Defect detection method and system for automobile rearview mirror

Abstract

The invention relates to the technical field of automobile part manufacturing, in particular to a defect detection method and system for an automobile rearview mirror, and the method comprises the steps: respectively collecting a mirror reflection signal of each sampling point of a mirror under a multispectral polarization condition, and obtaining the reflection response information of each sampling point according to the mirror reflection signal; a mirror surface detection environment is acquired, reflection response correction is performed on each sampling point based on the mirror surface detection environment, and a plurality of calibration reflection data sets are acquired; a specular reflection decomposition network is set, reflection information stripping is carried out on the calibration reflection data sets according to the specular reflection decomposition network, and a foreign matter reflection characteristic graph and a mirror film reflection deviation graph are obtained; and respectively carrying out abnormity judgment on the foreign matter reflection characteristic graph and the mirror film reflection deviation graph, and obtaining an automobile rearview mirror defect detection result according to a judgment result. According to the method, the technical limitation that foreign matter reflection interference and film reflection deviation are difficult to separate under a complex film structure in a traditional method is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of automotive parts manufacturing technology, and in particular to a method and system for detecting defects in automotive rearview mirrors. Background Technology

[0002] With the continuous improvement of intelligent manufacturing and automated testing in the automotive industry, rearview mirrors, as a key component of vehicle safety and driver assistance systems, directly affect the driver's visibility and driving safety through their surface quality. Modern rearview mirrors typically employ multi-layered film structures, including high-reflectivity metallic films and anti-glare films, to enhance reflectivity and resistance to environmental interference. During the rearview mirror manufacturing process, the surface is highly susceptible to the adhesion of foreign matter such as oil, dust, and water stains, as well as defects such as uneven film thickness and localized peeling. These problems not only affect the optical performance of the rearview mirror but may also lead to safety hazards.

[0003] Existing methods for inspecting automotive rearview mirrors typically rely on observation or simple reflection intensity measurements to determine surface quality. However, these methods often suffer from low inspection efficiency, high subjectivity, and incomplete defect identification when faced with various types of foreign objects, complex film structures, and the demands of large-scale automated production. They fail to meet the requirements of high-quality, high-consistency, and intelligent inspection of modern automotive rearview mirrors. In particular, under multi-spectral and polarization-based multi-dimensional inspection conditions, effectively separating foreign objects from film reflection characteristics has become one of the core technologies for improving the level of automated inspection and product quality of rearview mirrors.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of this disclosure and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a method and system for detecting defects in automotive rearview mirrors, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for detecting defects in a car rearview mirror, the method comprising: The mirror reflection signals under multispectral polarization conditions at each sampling point on the mirror surface are collected respectively, and the reflection response information of each sampling point is obtained based on the mirror reflection signals. Obtain the mirror detection environment, and perform reflection response correction on each sampling point based on the mirror detection environment to obtain several calibration reflection datasets; A specular reflection decomposition network is established, and the reflection information of several calibration reflection datasets is stripped according to the specular reflection decomposition network to obtain foreign object reflection feature map and mirror film reflection deviation map. The foreign object reflection feature map and the mirror film reflection deviation map are respectively judged for anomalies, and the defect detection results of the car rearview mirror are obtained based on the judgment results.

[0007] Further, the reflection information of several calibration reflection datasets is stripped according to the specular reflection decomposition network to obtain foreign object reflection feature maps and mirror film reflection deviation maps, including: A comprehensive mirror inspection database is established based on historical mirror production information. The comprehensive mirror inspection database includes the mirror film reference reflection characteristics and foreign object reflection reference characteristics of several types of foreign objects. Based on the reference reflectance characteristics of the lens film, a global lens film difference analysis is performed on the calibration reflectance dataset of each sampling point to obtain a comprehensive lens film difference degree. The global lens film difference analysis is used to evaluate the overall deviation of each sampling point from the reference reflectance characteristics of the lens film. For sampling points where the overall mirror film difference is greater than a preset difference threshold, the calibration reflection dataset and several foreign object reflection reference features are respectively subjected to local foreign object difference analysis to obtain the foreign object reflection difference. Based on the combined lens film difference degree and foreign object reflection difference degree, the sampling points are decoupled by feature layering to obtain the foreign object reflection feature map and the lens film reflection deviation map.

[0008] Further, feature layer decoupling is performed on the sampling points to obtain the foreign object reflection feature map and the mirror film reflection deviation map, including: S1: Establish a mirror film reflection feature prediction model based on the mirror film reference reflection features, adaptively adjust the mirror film thickness, and use the adjusted mirror film thickness as input based on the mirror film reflection feature prediction model to output several theoretical compensation reflection curves. S2: Extract the measured reflection curves from the calibration reflection dataset, select the theoretical reflection curves that meet the preset fitting conditions of the measured reflection curves, calculate the film thickness compensation residual, and calculate the residual fitting value based on the film thickness compensation residual and the foreign object reflection reference characteristics. Determine whether the residual fitting value is less than a preset threshold. If not, use the foreign object reflection reference feature to perform foreign object compensation on the measured reflection curve. The foreign object compensation step refers to the film thickness compensation steps S1 and S2. Alternate between film thickness compensation and foreign object compensation until the residual fitting value meets the preset threshold or reaches the preset number of iterations. Record the fitting thickness and foreign object characteristics of each sampling point, and draw the foreign object reflection feature map and the mirror film reflection deviation map.

