy effect calculation on the attention weight distribution to obtain a feature synergy weight matrix; and using the feature synergy weight matrix to perform weighted fusion and nonlinear transformation on the comprehensive feature representation space to generate a quality-oriented fusion feature representation.

[0010] Optionally, in the fifth implementation of the first aspect of the present invention, the step of performing multi-level quality evaluation calculations and importance ranking on the quality fusion features based on a preset quality hierarchy structure to obtain hierarchical quality evaluation results includes: calculating various mobile phone case evaluation indicators on the quality fusion features based on the preset quality hierarchy structure to obtain multi-level quality evaluation results; extracting overall performance factors and calculating the contribution of protection effects on the multi-level quality evaluation results to obtain an importance ranking of quality influence factors; allocating quality influence weights and factor weights based on the importance ranking to obtain a quality factor weight distribution; and performing hierarchical combination calculations on the multi-level quality evaluation results based on the quality factor weight distribution to obtain hierarchical quality evaluation results.

[0011] Optionally, in the sixth implementation of the first aspect of the present invention, the step of determining the quality level of the phone case based on the layered quality evaluation results and performing confidence cross-validation on the phone case usage scenario to obtain the phone case quality level status includes: performing multi-level quality classification calculation on the layered quality evaluation results of the phone case based on the usage scenario requirement parameters corresponding to the phone case to be tested and a preset adaptive classification model to obtain a quality level determination result; performing phone case protection performance matching degree calculation and usage adaptability assessment calculation on the quality level determination result to obtain a performance assessment result, and performing phone case quality reliability calculation and level confidence quantification on the performance assessment result to obtain a level confidence assessment; and performing secondary matching degree calculation and level correction on the phone case key indicators based on a preset phone case quality standard threshold and the level confidence assessment to generate a phone case quality level status.

[0012] Optionally, in the seventh implementation of the first aspect of the present invention, the step of performing abnormal pattern recognition and early warning level adjustment based on production batch information association of the quality grade status of the phone case to generate an automated detection result of the phone case to be tested includes: performing time series calculation and abnormal pattern recognition on the quality grade status of the phone case based on preset production batch information to obtain a quality anomaly detection result, and performing risk quantification on the quality anomaly detection result to obtain a quality risk prediction result; classifying the quality risk prediction result into early warning levels and determining the early warning intensity to obtain a quality anomaly early warning signal, and performing multi-level early warning judgment and early warning level strategy matching on the quality anomaly early warning signal based on a preset risk response strategy to generate an automated detection result of the phone case to be tested.

[0013] The second aspect of this invention provides an automated inspection system for mobile phone cases. The automated inspection system includes: a data preprocessing module for collecting raw multidimensional inspection data of the mobile phone case to be inspected, and performing multiple data preprocessing steps on the raw multidimensional inspection data to obtain multimodal standardized data; a feature fusion module for extracting multidimensional inspection features from the multimodal standardized data to obtain a multimodal quality feature set of the mobile phone case, and performing feature association calculations and cross-domain mapping transformations on the multimodal quality feature set of the mobile phone case at a preset mobile phone case structural hierarchy to obtain a comprehensive feature representation space; a quality evaluation module for performing multidimensional weight allocation and feature weighted fusion on the comprehensive feature representation space to obtain quality fusion features, and performing multi-level quality evaluation calculations and importance ranking on the quality fusion features based on a preset quality hierarchy structure to obtain hierarchical quality evaluation results; and a detection judgment module for determining the quality level of the mobile phone case based on the hierarchical quality evaluation results for the mobile phone case's usage scenario and performing confidence cross-validation to obtain the mobile phone case's quality level status, and performing abnormal pattern recognition and early warning level adjustment based on production batch information association on the mobile phone case's quality level status to generate automated inspection results for the mobile phone case to be inspected.

[0014] The above-mentioned automated detection method and system for mobile phone cases. In this embodiment of the invention, multi-dimensional detection data of the mobile phone case to be tested is collected and preprocessed to obtain multi-modal standardized data. Then, feature extraction and cross-domain mapping transformation are performed on these data to obtain a comprehensive feature representation space. Subsequently, through multi-dimensional weight allocation and feature weighted fusion, quality fusion features are obtained and multi-level quality evaluation is performed to obtain hierarchical quality evaluation results. Finally, based on the usage scenario, quality level determination and abnormal pattern recognition are performed to output automated detection results. Through hierarchical data fusion and feature association analysis, the multi-modal integration problem of mobile phone case quality detection is solved, especially in terms of appearance quality, material safety, geometric accuracy, and functional performance. The unique characteristics of mobile phone cases and usage scenario requirements are fully considered, effectively improving detection accuracy. Furthermore, the multi-level feature association and cross-domain mapping strategy not only realizes collaborative analysis between different quality dimensions but also enhances the reliability of quality evaluation. In addition, through scenario adaptation determination and batch association early warning, quality level and abnormal patterns are accurately identified, thereby achieving efficient and accurate detection of mobile phone case quality as a whole.

[0015] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the first embodiment of the automated detection method for mobile phone cases in this invention;

[0018] Figure 2 This is a schematic diagram of an embodiment of the automated detection system for mobile phone cases according to the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. 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.

[0020] The terms "comprising" and "having," and any variations thereof, used in the embodiments of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0021] To facilitate understanding of this embodiment, the specific process of this embodiment is described below. Please refer to [link / reference]. Figure 1 The first embodiment of the automated detection method for mobile phone cases in this invention includes:

[0022] 101. Collect the original multidimensional detection data of the mobile phone case to be tested, and perform multiple data preprocessing on the original multidimensional detection data to obtain multimodal standardized data;