[0009] Furthermore, the lens coating thickness is adaptively adjusted, including: The average reflection error is calculated based on the extracted measured reflection curves and several of the aforementioned mirror film reference reflection characteristics, and the interpolation thickness range is obtained based on the aforementioned average reflection errors. The difference step size is set according to the interpolation thickness range, and several step points are obtained. The reflection interpolation of several step points is calculated based on the mirror detection environment to obtain several interpolation curves. The fitting error between the interpolation curves and the measured reflection curve is calculated based on the interpolation curves. The corresponding lens thickness with the smallest fitting error is identified and adjusted based on the fitting errors.

[0010] Furthermore, a comprehensive mirror testing database will be established based on historical mirror production information, including: Historical mirror reflection information and historical foreign object reflection information are extracted based on historical mirror production information. The historical mirror reflection information includes mirror film reflection information of several mirror detection environments, and the historical foreign object reflection information includes foreign object reflection information of several types of foreign objects. The mirror film reflection information is mapped to environmental detection conditions, foreign object type and foreign object reflection information respectively; A combined index relationship is created based on the environmental detection conditions and foreign object types, and the environmental detection conditions and foreign object types are classified and managed according to the combined index relationship. A comprehensive mirror detection database is constructed based on the mapping relationship and the combined index relationship.

[0011] Furthermore, based on the aforementioned mirror detection environment, reflection response correction is performed on each sampling point to obtain several calibration reflection datasets, including: Based on the empty rearview mirror state, multi-band spectral baseline data and polarization baseline data at different polarization angles were collected; Spectral correction coefficients are calculated based on the spectral baseline data, and a polarization correction mapping matrix is ​​constructed based on the polarization baseline data; The specular reflection signal is spectrally and polarized based on the spectral correction coefficients and polarization correction mapping matrix to form a calibration reflection dataset.

[0012] Further, the specular reflection signal is subjected to spectral and polarization correction based on the spectral correction coefficients and polarization correction mapping matrix to form a calibration reflection dataset, including: The specular reflection signal is weighted based on the spectral correction coefficients corresponding to the wavelength to obtain the spectral correction signal; For each sampling point and each sampling wavelength, the spectral correction signals with different polarization angles are combined to obtain a polarization signal vector; The polarization signal vector is linearly transformed with the polarization correction mapping matrix corresponding to the wavelength to obtain the polarization correction signal; Several polarization correction signals are obtained according to several mirror detection environments, and the several polarization correction signals are aggregated to obtain a calibration reflection dataset.

[0013] Furthermore, anomaly determinations are made on the foreign object reflection feature map and the mirror film reflection deviation map, respectively, and the rearview mirror defect detection results are obtained based on the determination results, including: Based on the foreign object reflection feature map and the mirror film reflection deviation map, the foreign object feature parameters and film thickness deviation values ​​are extracted, and compared with their respective preset thresholds to determine the foreign object contamination area and the film thickness abnormal area. Based on the foreign matter contamination area and the film thickness anomaly area, respectively obtain the contamination anomaly mask and the film thickness anomaly mask, and then perform spatial pixel superposition of the contamination anomaly mask and the film thickness anomaly mask; The abnormal defect type is identified based on the spatial pixel overlay result, and the rearview mirror defect detection result is obtained.

[0014] A defect detection system for automotive rearview mirrors, the system comprising: The signal acquisition module collects the mirror reflection signals under multispectral polarization conditions at each sampling point on the mirror surface, and obtains the reflection response information of each sampling point based on the mirror reflection signals. The response correction module acquires the mirror detection environment, performs reflection response correction on each sampling point based on the mirror detection environment, and acquires several calibration reflection datasets. The reflection decomposition module establishes a specular reflection decomposition network, and performs reflection information stripping on several calibration reflection datasets based on the specular reflection decomposition network to obtain foreign object reflection feature maps and mirror film reflection deviation maps. The anomaly detection module performs anomaly detection on the foreign object reflection feature map and the mirror film reflection deviation map, and obtains the defect detection results of the car rearview mirror based on the detection results.

[0015] Furthermore, the response correction module includes: The data acquisition unit, based on the empty rearview mirror state, acquires multi-band spectral baseline data and polarization baseline data at different polarization angles; The mapping matrix unit calculates spectral correction coefficients based on spectral baseline data and constructs a polarization correction mapping matrix based on polarization baseline data; The data generation unit performs spectral and polarization correction on the specular reflection signal based on the spectral correction coefficient and polarization correction mapping matrix to form a calibrated reflection dataset.