[0023] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0024] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0025] In this embodiment, a preset multimodal sensor array is used to collect raw multidimensional detection data of the phone case under test. This raw multidimensional detection data includes raw visible light image data, raw near-infrared spectral data, raw X-ray fluorescence data, raw laser scanning data, raw mechanical test data, and raw electromagnetic performance data. The intensity gradient and directional distribution of the reflective area on the phone case surface are calculated from the raw visible light image data to obtain a reflective interference distribution map. Based on this map, regional illumination compensation and local contrast enhancement are performed to obtain a homogenized visible light image. Furthermore, baseline drift detection and polynomial fitting correction are performed on the raw near-infrared spectral data to obtain standardized spectral data. Finally, characteristic peak identification and quantitative component calculation are performed on the standardized spectral data to obtain the material composition characteristics. The system extracts data from various sources, including: elemental composition data from raw X-ray fluorescence data (through energy spectrum deconvolution and elemental quantification); geometric structure point cloud data from raw laser scanning data (through point cloud filtering and coordinate system registration); mechanical performance data from raw mechanical test data (through temperature compensation and performance parameter standardization); and electromagnetic property data from raw electromagnetic performance data (through frequency domain transformation and feature extraction). It also performs edge detection and texture feature extraction on homogenized visible light images to obtain structured visual feature data. Finally, it performs timestamp alignment and data format standardization on the structured visual feature data, material composition feature data, elemental composition data, geometric structure point cloud data, mechanical performance data, and electromagnetic property data to generate multimodal standardized data.

[0026] In practical applications, a pre-set multimodal sensor array is used to collect comprehensive data from the phone case under test. This array includes six types of testing equipment: a high-resolution CCD camera, a near-infrared spectrometer, an X-ray fluorescence analyzer, a 3D laser scanner, a materials testing machine, and a network analyzer. During the actual testing process, the phone case is fixed on a rotatable testing platform, and each sensor synchronously collects data according to a pre-set spatial layout and timing control program. For example, for a transparent PC phone case, the CCD camera captures visible light images of its surface texture and defects, the near-infrared spectrometer detects the vibrational characteristics of the material's molecules, the X-ray fluorescence analyzer detects the elemental composition, the laser scanner acquires the 3D geometry, the materials testing machine tests the impact resistance, and the network analyzer tests the electromagnetic shielding effect. This multimodal synchronous acquisition method can obtain comprehensive quality information about the phone case, avoiding the limitations of a single testing method.

[0027] Furthermore, for the raw visible light image data, the intensity gradient is first calculated using the Sobel operator. This operator detects the rate of change of pixel intensity using convolutional kernels in the horizontal and vertical directions, and then statistically analyzes the distribution characteristics of the gradient direction to identify reflective areas on the phone case surface caused by the material's gloss, resulting in a reflective interference distribution map (a two-dimensional matrix recording the reflective intensity and distribution range of each pixel). Based on this distribution map, a regional adaptive histogram equalization technique is used for illumination compensation, reducing brightness gain in severely reflective areas and increasing brightness gain in shadow areas. Simultaneously, a contrast-limited adaptive histogram equalization algorithm is used to enhance local contrast. For example, when inspecting a black glossy phone case, surface reflections can obscure minor scratches; this processing can significantly improve image quality, making defect features clearer and more identifiable. For the raw near-infrared spectral data, the least squares method is first used to detect spectral baseline drift, which is typically caused by instrument temperature changes and environmental factors. Then, a fifth-order polynomial fitting algorithm is used to correct the baseline, eliminating systematic errors. Based on standardized spectral data, the second derivative method is used for characteristic peak identification. This method finds the absorption peak position by calculating the second derivative of the spectral curve, avoiding the influence of baseline tilt. The quantitative calculation of components adopts the Beer-Lambert law, establishing a linear relationship between the integral area of ​​the characteristic peak and the concentration. For example, when testing TPU material mobile phone cases, at 1730 cm⁻¹... -1The characteristic peaks at certain locations correspond to the stretching vibrations of polyurethane bonds. Peak area calculations allow for quantitative analysis of plasticizer content, determining whether safety standards are exceeded. This spectral processing method accurately identifies the composition of mobile phone case materials. For raw X-ray fluorescence data, a Gaussian-Lorentz mixture function is used for energy spectrum deconvolution. This function effectively separates overlapping characteristic peaks, improving the accuracy of element identification. Energy spectrum deconvolution is a mathematical process that decomposes composite peaks into single-element characteristic peaks. An iterative fitting algorithm determines the peak position, peak height, and peak width parameters for each element. Elemental quantification uses the basic parameter method, considering matrix effects and inter-element interactions. The element content is determined by the ratio of theoretically calculated coefficients to measured intensities. For example, when testing colored mobile phone cases, this method can accurately detect the content of harmful heavy metals such as lead, cadmium, mercury, and hexavalent chromium, ensuring compliance with RoHS directives. This processing method achieves ppm-level detection accuracy, meeting stringent environmental regulations. For raw laser scanning data, a statistical filtering algorithm is first used to remove noise points from the point cloud data. This algorithm identifies outliers based on the statistical characteristics of the distance between each point and its neighbors. Then, coordinate system registration is performed. The ICP iterative nearest point algorithm is used to unify the point cloud data scanned from different angles into the same coordinate system, achieving complete 3D geometric reconstruction and obtaining geometric structure point cloud data (a set of 3D points containing xyz coordinate information, each point representing a spatial position on the surface of the phone case). For example, for a phone case with a camera opening, the point cloud data can be used to accurately measure the opening diameter, positional deviation, and edge roundness to determine whether it meets assembly requirements. The accuracy of this geometric measurement method can reach 0.01mm, which can meet the quality requirements of precision manufacturing of high-end phone cases. Furthermore, the raw data for mechanical testing and electromagnetic performance are standardized. The mechanical testing data uses a temperature compensation algorithm to eliminate the influence of ambient temperature on material properties, and linear interpolation is used to normalize the test results at different temperatures to standard temperature conditions. Performance parameter standardization uses the Z-score standardization method to ensure the comparability of parameters with different dimensions. The raw electromagnetic performance data is converted from the time domain to the frequency domain using a fast Fourier transform to extract the amplitude and phase information of key frequency points. For example, when testing the impact resistance of a phone case, temperature compensation can eliminate the impact of seasonal temperature differences on the test results; when testing electromagnetic shielding effectiveness, frequency domain analysis can obtain shielding effectiveness data for different frequency bands. Finally, multimodal data undergoes time synchronization and format standardization processing. High-precision timestamp alignment technology ensures the time consistency of the six types of test data, eliminating data deviations caused by differences in device response time. Data format standardization employs a unified data structure and encoding format, integrating different types of test data into a standardized data matrix. Timestamp alignment accuracy reaches the millisecond level, ensuring that multimodal data accurately reflects the state characteristics of the phone case at the same moment.