[0016] The technical solution of this invention can achieve the following technical effects: It simultaneously achieves high-precision automated detection of defects such as foreign object adhesion and uneven film thickness, overcoming the technical limitations of traditional methods in separating foreign object reflection interference and film reflection deviation under complex film structures.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a defect detection method for automotive rearview mirrors. Figure 2 A flowchart illustrating the process of obtaining the foreign object reflection feature map and the mirror film reflection deviation map; Figure 3 A flowchart illustrating the process of obtaining the calibration reflection dataset; Figure 4 A flowchart illustrating the process of obtaining rearview mirror defect detection results. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] Example 1; like Figure 1 As shown, this application provides a defect detection method for automotive rearview mirrors, the method comprising: S10: Collect the mirror reflection signal at each sampling point of the mirror under multispectral polarization conditions, and obtain the reflection response information of each sampling point based on the mirror reflection signal; S20: Obtain the mirror detection environment, perform reflection response correction on each sampling point based on the mirror detection environment, and obtain several calibration reflection datasets; S30: Establish a specular reflection decomposition network, and extract reflection information from several calibration reflection datasets based on the specular reflection decomposition network to obtain foreign object reflection feature map and mirror film reflection deviation map. S40: Perform anomaly judgment on the foreign object reflection feature map and the mirror film reflection deviation map respectively, and obtain the defect detection results of the car rearview mirror based on the judgment results.

[0023] Specifically, during the detection process, a mirror detection environment with a controllable light source is first constructed. This environment provides illumination conditions with multiple spectral combinations and adjustable polarization angles to simulate the impact of complex external lighting changes on mirror reflection. Before the detection begins, the mirror is positioned using a high-precision three-axis motorized slide, and multiple sampling points are set on the mirror at regular intervals. These sampling points cover the entire mirror area, ensuring comprehensive acquisition of reflection signals. Subsequently, using a multispectral polarization imaging device, the mirror reflection signal of each sampling point is acquired in multiple spectral channels and at different polarization angles. By recording both the intensity and polarization direction of these reflection signals, a reflection response dataset containing three-dimensional features of space, spectrum, and polarization is formed. The acquired reflection response signals are then input into the mirror detection environment, including environmental parameters such as temperature, humidity, background illumination, and mirror curvature. These parameters are dynamically monitored and recorded by sensors for subsequent reflection response correction. During the correction process, a combination of physical modeling and machine learning is used. The method involves normalizing the reflection response of each sampling point to account for environmental factors, generating a calibration reflection dataset that has eliminated interference from the detection environment. To further extract target defect features from the calibration data, this embodiment establishes a specular reflection decomposition network. This network adopts a dual-channel convolutional neural network structure, with one channel used to extract potential foreign object reflection features and the other channel used to fit the uniform reflection pattern of the mirror coating region and its deviation, thereby outputting a foreign object reflection feature map and a mirror coating reflection deviation map, respectively. This network is pre-trained with a large amount of data containing labeled defect samples and combines a self-supervised learning mechanism to enhance the features of unlabeled samples, giving it good generalization ability. In the output foreign object reflection feature map, anomaly judgment is made based on features such as the reflection intensity gradient change and edge contour shape of the foreign object to determine whether there is foreign object contamination such as dust, water spots, or oil stains. In the mirror coating reflection deviation map, the presence of defects in the mirror coating is determined by detecting large-area reflection anomalies, local deformation, coating peeling, and other phenomena. The system performs a comprehensive analysis based on the two judgment results mentioned above, and finally outputs the detection results such as the type, location, and severity of the rearview mirror defects. For example, in a real detection scenario, after multi-channel sampling of the surface of a rearview mirror, it was found that the lower left area showed an abnormal reflection peak under a polarization of 45 degrees. Through reflection decomposition network analysis, this area was determined to be a water stain. At the same time, an area that continuously deviates from the normal reflection curve was detected in the center area of ​​the mirror surface, which was determined to be local aging of the mirror film. Finally, the system outputs a comprehensive report based on the detection model, indicating that the rearview mirror needs to be cleaned and partially replaced.

[0024] The technical solution of this invention enables high-precision automated detection of defects such as foreign object adhesion and uneven film thickness, overcoming the technical limitations of traditional methods in separating foreign object reflection interference and film reflection deviation under complex film structures.

[0025] Furthermore, such as Figure 2 As shown, reflection information is stripped from several calibration reflection datasets using a specular reflection decomposition network to obtain foreign object reflection feature maps and mirror film reflection deviation maps, including: A comprehensive mirror inspection database was established based on historical mirror production information. The comprehensive mirror inspection database includes the mirror film reference reflection characteristics and the foreign object reflection reference characteristics of several types of foreign objects. Based on the reference reflectance characteristics of the lens film, a global lens film difference analysis is performed on the calibration reflectance dataset of each sampling point to obtain the comprehensive lens film difference degree. The global lens film difference analysis is used to evaluate the overall deviation of each sampling point from the reference reflectance characteristics of the lens film. For sampling points where the overall mirror film difference is greater than the preset difference threshold, the calibration reflection dataset and several foreign object reflection reference features are respectively subjected to local foreign object difference analysis to obtain the foreign object reflection difference. Based on the combined lens film difference and foreign object reflection difference, the sampling points are decoupled by feature layering to obtain the foreign object reflection feature map and the lens film reflection deviation map.