[0028] 102. Extract multi-dimensional detection features from the multimodal standardized data to obtain a multimodal quality feature set for mobile phone cases. Perform feature association calculation and cross-domain mapping transformation on the multimodal quality feature set of mobile phone cases according to the preset structure level of the mobile phone case to obtain a comprehensive feature representation space.

[0029] In this embodiment, dual-path convolutional defect feature extraction and multi-scale fusion are performed on the structured visual feature data in the multimodal standardized data to obtain an appearance defect feature map. Defect region segmentation and defect level calculation are then performed on the appearance defect feature map to obtain an appearance quality feature vector. Hazardous substance concentration calculation and material safety index calculation are performed on the material composition feature data and elemental composition data in the multimodal standardized data to obtain a material safety evaluation result. Risk level numerical mapping and safety encoding are then performed on the material safety evaluation result to obtain a material safety feature vector. Opening accuracy measurement and surface quality calculation are performed on the geometric structure point cloud data in the multimodal standardized data to obtain a geometric accuracy evaluation result. Assembly matching degree calculation and feature encoding are then performed on the geometric accuracy evaluation result to obtain a geometric quality feature vector. Three-proof performance quantification and wireless charging efficiency calculation are performed on the mechanical performance data and electromagnetic property data in the multimodal standardized data to obtain a functional performance evaluation result. Performance index standardization and feature encoding are then performed on the functional performance evaluation result to obtain a functional performance feature vector. The system generates a multimodal quality feature set for mobile phone cases by unifying and encoding the feature dimensions of appearance quality feature vectors, material safety feature vectors, geometric quality feature vectors, and functional performance feature vectors. It then performs feature domain partitioning and correlation calculation on the multimodal quality feature set, based on the relationship between appearance and material, to obtain a cross-domain feature correlation matrix. Based on this matrix, it aligns the geometric precision and functional performance features semantically and unifies the dimensions, resulting in multimodal features with unified quality attributes. The system performs feature interaction calculation and synergy effect calculation on these unified quality attribute multimodal features, obtaining a feature synergy effect representation. This representation is then weighted and fused using adaptive weighting of user demand features for mobile phone cases, resulting in a fused feature representation. The fused feature representation undergoes high-dimensional mapping and spatial projection to obtain a high-dimensional feature representation. A hierarchical feature representation structure is constructed using this high-dimensional representation, resulting in a multidimensional quality subspace for mobile phone cases. Finally, the system integrates and indexes the multidimensional quality subspace to generate a comprehensive feature representation space.

[0030] In practical applications, dual-path convolutional defect feature extraction is performed on structured visual feature data from multimodal standardized data. Parallel coarse-grained and fine-grained convolutional paths are used to simultaneously process the homogenized visible light image. The coarse-grained path uses a 7×7 convolutional kernel to detect large-area defects such as cracks and deformations, while the fine-grained path uses a 3×3 convolutional kernel to detect minute defects such as scratches and blemishes. Dual-path convolutional defect feature extraction refers to a processing method that simultaneously extracts defect features at different scales through two branches of convolutional neural networks with different receptive fields, and then performs feature fusion. Multi-scale fusion employs a feature pyramid structure, unifying feature maps of different resolutions to the same size through upsampling and downsampling operations, and then fusing them element-wise. For example, when detecting transparent phone cases, the coarse-grained path can identify overall bubble defects, while the fine-grained path can detect minute scratches on the surface. The fused appearance defect feature map contains complete defect information. Defect region segmentation uses a threshold-based region growing algorithm, starting from the defect center point and expanding outwards to regions with similar pixel values ​​to determine the precise boundary of the defect. The defect level calculation is based on defect area, depth, and location information, classifying defects into three levels: minor, moderate, and severe, forming a 128-dimensional appearance quality feature vector. This allows for the simultaneous capture of defect features at different scales, significantly improving the comprehensiveness and accuracy of defect detection.

[0031] A comprehensive safety assessment is conducted on material composition and elemental data. The concentration of hazardous substances is calculated using a weighted cumulative method, with different weighting coefficients assigned to hazardous substances such as lead, mercury, cadmium, hexavalent chromium, polybrominated biphenyls (PBBBs), and polybrominated diphenyl ethers (PBDEs) according to RoHS and REACH regulations. The material safety index calculation includes assessments of three dimensions: total hazardous substance index, volatile organic compound (VOC) release, and biocompatibility index. For example, for a silicone phone case, near-infrared spectroscopy detected a plasticizer content of 15 ppm, and X-ray fluorescence detected a lead content of 8 ppm. The safety index is calculated based on the limits, and a risk warning is generated if the lead content exceeds the limit. The risk level numerical mapping converts continuous safety indicators into discrete risk levels, using a piecewise linear mapping function to map indicator values ​​to a risk level of 1-5, where 1 represents safe and 5 represents high risk. The safety coding uses a binary encoding method, combining multiple safety indicators into a 64-bit material safety feature vector, with each 8 bits representing the assessment result of one safety dimension.

[0032] Precise dimensional and quality assessments are performed on the geometric point cloud data. The opening accuracy measurement employs a least-squares circle fitting algorithm, fitting the theoretical circle of the opening to the point cloud data and calculating the deviation between the actual opening and the design dimensions. Surface quality calculations include assessments of three indices: surface roughness, flatness, and roundness. Surface roughness is obtained by calculating the dispersion of the point cloud data on the fitting plane; flatness is calculated by determining the maximum distance difference from a point to the plane using least-squares plane fitting; and roundness is quantified by the maximum deviation between the fitted circle and the actual contour. For example, for a phone case with a charging port opening, the measured opening diameter deviation is 0.02 mm, and the surface roughness is 0.8 μm, meeting precision manufacturing requirements. Assembly matching calculations are based on tolerance analysis of key dimensions, assessing the matching between the actual dimensions of the phone case and the design dimensions of the phone body, and calculating assembly gaps and interference fits. Feature encoding normalizes multiple geometric parameters to the 0-1 range, forming a 96-dimensional geometric quality feature vector, with each dimension representing a geometric quality index. This precise measurement method ensures a perfect match between the phone case and the phone body, avoiding functional defects caused by poor assembly.

[0033] A comprehensive functional evaluation is conducted using mechanical and electromagnetic property data. The tri-proof performance quantification includes performance assessments across three dimensions: waterproof, dustproof, and drop-proof. Waterproof performance is quantified through material sealing test data, dustproof performance is calculated through particulate matter blocking efficiency, and drop-proof performance is assessed through energy absorption capacity from impact tests. Wireless charging efficiency is calculated based on the permeability and dielectric constant parameters from the electromagnetic property data, using an electromagnetic field simulation algorithm to calculate the attenuation of the wireless charging signal by the phone case. For example, when testing metal phone cases, their electromagnetic shielding effect significantly reduces wireless charging efficiency, resulting in a 15% reduction in charging efficiency through quantitative calculation. Performance index standardization employs the max-min normalization method to unify performance parameters with different dimensions to the same numerical range. Feature encoding organizes the standardized functional performance parameters into an 80-dimensional functional performance feature vector, encompassing multiple functional dimensions such as mechanical strength, electromagnetic compatibility, and thermal conductivity. This functional evaluation method comprehensively quantifies the phone case's performance, ensuring the product meets practical application requirements. Furthermore, based on four types of quality feature vectors, dimensional unification and encoding integration are performed. Principal component analysis is used to unify feature dimensions, compressing feature vectors from different dimensions into the same feature space dimension. The unified coding system employs feature standardization and normalization to ensure that different types of features have the same numerical range and distribution characteristics, thereby obtaining a multimodal quality feature set for mobile phone cases (a 368-dimensional comprehensive feature matrix containing complete information on four major categories of quality features: appearance, material, geometry, and function). For example, for a leather mobile phone case, its multimodal quality feature set includes multidimensional information such as surface texture features, leather composition features, edge precision features, and protective performance features. This achieves effective fusion of quality information obtained from different detection methods.

[0034] For the multimodal quality feature set of mobile phone cases, appearance and material features are grouped into one association domain, and geometric and functional features into another, based on the inherent logical relationship of mobile phone case quality. The correlation is calculated using the Pearson correlation coefficient method to determine the linear correlation between different feature domains, forming a cross-domain feature correlation matrix (a symmetric matrix where element values ​​represent the correlation strength between different features). Then, the cross-domain feature correlation matrix is ​​semantically aligned using linear transformation to map feature vectors from different feature domains to a unified semantic space, eliminating differences in feature representation. Dimensionality is also unified through feature selection and dimensionality reduction techniques, unifying features from multiple domains to the same dimensional space. For example, a strong correlation of 0.85 is found between material hardness and impact resistance; semantic alignment establishes the mapping relationship between the two. The unified multimodal features of quality attributes are standardized feature representations after cross-domain correlation analysis and semantic alignment, eliminating differences in feature representation between different detection methods.

[0035] Deep interactive analysis is performed on multimodal features with unified quality attributes. This involves calculating second-order interaction effects between different feature dimensions using bilinear pooling through feature interaction computation, revealing nonlinear correlations between features. Synergistic effect calculation is also performed, identifying feature combinations that synergistically contribute to quality assessment based on feature complementarity and enhancement analysis, resulting in a feature synergistic effect representation (a high-dimensional tensor recording the complex interaction relationships and synergistic contribution levels among multiple features). Furthermore, the adaptive weighting of user demand features for phone cases dynamically adjusts the importance weights of each feature dimension based on the differences in needs across different usage scenarios. For example, business scenarios prioritize appearance quality, while sports scenarios prioritize protective performance, automatically adjusting feature weights according to the scenario. Weighted fusion employs an attention mechanism, weighting features according to their weight distribution to form a fused feature representation. This adaptive weighting mechanism enables quality assessment to adapt to different application scenarios and user needs.

[0036] The fused feature representation undergoes high-dimensional spatial transformation and structured organization. High-dimensional mapping employs kernel principal component analysis to map the fused features to a higher-dimensional feature space, enhancing the linear separability of the features. Spatial projection, through dimensionality reduction techniques, projects the high-dimensional features onto a visualized low-dimensional space, facilitating feature analysis and quality assessment, resulting in a high-dimensional feature representation (which contains complex feature information after nonlinear transformation, better representing the quality status of the phone case). The hierarchical feature representation structure adopts a tree-like hierarchical structure, organizing related features into different hierarchical nodes, forming a quality feature hierarchy from coarse-grained to fine-grained. For example, the top layer represents the overall quality level, the middle layer represents the assessment of each quality dimension, and the bottom layer represents specific quality indicators. The multidimensional quality subspace of the phone case is a composite spatial structure composed of multiple interrelated feature subspaces, each representing a feature representation of a quality dimension.

[0037] A unified and systematic organization of the multi-dimensional quality subspace of phone cases is implemented. The unified integration of phone case quality employs a feature fusion algorithm to merge information from multiple quality subspaces into a unified quality representation. Index construction utilizes hash indexing technology, assigning a unique index identifier to each quality feature and establishing a fast retrieval and matching mechanism. This generates a comprehensive feature representation space (a complete quality feature knowledge base containing all feature information and relationships required for phone case quality assessment). This space features a hierarchical organizational structure and an efficient indexing mechanism, supporting rapid quality querying and assessment calculations.

[0038] 103. Perform multi-dimensional weight allocation and feature weighted fusion on the comprehensive feature representation space to obtain quality fusion features. Based on the preset quality hierarchy structure, perform multi-level quality evaluation calculation and importance ranking on the quality fusion features to obtain hierarchical quality evaluation results.