[0026] As a preferred embodiment of the above, a comprehensive mirror inspection database is first constructed. This database is built based on historical mirror production and quality inspection data, specifically including a large number of mirror film reference reflection characteristics and foreign object reflection reference characteristics corresponding to various typical foreign objects. The mirror film reference reflection characteristics are obtained by feature normalization and statistical modeling of large-scale imaging data of defect-free mirror samples under standard illumination and polarization conditions, possessing high stability and representativeness. The foreign object reflection reference characteristics are derived from labeled defect samples collected in actual production and are classified and processed separately to enable them to identify the difference characteristics of various foreign objects. During the inspection process, a first-stage global mirror film difference analysis is performed on the calibration reflection dataset collected at each sampling point. This analysis compares the data with the mirror film reference reflection characteristics in the database dimension by dimension, comprehensively considering the deviation of each sampling point in dimensions such as multispectral, polarization intensity, and reflection direction, to generate a comprehensive mirror film for evaluating reflection consistency. To ensure the accuracy of the analysis results, a weighted average deviation algorithm was used, combined with a polarization stability factor and a reflection curvature model, to distinguish between natural differences caused by changes in the lens coating material and heterogeneity caused by abnormal reflection. For sampling points with a comprehensive lens coating difference exceeding a set threshold, the analysis proceeded to the second stage: local foreign object difference analysis. In this stage, the calibrated reflection data of the point to be analyzed was matched with the reflection benchmark features of each type of foreign object in the database for similarity matching and local feature fitting, thereby obtaining the reflection difference index corresponding to each type of foreign object. During this process, a convolutional feature enhancement network was used to compare the texture morphology of local reflection modes, abrupt edges in the polarization angle domain, and color deviations in multispectral channels, further distinguishing between sudden anomalies caused by foreign object attachment and continuous anomalies caused by lens coating material degradation. Finally, the comprehensive lens coating difference and foreign object reflection difference were combined to perform feature layer decoupling on the sampling points. This decoupling process employs a two-layer feature fusion map mechanism: the first layer maps reflection data to a mirror film deviation heatmap, showing the overall trend of mirror film condition changes; the second layer projects foreign object difference indicators to a foreign object distribution map, clearly indicating the location and type of contaminated areas. These two maps together constitute the final output foreign object reflection feature map and mirror film reflection deviation map. For example, in a certain inspection, multiple sampling points with excessive overall mirror film difference were identified on a certain mirror surface. The reflection data in the upper right area of ​​the mirror surface highly matched the oil spot baseline characteristics, and its foreign object reflection difference was significantly higher than other types, thus it was determined to be oil spot contamination. Although the lower left area of ​​the mirror surface did not have obvious foreign object similarity characteristics, it continuously deviated from the mirror film baseline reflection characteristics, and was ultimately classified as a mirror film aging area. This layered decoupling mechanism not only improves the resolution of mixed defects, but also provides a basis for subsequent quantitative defect assessment and repair recommendations.

[0027] Furthermore, feature layer decoupling is performed on the sampling points to obtain the foreign object reflection feature map and the mirror film reflection deviation map, including: S1: Establish a mirror film reflection feature prediction model based on the mirror film reference reflection characteristics, adaptively adjust the mirror film thickness, take the adjusted mirror film thickness as input based on the mirror film reflection feature prediction model, and output several theoretical compensation reflection curves. S2: Extract the measured reflection curves from the calibration reflection dataset, select the theoretical reflection curves that meet the preset fitting conditions of the measured reflection curves, calculate the film thickness compensation residual, and calculate the residual fitting value based on the film thickness compensation residual and the foreign object reflection reference characteristics. Determine whether the residual fitting value is less than the preset threshold. If not, use the foreign object reflection reference feature to perform foreign object compensation on the measured reflection curve. The foreign object compensation step refers to the film thickness compensation steps S1 and S2. Alternate between film thickness compensation and foreign object compensation until the residual fitting value meets the preset threshold or reaches the preset number of iterations. Record the fitting thickness and foreign object characteristics of each sampling point, and draw the foreign object reflection feature map and the mirror film reflection deviation map.

[0028] As a preferred embodiment of the above, firstly, a mirror film reflection feature prediction model is established. This model adaptively adjusts the mirror film thickness based on the mirror film reference reflection features. In this step, the adjusted mirror film thickness is input to output multiple theoretical compensation reflection curves. The key is to establish a prediction model that can dynamically respond to different mirror film thickness changes, thereby ensuring the accuracy and reliability of the reflection curve data. Next, measured reflection curves are extracted from the calibration reflection dataset, and theoretical reflection curves that meet the preset fitting conditions are selected. Based on this, the film thickness compensation residual is calculated, and the fitting value with the foreign object reflection reference feature is calculated using this residual. This process can be implemented through an iterative calculation model to ensure the accuracy and stability of the residual fitting value. The calculation of the film thickness compensation residual involves... The measured data is matched with the theoretical reflection curve to obtain the deviation data. When judging whether the residual fitting value is less than the preset threshold, if the residual fitting value exceeds the threshold, foreign object compensation is performed. During foreign object compensation, the previous film thickness compensation step is referred to ensure that the foreign object features can be correctly identified and compensated in the reflection curve. In order to further improve the accuracy of compensation, an adaptive compensation algorithm can be introduced in this step to improve the accuracy of detection by dynamically adjusting the compensation amount. Finally, film thickness compensation and foreign object compensation are alternately performed until the residual fitting value meets the preset threshold or reaches the preset number of iterations. At this time, the fitting thickness and foreign object features of each sampling point need to be recorded in detail, and the results are plotted graphically as a foreign object reflection feature map and a mirror film reflection deviation map to provide an effective reference for maintenance and quality control.