[0039] In this embodiment, multi-head attention calculation is performed on the comprehensive feature representation space to correlate the appearance, material, geometry, and function of the phone case, resulting in the attention weight distribution of multi-dimensional features. The feature interaction intensity and synergy effect are then calculated on the attention weight distribution to obtain a feature synergy weight matrix. This matrix is ​​used to perform weighted fusion and nonlinear transformation on the comprehensive feature representation space to generate a quality-oriented fusion feature representation. Based on a preset quality hierarchy, various phone case evaluation indicators are calculated on the quality fusion features to obtain multi-level quality evaluation results. Overall performance factors are extracted and the contribution of protection effects is calculated on the multi-level quality evaluation results to obtain a ranking of the importance of quality impact factors. Based on this ranking, quality impact weights and factor weights are allocated to obtain a quality factor weight distribution. Finally, based on this quality factor weight distribution, a hierarchical combination calculation is performed on the multi-level quality evaluation results to obtain a hierarchical quality evaluation result.

[0040] In practical applications, multi-head attention computation is performed on the comprehensive feature representation space, employing four independent attention heads to handle feature associations across four quality dimensions: appearance, material, geometry, and function. Multi-head attention computation is a parallel processing mechanism where each attention head calculates the association weights between different features using query-key-value triples. Specifically, the comprehensive feature representation space is decomposed into four subspaces, and each attention head uses a different linear transformation matrix to process the corresponding quality dimension features, resulting in a multi-dimensional feature attention weight distribution. For example, the appearance attention head calculates the correlation between surface defects and material texture, while the material attention head calculates the correlation between chemical composition and safety indicators. Furthermore, the feature interaction strength is calculated using a dot product attention mechanism. The interaction strength score is obtained by calculating the dot product of the query vector and the key vector, and then normalized using a softmax function to obtain the attention weight distribution. Finally, a synergistic effect calculation is performed. Based on the outputs of multiple attention heads, a weighted average method is used to fuse the attention weights of different dimensions, forming a feature synergistic weight matrix (this matrix is ​​symmetric, and its elements represent the synergistic strength between different feature dimensions). This multi-head attention mechanism can simultaneously capture the multi-dimensional correlation characteristics of phone case quality, avoiding the one-sidedness of single-dimensional evaluation and significantly improving the accuracy of identifying the correlation between quality features.

[0041] This method utilizes a feature-coordinated weighting matrix to intelligently fuse the comprehensive feature representation space. Weighted fusion employs matrix multiplication, combining the feature-coordinated weighting matrix with the original feature representations to highlight the contributions of important feature dimensions. The nonlinear transformation uses the ReLU activation function and batch normalization to perform nonlinear mapping and numerical stabilization on the weighted fused features. The quality-oriented fusion feature representation is a high-dimensional feature vector after attention weighting and nonlinear transformation, retaining key quality information and eliminating redundant features. For example, when detecting sporty phone cases, the weights of drop-proof performance and material durability features are automatically increased, while the weights of aesthetic features are decreased, generating a fusion feature representation focused on protective performance. This quality-oriented feature fusion method can dynamically adjust feature importance based on the phone case's functional positioning and usage requirements, ensuring that the quality assessment results meet practical application requirements.

[0042] Based on a pre-defined quality hierarchy structure, multi-level index calculations are performed on the quality fusion characteristics. The quality hierarchy structure adopts a three-layer tree structure: the first layer is the overall quality level, the second layer is the evaluation of four quality dimensions, and the third layer is the specific quality indicators. The phone case appearance quality calculation includes three indicators: surface defect density, color difference, and texture consistency; the material safety calculation includes three indicators: harmful substance content, volatile substance release, and biocompatibility; the geometric accuracy calculation includes three indicators: dimensional deviation, shape error, and assembly gap; and the functional performance calculation includes three indicators: protection level, electromagnetic compatibility, and thermal conductivity. Each indicator extracts its corresponding value from the quality fusion characteristics using a specific calculation algorithm. For example, surface defect density is calculated by counting the number of elements exceeding a threshold in the statistical defect feature vector, and material safety is calculated by weighted summation of harmful substance concentration characteristics. Overall performance factors are extracted and the contribution of protection effects is calculated based on the multi-level quality evaluation results. Principal component analysis is used to extract the main quality influencing factors from the multi-level quality evaluation results. The contribution of protection effects is calculated based on the contribution of each quality indicator to the overall protection performance of the phone case, using regression analysis to establish a quantitative relationship between the indicators and the protection effect. The importance ranking of quality impact factors is based on a comprehensive ranking of each factor's variance contribution rate and contribution to protective effect, forming an importance sequence from high to low. This multi-level evaluation system can comprehensively quantify the quality status of mobile phone cases, providing accurate data support for quality control and improvement.

[0043] Dynamic weight allocation and hierarchical combination calculation are performed based on importance ranking. The weight allocation for quality impact uses an exponential decay function, assigning corresponding weight values ​​according to the importance ranking position, with factors ranking higher receiving higher weights. Factor weight allocation considers the different usage scenarios of phone cases, giving additional weight to key quality factors in different application scenarios, resulting in a quality factor weight distribution (a normalized weight vector where the sum of all weight values ​​equals 1, ensuring the rationality of weight allocation). Based on the quality factor weight distribution, hierarchical combination calculation is performed on the multi-level quality evaluation results. A weighted summation method is used to combine the multi-level quality evaluation results according to the quality factor weight distribution. The first layer calculates the weighted average of each indicator within the same quality dimension; the second layer calculates the weighted average between different quality dimensions; and the third layer calculates the comprehensive score of the overall quality level. For example, for phone cases in business scenarios, the appearance quality weight is set to 0.4, the material safety weight to 0.3, the geometric accuracy weight to 0.2, and the functional performance weight to 0.1, reflecting the importance that business users place on appearance and safety. The tiered quality evaluation results include complete quality assessment information from specific indicators to overall levels, forming a hierarchical quality evaluation system. This adaptive weight allocation and tiered combination mechanism can generate personalized quality evaluation results based on different application needs and quality concerns, significantly improving the relevance and practicality of quality assessment.

[0044] 104. The quality level of the phone case is determined and the confidence level is cross-validated based on the usage scenario of the layered quality evaluation results to obtain the quality level status of the phone case. The abnormal pattern recognition and early warning level adjustment of the phone case quality level status are performed by associating production batch information, and the automated detection results of the phone case to be tested are generated.

[0045] In this embodiment, based on the usage scenario requirements parameters corresponding to the phone case to be tested and a preset adaptive classification model, multi-level quality classification calculations are performed on the layered quality evaluation results of the phone case to obtain quality level determination results; the phone case protection performance matching degree calculation and usage adaptability assessment calculations are performed on the quality level determination results to obtain performance assessment results, and the phone case quality reliability calculation and level confidence quantification are performed on the performance assessment results to obtain level confidence assessment; based on the preset phone case quality standard threshold and level confidence assessment, the secondary matching degree calculation and level correction of the phone case key indicators are performed on the quality level determination results to generate the phone case quality level status; based on the preset production batch information, time series calculation and abnormal pattern recognition are performed on the phone case quality level status to obtain quality anomaly detection results, and the quality anomaly detection results are risk quantified to obtain quality risk prediction results; the quality risk prediction results are classified into warning levels and the warning intensity is determined to obtain quality anomaly warning signals, and based on the preset risk response strategy, multi-level warning judgment and warning level strategy matching are performed on the quality anomaly warning signals to generate automated detection results for the phone case to be tested.

[0046] In practical applications, quality levels are determined based on usage scenario requirements and an adaptive classification model. These requirements include quantitative indicators across four dimensions: protection level requirements, emphasis on appearance, price sensitivity, and usage frequency. The adaptive classification model employs a support vector machine algorithm, dynamically adjusting classification boundaries and judgment thresholds according to the specific needs of different usage scenarios. Specifically, multi-level quality classification calculations are performed on the layered quality evaluation results of phone cases, dividing them into four levels: excellent, good, acceptable, and unacceptable, each corresponding to a specific quality score range. In practice, appearance quality is weighted at 0.5 for business scenarios, protection performance at 0.6 for sports scenarios, and price-performance ratio at 0.4 for student scenarios. For example, when testing sports phone cases, based on the high protection requirements of sports scenarios, products with a drop resistance rating of 8 or higher and a waterproof rating of IPX7 or higher are classified as excellent, while those with a drop resistance rating of 6-7 and a waterproof rating of IPX5-6 are classified as good. The quality level determination results include the level label, confidence score, and the compliance status of key quality indicators. This scenario-adaptive classification method can dynamically adjust quality standards according to actual usage needs, avoiding evaluation biases caused by uniform standards and significantly improving the accuracy and practicality of quality classification.

[0047] The quality grade determination results are used for protective performance matching and reliability assessment. The Euclidean distance algorithm is used to calculate the matching degree of the phone case's protective performance, calculating the vector distance between the actual protective performance indicators and the ideal protective performance requirements; the smaller the distance, the higher the matching degree. Adaptability assessment is used, including a comprehensive evaluation across three dimensions: functional adaptability, size adaptability, and material adaptability. Functional adaptability is calculated by comparing the degree of matching between the phone case's functions and user needs; size adaptability is calculated by the degree of conformity between the geometric dimensions and the phone model; and material adaptability is calculated by the adaptability of material properties to the usage environment, thus obtaining the performance assessment results. Furthermore, the performance assessment results are used to calculate the phone case's quality reliability. This involves statistical analysis of historical quality data, using the Weibull distribution function to assess the product's reliability indicators and expected lifespan, and quantifying the confidence level. A Bayesian probability method is used, combining prior quality distribution and current test results to calculate the confidence probability of the quality grade. For example, a silicone phone case might have a protective performance matching degree of 0.85, a usability assessment of 0.78, a quality reliability index of 95%, and a grade confidence level of 0.82. This indicates that the product's quality grade assessment has high reliability. The grade confidence level assessment provides a reliable basis for subsequent quality decisions, avoiding erroneous judgments due to testing errors.

[0048] The quality grade and status are determined based on preset quality standard thresholds (based on industry standards and user satisfaction surveys, including minimum pass and excellent standards for each quality indicator) and confidence level assessment. The secondary matching degree calculation for key indicators of the phone case uses a weighted similarity algorithm to precisely match key quality indicators with standard thresholds, calculating the degree of deviation and compliance. The grade correction employs a confidence-weighted adjustment method: when the confidence level is below 0.7, the quality grade is automatically lowered by one level; when the confidence level is above 0.9 and all key indicators meet the standards, the quality grade is considered to be raised by one level. For example, a phone case initially judged to be at a good level, but with a confidence level of only 0.65 and its drop resistance slightly below the standard threshold, is automatically corrected to a pass level. The phone case quality grade status is the final quality assessment result after scenario adaptation, reliability assessment, and grade correction, including the quality grade, key indicator compliance status, reliability assessment, and applicable scenario suggestions. This multi-verification and correction mechanism ensures the accuracy and stability of the quality grade determination, effectively reducing the risk of misjudgment.

[0049] Anomaly detection and risk warning are performed based on production batch information. The time series calculation uses a sliding window method, with each production batch as the time unit, to statistically analyze the distribution and trend of quality levels within a continuous time window. Anomaly pattern recognition employs a statistical control chart-based approach, setting upper and lower limits for quality levels. When the quality level of three consecutive batches exceeds the control limits or shows a significant downward trend, it is identified as an anomaly pattern, yielding quality anomaly detection results (including anomaly type, impact level, duration, and preliminary analysis of possible causes). The quality anomaly detection results are then quantified for risk, using a risk assessment matrix method to quantitatively combine the probability of anomaly occurrence and the severity of impact, calculating a comprehensive risk level and obtaining a quality risk prediction result. For example, if the drop resistance of five consecutive batches of mobile phone cases shows a downward trend, from excellent to good, it is identified as a material performance anomaly pattern, with a risk level assessed as medium risk. The quality risk prediction results are then categorized into warning levels and their intensity. The warning level classification divides the risk level into four levels: low, medium, high, and urgent. The warning intensity determination sets corresponding warning response measures based on the risk level and the scope of impact. Furthermore, based on a preset risk response strategy, multi-level early warning judgments and matching of early warning levels are performed on quality anomaly warning signals. The multi-level early warning judgment employs a decision tree algorithm, intelligently determining the early warning level based on factors such as anomaly type, risk level, and scope of impact. The matching of early warning level strategies automatically matches different early warning levels with corresponding countermeasures, forming a complete risk response plan. The automated inspection results of the phone cases under inspection include quality level status, risk warning information, improvement suggestions, and subsequent monitoring requirements, providing comprehensive decision support for production quality management. This batch-information-based anomaly early warning mechanism can promptly detect early signs of quality problems, achieving a shift from passive detection to proactive prevention, significantly improving the foresight and effectiveness of quality management, thereby realizing efficient and accurate overall inspection of phone case quality.