[0029] Furthermore, adaptive adjustment of the lens coating thickness includes: The average reflection error is calculated based on the extracted measured reflection curves and several mirror film reference reflection characteristics, and the interpolation thickness range is obtained based on the several average reflection errors. The difference step size is set according to the interpolation thickness range, and several step points are obtained. Based on the mirror detection environment, the reflection interpolation of several step points is calculated respectively, and several interpolation curves are obtained. The fitting error between the interpolation curve and the measured reflection curve is calculated based on several interpolation curves. The corresponding lens thickness with the smallest fitting error is then identified and adjusted.

[0030] As a preferred embodiment of the above, firstly, the measured reflection curve is extracted through calculation, and the average reflection error is calculated by comparing it with the mirror film reference reflection characteristics. During the calculation of the average reflection error, multiple different mirror film reference reflection characteristics are selected for comparison to ensure the representativeness of the error results. Based on the calculation of all average reflection errors, an interpolation thickness range is obtained. The definition of the interpolation thickness range should reflect the variation range of the mirror film thickness to ensure the accuracy of subsequent interpolation calculations. After obtaining the interpolation thickness range, a difference step size is set to finely define several step points. The setting of the step points should adapt to the accuracy requirements of subsequent reflection interpolation calculations. The determination of the difference step size needs to consider the balance between computational efficiency and accuracy. Then, based on the determined step points, the reflection interpolation of these points is calculated in a mirror detection environment. Here, the mirror detection environment refers to the lighting and angle conditions of the rearview mirror under normal use conditions. Data obtained through real-scene testing is used to calibrate the measurement. Interpolation is performed to generate multiple interpolation curves. Next, by comparing with the measured reflection curves, the fitting error of each interpolation curve is calculated. The calculation mechanism of the fitting error provides an adaptive analysis for different thicknesses of the mirror film. Each error curve indicates the possible thickness deviation of the mirror film. After the fitting error is calculated, the corresponding mirror film thickness with the smallest error is identified and the thickness is adjusted. This identification process relies on the sorting and selection of fitting errors to determine the optimal thickness in an automated manner. Suppose that a mirror film reflection characteristic is obtained during the detection process, and the average error obtained from the preliminary calculation shows that the thickness may be uneven. Then, a thickness range is drawn based on this error. By calculating the light reflection interpolation at the step point, the final light wave data can be displayed as a series of curves, namely interpolation curves. By comparing these curves with the measured curves, the thickness with the smallest error is obtained, and the mirror film is precisely adjusted to ensure the normal working condition of the rearview mirror.

[0031] Furthermore, 5. The method for defect detection of automotive rearview mirrors according to claim 2, characterized in that a comprehensive mirror inspection database is established based on historical mirror production information, including: Historical mirror reflection information and historical foreign object reflection information are extracted based on historical mirror production information. The historical mirror reflection information includes mirror film reflection information of several mirror detection environments, and the historical foreign object reflection information includes foreign object reflection information of several types of foreign objects. The mirror film reflection information is mapped to environmental detection conditions, foreign object type and foreign object reflection information respectively; A combined index relationship is created based on the environmental detection conditions and foreign object types, and the environmental detection conditions and foreign object types are classified and managed according to the combined index relationship. A comprehensive mirror detection database is constructed based on the mapping relationship and the combined index relationship.

[0032] As a preferred embodiment of the above embodiments, to establish a comprehensive mirror inspection database, it is first necessary to extract relevant data based on historical mirror production information. This data includes historical mirror reflection information and historical foreign object reflection information. Historical mirror reflection information covers mirror film reflection information in multiple mirror inspection environments, while historical foreign object reflection information includes reflection characteristic data of several foreign object types. This data originates from inspection records accumulated during mirror production and an information database of anomaly handling. For example, representative reflection data is extracted from inspection maps of different lighting conditions, different mirror film thicknesses, or foreign object areas. Next, the extracted mirror film reflection information is correlated with environmental inspection conditions, and the foreign object type is correlated with foreign object reflection information. The core of this process is to construct a many-to-many mapping relationship based on reflection characteristics. Specifically, the mapping relationship for mirror film reflection information needs to consider the dependence between mirror thickness changes, the inspection environment, and the reflection results, while the mapping relationship for foreign object information focuses on associating the morphological characteristics of different foreign object types with their reflection patterns. For example, foreign objects may include dust, stains, and scratches. The reflective characteristics of different foreign objects can be represented through modeling and data classification. Then, based on environmental detection conditions and foreign object types, a combined index relationship is created. The purpose of the combined index relationship is to efficiently aggregate environmental conditions and foreign object reflection information, facilitating subsequent classification and management. For example, through index rules, the reflective features of mirrors with different light angles can be classified into corresponding environmental categories. At the same time, based on the size or shape of the foreign object, the reflective features of the foreign object can be classified into corresponding foreign object type categories. The index creation can adopt a weighted scoring mechanism to intelligently classify based on the complexity of the detection environment and the image characteristics of the foreign object type. Finally, a comprehensive mirror detection database is constructed based on the above mapping relationship and combined index relationship. In the process of database construction, data storage efficiency and information retrieval convenience are given priority. The database can be based on a relational design, storing mirror reflection features, detection environmental conditions, and foreign object information as multiple interconnected tables. The tables are logically connected through combined indexes. Through this data structure design, specific foreign object reflection features in the corresponding environment can be quickly retrieved, and mirror condition assessment can be performed.