[0050] In this embodiment of the invention, multi-dimensional detection data of the phone case to be tested is collected and preprocessed to obtain multi-modal standardized data. Then, feature extraction and cross-domain mapping transformation are performed on this data to obtain a comprehensive feature representation space. Furthermore, through multi-dimensional weight allocation and feature weighted fusion, quality fusion features are obtained, and multi-level quality evaluation is performed to obtain hierarchical quality evaluation results. Finally, quality level determination and abnormal pattern recognition are performed based on the usage scenario, outputting automated detection results. Through hierarchical data fusion and feature association analysis, the multi-modal integration problem of phone case quality detection is solved, particularly in terms of appearance quality, material safety, geometric accuracy, and functional performance. It fully considers the unique characteristics of phone cases and the needs of usage scenarios, effectively improving detection accuracy. Moreover, the multi-level feature association and cross-domain mapping strategy not only achieves collaborative analysis between different quality dimensions but also enhances the reliability of quality evaluation. In addition, through scenario adaptation determination and batch association early warning, quality levels and abnormal patterns are accurately identified, thus achieving efficient and accurate detection of phone case quality overall.

[0051] The above describes the automated detection method for mobile phone cases in the embodiments of the present invention. The following describes the automated detection system for mobile phone cases in the embodiments of the present invention. Please refer to [link / reference]. Figure 2 One embodiment of the automated detection system for mobile phone cases in this invention includes:

[0052] The data preprocessing module 201 is used to collect the original multidimensional detection data of the mobile phone case to be tested, and to perform multiple data preprocessing on the original multidimensional detection data to obtain multimodal standardized data.

[0053] The feature fusion module 202 is used to extract multi-dimensional detection features from the multimodal standardized data to obtain a multimodal quality feature set of the phone case, and to perform feature association calculation and cross-domain mapping transformation on the multimodal quality feature set of the phone case according to a preset phone case structure level to obtain a comprehensive feature representation space.

[0054] The quality evaluation module 203 is used to perform multi-dimensional weight allocation and feature weighted fusion on the comprehensive feature representation space to obtain quality fusion features, and to perform multi-level quality evaluation calculation and importance ranking on the quality fusion features based on a preset quality hierarchy structure to obtain hierarchical quality evaluation results.

[0055] The detection and judgment module 204 is used to determine the quality level of the phone case in the usage scenario and perform confidence cross-validation on the layered quality evaluation results to obtain the quality level status of the phone case, and to perform abnormal pattern recognition and early warning level adjustment on the quality level status of the phone case by associating it with production batch information, so as to generate the automated detection results of the phone case to be tested.

[0056] In this embodiment of the invention, multi-dimensional detection data of the phone case to be tested is collected and preprocessed to obtain multi-modal standardized data. Then, feature extraction and cross-domain mapping transformation are performed on this data to obtain a comprehensive feature representation space. Furthermore, through multi-dimensional weight allocation and feature weighted fusion, quality fusion features are obtained, and multi-level quality evaluation is performed to obtain hierarchical quality evaluation results. Finally, quality level determination and abnormal pattern recognition are performed based on the usage scenario, outputting automated detection results. Through hierarchical data fusion and feature association analysis, the multi-modal integration problem of phone case quality detection is solved, particularly in terms of appearance quality, material safety, geometric accuracy, and functional performance. It fully considers the unique characteristics of phone cases and the needs of usage scenarios, effectively improving detection accuracy. Moreover, the multi-level feature association and cross-domain mapping strategy not only achieves collaborative analysis between different quality dimensions but also enhances the reliability of quality evaluation. In addition, through scenario adaptation determination and batch association early warning, quality levels and abnormal patterns are accurately identified, thus achieving efficient and accurate detection of phone case quality overall.

[0057] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0058] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention 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 the present invention.

Claims

1. An automated method of detecting a phone case, the method comprising: The automatic detection method of the mobile phone shell comprises: Collecting original multi-dimensional detection data of a mobile phone shell to be detected, and performing multiple data preprocessing on the original multi-dimensional detection data to obtain multi-modal standardized data; Extracting multi-dimensional detection features from the multi-modal standardized data to obtain a mobile phone shell multi-modal quality feature set, and performing cross-domain feature correlation calculation and cross-domain mapping transformation between multiple quality feature dimensions in the mobile phone shell multi-modal quality feature set to obtain a comprehensive feature representation space, wherein the multiple quality feature dimensions include appearance quality features, material safety features, geometric quality features, and functional performance features; Performing multi-dimensional weight distribution and feature weighted fusion on the comprehensive feature representation space to obtain quality fusion features, and performing multi-level quality evaluation calculation and importance sorting on the quality fusion features based on a pre-set quality hierarchy to obtain a hierarchical quality evaluation result; Performing quality level determination and confidence cross-validation of a mobile phone shell use scenario on the hierarchical quality evaluation result to obtain a mobile phone shell quality level state, and performing abnormal mode recognition and pre-warning level adjustment of production batch information association on the mobile phone shell quality level state to generate an automatic detection result of the mobile phone shell to be detected.

2. The method of claim 1, wherein, The original multi-dimensional detection data includes visible light image original data, near-infrared spectrum original data, X-ray fluorescence original data, laser scanning original data, mechanical test original data, and electromagnetic performance original data, and the multiple data preprocessing on the original multi-dimensional detection data to obtain the multi-modal standardized data comprises: Performing intensity gradient calculation and direction distribution statistics of the mobile phone shell surface reflection area on the visible light image original data to obtain a reflection interference distribution map, and performing regional illumination compensation and local contrast enhancement based on the reflection interference distribution map to obtain a homogenized visible light image, and performing baseline drift detection and polynomial fitting correction on the near-infrared spectrum original data to obtain standardized spectrum data, and performing feature peak identification and component quantitative calculation on the standardized spectrum data to obtain material component feature data, and performing energy spectrum deconvolution and element quantitative calculation on the X-ray fluorescence original data to obtain element composition data, and performing point cloud filtering and coordinate system registration on the laser scanning original data to obtain geometric structure point cloud data, and performing temperature compensation and performance parameter standardization on the mechanical test original data to obtain mechanical performance data, and performing frequency domain transformation and feature extraction on the electromagnetic performance original data to obtain electromagnetic characteristic data; Performing edge detection and texture feature extraction of the mobile phone shell contour on the homogenized visible light image to obtain structured visual feature data, and performing timestamp alignment and data format standardization on the structured visual feature data, the material component feature data, the element composition data, the geometric structure point cloud data, the mechanical performance data, and the electromagnetic characteristic data to generate multi-modal standardized data.

3. The method of claim 2, wherein, The extraction of multi-dimensional detection features from the multi-modal standardized data to obtain a mobile phone shell multi-modal quality feature set comprises: The structured visual feature data in the multi-modal standardized data is subjected to two-way convolution defect feature extraction and multi-scale fusion to obtain an appearance defect feature map, and the appearance defect feature map is subjected to defect region segmentation and defect level calculation to obtain an appearance quality feature vector; The material composition feature data and element composition data in the multi-modal standardized data are subjected to harmful substance concentration calculation and material safety index calculation to obtain a material safety evaluation result, and the material safety evaluation result is subjected to risk level numerical mapping and safety coding to obtain a material safety feature vector; The geometric structure point cloud data in the multi-modal standardized data is subjected to hole opening precision measurement and surface quality calculation to obtain a geometric precision evaluation result, and the geometric precision evaluation result is subjected to assembly matching degree calculation and feature coding to obtain a geometric quality feature vector; The mechanical property data and electromagnetic characteristic data in the multi-modal standardized data are subjected to three-proofing performance quantification and wireless charging efficiency calculation to obtain a functional performance evaluation result, and the functional performance evaluation result is subjected to performance index standardization and feature coding to obtain a functional performance feature vector; The appearance quality feature vector, the material safety feature vector, the geometric quality feature vector, and the functional performance feature vector are subjected to feature dimension unification and unified coding to generate a mobile phone shell multi-modal quality feature set.

4. The method of claim 1, wherein, The multi-modal quality feature set is subjected to cross-domain feature association calculation and cross-domain mapping transformation between multiple quality feature dimensions to obtain a comprehensive feature representation space, including: The multi-modal quality feature set is subjected to feature domain division and association calculation of appearance and material association to obtain a cross-domain feature association matrix, and the geometric precision and functional performance are subjected to feature semantic alignment and dimension unification based on the cross-domain feature association matrix to obtain a multi-modal feature with unified quality attributes; The multi-modal feature with unified quality attributes is subjected to feature interaction calculation and synergistic effect calculation to obtain a feature synergistic effect representation, and the feature synergistic effect representation is subjected to weight adaptive calculation and weighted fusion of mobile phone shell user demand features to obtain a fused feature representation; The fused feature representation is subjected to high-dimensional mapping and spatial projection to obtain a high-dimensional feature representation, and the high-dimensional feature representation is used to construct a hierarchical feature representation structure to obtain a mobile phone shell multi-dimensional quality subspace; The mobile phone shell multi-dimensional quality subspace is subjected to unified integration and index construction of mobile phone shell quality to generate a comprehensive feature representation space.

5. The method of claim 1, wherein, The comprehensive feature representation space is subjected to multi-dimensional weight allocation and feature weighted fusion to obtain a quality fused feature, including: The comprehensive feature representation space is subjected to multi-head attention calculation of mobile phone shell appearance-material-geometry-function association to obtain an attention weight distribution of multi-dimensional features, and the attention weight distribution is subjected to feature interaction intensity calculation and synergistic effect calculation to obtain a feature synergistic weight matrix; The comprehensive feature representation space is subjected to weighted fusion and nonlinear transformation using the feature synergistic weight matrix to generate a quality-oriented fused feature representation.

6. The method of claim 1, wherein, The multi-level quality evaluation result is obtained by performing multi-level quality evaluation calculation and importance sorting on the quality fusion feature based on the preset quality hierarchy. The multi-level quality evaluation result is obtained by performing multi-level quality evaluation calculation and importance sorting on the quality fusion feature based on the preset quality hierarchy. The multi-level quality evaluation result is obtained by performing multi-level quality evaluation calculation and importance sorting on the quality fusion feature based on the preset quality hierarchy.

7. The method of claim 1, wherein, The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level determination result is obtained by performing multi-level quality classification calculation on the hierarchical quality evaluation result of the mobile phone shell based on the use scene demand parameter corresponding to the mobile phone shell to be detected and the preset adaptive classification model. The performance evaluation result is obtained by performing mobile phone shell protection performance matching degree calculation and use adaptability evaluation calculation on the quality level determination result, and the quality level reliability calculation and level confidence quantification are performed on the performance evaluation result to obtain the level confidence evaluation. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell.

8. The method of claim 1, wherein, The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell.

9. An automated detection system for mobile phone cases, the system comprising: The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality level state of the mobile phone shell is obtained by performing quality level determination and confidence cross-validation on the hierarchical quality evaluation result of the mobile phone shell. The quality evaluation module is configured to perform multi-dimensional weight distribution and feature weighting fusion on the comprehensive feature representation space to obtain quality fusion features, perform multi-level quality evaluation calculation and importance sorting on the quality fusion features based on a preset quality hierarchy, and obtain hierarchical quality evaluation results. The detection determination module is configured to perform quality level determination and confidence cross-validation of a mobile phone shell use scene on the hierarchical quality evaluation results, obtain a mobile phone shell quality level state, perform abnormal mode recognition and early warning level adjustment on the mobile phone shell quality level state in association with production batch information, and generate an automatic detection result of the mobile phone shell to be detected.

Images