[0033] Furthermore, such as Figure 3 As shown, reflection response correction is performed on each sampling point based on the mirror detection environment, and several calibration reflection datasets are obtained, including: Based on the empty rearview mirror state, multi-band spectral baseline data and polarization baseline data at different polarization angles were collected; Spectral correction coefficients are calculated based on spectral baseline data, and a polarization correction mapping matrix is ​​constructed based on polarization baseline data; The specular reflection signal is spectrally and polarized based on the spectral correction coefficients and polarization correction mapping matrix to form a calibrated reflection dataset.

[0034] As a preferred embodiment of the above, firstly, with the rearview mirror in an empty mirror state (i.e., a standard state without external influence), multi-band spectral baseline data and polarization baseline data at different polarization angles are collected. The spectral baseline data collection needs to cover multiple bands, such as visible light, near-infrared, and ultraviolet bands, to capture the full-band spectral reflectance characteristics of the rearview mirror. The polarization baseline data is obtained by measuring the mirror reflection signal based on different polarization angles to reflect the optical behavior of the mirror under polarization conditions. The above data is collected using professional testing equipment to ensure data accuracy and reliability. Next, spectral correction coefficients are calculated based on the spectral baseline data. Obtaining the spectral correction coefficients requires analyzing the reflection intensity of the mirror in different bands and identifying the variation law of its band characteristics to obtain the correction coefficients for each band. These coefficients are used to correct the spectral errors of mirror reflection in actual testing, such as signal shifts caused by inconsistent band reflections. At the same time, based on the polarization baseline data collected at different polarization angles, a polarization correction mapping is constructed through data modeling and algorithm processing. A calibration matrix is ​​used to correct for reflection intensity or direction deviations in specular reflection signals caused by changes in polarization angle. A preferred approach is to perform multi-angle fitting on different polarization baseline data to generate a calibration matrix with high universality and accuracy. Then, using the obtained spectral correction coefficients and polarization correction mapping matrix, reflection signals under all specular inspection environments are corrected. Specifically, firstly, the spectral correction coefficients are used to perform spectral compensation on the detected specular reflection data to adjust the reflection deviation in each band and unify its baseline. Secondly, the polarization correction mapping matrix is ​​applied to correct the polarization of the reflection signal, making it conform to standard optical reflection laws. Both correction processes can be implemented modularly, preferably using a data processing unit combined with algorithm control for rapid response correction. Finally, the corrected data forms a calibration reflection dataset, which includes spectral and polarization correction information for all sampling points, possessing high accuracy and uniformity. The calibration reflection dataset can provide a reference benchmark for subsequent automotive rearview mirror defect detection, ensuring the accuracy of the detection results.

[0035] Furthermore, based on the spectral correction coefficients and polarization correction mapping matrix, the specular reflection signal is subjected to spectral and polarization correction to form a calibrated reflection dataset, including: The specular reflection signal is weighted based on the spectral correction coefficients of the corresponding wavelength to obtain the spectral correction signal; For each sampling point and each sampling wavelength, the spectral correction signals with different polarization angles are combined to obtain the polarization signal vector; The polarization correction signal is obtained by linearly transforming the polarization signal vector with the polarization correction mapping matrix of the corresponding wavelength. Several polarization correction signals are acquired based on several mirror detection environments, and these polarization correction signals are aggregated to obtain a calibration reflection dataset.

[0036] As a preferred embodiment of the above, firstly, the specular reflection signal is weighted based on the spectral correction coefficients of the corresponding wavelength to obtain a spectral correction signal. This ensures the accurate calculation of the correction coefficients for each sampling wavelength and applies them to the processing of the specular reflection signal, thereby correcting the inconsistencies of each wavelength. The sampling wavelengths typically cover the visible to infrared light spectrum, and the intensity of the reflected signal in these spectra reflects the performance of the mirror in different lighting environments. Next, for each sampling point and sampling wavelength, the spectral correction signals at different polarization angles are merged to form a polarization signal vector. This process requires integrating the spectral correction signals collected at each sampling point under different polarization angles. The resulting vector reflects the comprehensive performance of the reflected signal under multiple polarization conditions. For example, the correction signals of the same sampling point at 0 degrees, 30 degrees, and 60 degrees of polarization angle are merged to obtain the polarization signal vector of that point in this environment. A preferred merging strategy can be based on weighted averaging or best fitting methods to ensure the true validity of the polarization signal vector. Subsequently, based on the polarization correction mapping matrix of each corresponding wavelength, the polarization signal vector is... A linear transformation is performed to obtain a polarization correction signal. This linear transformation allows for the correction of signal distortion under different polarization conditions using a polarization mapping matrix, closely following the polarization variation of the main baseline signal. This transformation ensures that the polarization reflection characteristics of different wavelengths can be accurately corrected to a standard state, enabling the instrument to provide stable test data. Subsequently, multiple sets of polarization correction signals are extracted according to different mirror testing environments. These testing environments refer to the test scenarios of the rearview mirror under different physical conditions. The obtained polarization correction signals under each testing environment need to be aggregated and analyzed to establish stable reflection performance. If the test signal in a certain environment is abnormal, external interference should be considered, and appropriate data filtering and recalibration should be performed. Finally, by separately acquiring and aggregating the polarization correction signals from each environment, a complete calibration reflection dataset is formed. This dataset integrates the calibration data of the rearview mirror under various environmental conditions, ensuring that the test results of the rearview mirror in practical applications are reliable and consistent. This calibration dataset can be used as benchmark data for reflection characteristics in subsequent testing stages, optimizing the quality control and defect identification of the rearview mirror.

[0037] Furthermore, such as Figure 4 As shown, anomalies are determined by examining the foreign object reflection feature map and the mirror film reflection deviation map, respectively. Based on the determination results, the rearview mirror defect detection results are obtained, including: Based on the foreign object reflection feature map and the mirror film reflection deviation map, the foreign object feature parameters and film thickness deviation values ​​are extracted, and compared with their respective preset thresholds to determine the foreign object contamination area and the film thickness abnormal area. Based on the foreign object contamination area and the film thickness anomaly area, respectively obtain the contamination anomaly mask and the film thickness anomaly mask, and then perform spatial pixel superposition of the contamination anomaly mask and the film thickness anomaly mask. The abnormal defect type is identified based on the spatial pixel overlay result, and the rearview mirror defect detection result is obtained.

[0038] As a preferred embodiment of the above, firstly, corresponding feature parameters are extracted based on the foreign object reflection feature map and the mirror film reflection deviation map. For the foreign object reflection feature map, the extracted feature parameters include the area, position, reflection intensity variation, and shape features of the foreign object. For the mirror film reflection deviation map, the film thickness deviation value and its distribution are extracted. These feature parameters are calculated from the feature map using image processing algorithms, for example, edge detection and region segmentation algorithms are used to determine the boundary, and then regional intensity variation data are extracted. After the parameter extraction is completed, the foreign object feature parameters and the mirror film deviation value are compared with preset thresholds to determine whether there is a foreign object. For anomalies, the threshold settings should be combined with statistical data from normal mirror production. For example, the preset threshold for foreign object area can consider user requirements for the rearview mirror's appearance, while the threshold for film thickness deviation is related to optical performance requirements. This comparison method can accurately determine whether there are foreign object contamination areas or film thickness anomaly areas on the rearview mirror surface. Furthermore, based on the identified foreign object contamination areas and film thickness anomaly areas, contamination anomaly masks and film thickness anomaly masks are generated respectively. The contamination anomaly mask is used to mark the foreign object contamination portion and record its spatial location, while the film thickness anomaly mask identifies areas where the film thickness exceeds the normal range. When generating the masks... Abnormal areas in the image can be binarized, with areas containing foreign objects or exhibiting abnormal film thickness marked as high-value pixels and normal areas marked as low-value pixels. This allows the spatial location and information within the mask to clearly represent the distribution of abnormal areas. Subsequently, the contamination anomaly mask and the film thickness anomaly mask are spatially superimposed at the pixel level to form a superimposed overall mask image. This spatial pixel superposition operation enables comprehensive analysis of abnormal areas on the rearview mirror surface. For example, when both foreign object contamination and film thickness anomaly are present at a certain location, the superimposed pixel value at that location will be marked as an area with a combined defect. Conversely, if a certain location... If only one problem occurs, it is reflected as a single defect. During the pixel overlay process, positional weights can be added to preferentially consider the severity and scope of the defect. Finally, the defect type is identified through the overlay result and a rearview mirror defect detection result is generated. For example, if the overlay mask shows a distribution of high-value pixels in a large continuous area, it may indicate a large structural defect. If the high-value pixels are mainly concentrated in the edge area, it may be contamination introduced during installation or use. Furthermore, the defect detection result can also provide detailed classification information, such as the type of foreign object, the size and location distribution of the abnormal area, etc.

[0039] Example 2; Based on the same inventive concept as the defect detection method for a car rearview mirror in the foregoing embodiments, the present invention also provides a defect detection system for a car rearview mirror, the system comprising: The signal acquisition module collects the mirror reflection signals under multispectral polarization conditions at each sampling point on the mirror surface, and obtains the reflection response information of each sampling point based on the mirror reflection signals. The response correction module acquires the mirror detection environment, performs reflection response correction on each sampling point based on the mirror detection environment, and acquires several calibration reflection datasets. The reflection decomposition module establishes a specular reflection decomposition network, and performs reflection information stripping on several calibration reflection datasets based on the specular reflection decomposition network to obtain foreign object reflection feature maps and mirror film reflection deviation maps. The anomaly detection module performs anomaly detection on the foreign object reflection feature map and the mirror film reflection deviation map, and obtains the defect detection results of the car rearview mirror based on the detection results.

[0040] The adjustment system described above in this invention can effectively realize a method for detecting defects in automotive rearview mirrors, and the technical effects it can achieve are as described in the above embodiments, and will not be repeated here.

[0041] Furthermore, the response correction module includes: The data acquisition unit, based on the empty rearview mirror state, acquires multi-band spectral baseline data and polarization baseline data at different polarization angles; The mapping matrix unit calculates spectral correction coefficients based on spectral baseline data and constructs a polarization correction mapping matrix based on polarization baseline data; The data generation unit performs spectral and polarization correction on the specular reflection signal based on the spectral correction coefficient and polarization correction mapping matrix to form a calibrated reflection dataset.

[0042] Similarly, the above-mentioned optimization schemes for the system can also achieve the optimization effects corresponding to the methods in Embodiment 1, which will not be repeated here.

[0043] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for detecting defects in a car rearview mirror, characterized in that, The method includes: The mirror reflection signals under multispectral polarization conditions at each sampling point on the mirror surface are collected respectively, and the reflection response information of each sampling point is obtained based on the mirror reflection signals. Obtain the mirror detection environment, and perform reflection response correction on each sampling point based on the mirror detection environment to obtain several calibration reflection datasets; A specular reflection decomposition network is established, and the reflection information of several calibration reflection datasets is stripped according to the specular reflection decomposition network to obtain foreign object reflection feature map and mirror film reflection deviation map. The foreign object reflection feature map and the mirror film reflection deviation map are respectively judged for anomalies, and the defect detection results of the car rearview mirror are obtained based on the judgment results.

2. The defect detection method for automobile rearview mirrors according to claim 1, characterized in that, The specular reflection decomposition network is used to strip reflection information from several calibration reflection datasets to obtain foreign object reflection feature maps and mirror film reflection deviation maps, including: A comprehensive mirror inspection database is established based on historical mirror production information. The comprehensive mirror inspection database includes the mirror film reference reflection characteristics and foreign object reflection reference characteristics of several types of foreign objects. Based on the reference reflectance characteristics of the lens film, a global lens film difference analysis is performed on the calibration reflectance dataset of each sampling point to obtain a comprehensive lens film difference degree. The global lens film difference analysis is used to evaluate the overall deviation of each sampling point from the reference reflectance characteristics of the lens film. For sampling points where the overall mirror film difference is greater than a preset difference threshold, the calibration reflection dataset and several foreign object reflection reference features are respectively subjected to local foreign object difference analysis to obtain the foreign object reflection difference. Based on the combined lens film difference degree and foreign object reflection difference degree, the sampling points are decoupled by feature layering to obtain the foreign object reflection feature map and the lens film reflection deviation map.

3. The method for detecting defects in a car rearview mirror according to claim 2, characterized in that, The sampling points are decoupled by feature layering to obtain the foreign object reflection feature map and the mirror film reflection deviation map, including: S1: Establish a mirror film reflection feature prediction model based on the mirror film reference reflection features, adaptively adjust the mirror film thickness, and use the adjusted mirror film thickness as input based on the mirror film reflection feature prediction model to output several theoretical compensation reflection curves. S2: Extract the measured reflection curves from the calibration reflection dataset, select the theoretical reflection curves that meet the preset fitting conditions of the measured reflection curves, calculate the film thickness compensation residual, and calculate the residual fitting value based on the film thickness compensation residual and the foreign object reflection reference characteristics. Determine whether the residual fitting value is less than a preset threshold. If not, use the foreign object reflection reference feature to perform foreign object compensation on the measured reflection curve. The foreign object compensation step refers to the film thickness compensation steps S1 and S2. Alternate between film thickness compensation and foreign object compensation until the residual fitting value meets the preset threshold or reaches the preset number of iterations. Record the fitting thickness and foreign object characteristics of each sampling point, and draw the foreign object reflection feature map and the mirror film reflection deviation map.

4. The method for detecting defects in a car rearview mirror according to claim 3, characterized in that, Adaptive adjustment of the lens coating thickness, including: The average reflection error is calculated based on the extracted measured reflection curves and several of the aforementioned mirror film reference reflection characteristics, and the interpolation thickness range is obtained based on the aforementioned average reflection errors. The difference step size is set according to the interpolation thickness range, and several step points are obtained. The reflection interpolation of several step points is calculated based on the mirror detection environment to obtain several interpolation curves. The fitting error between the interpolation curves and the measured reflection curve is calculated based on the interpolation curves. The corresponding lens thickness with the smallest fitting error is identified and adjusted based on the fitting errors.

5. The method for detecting defects in a car rearview mirror according to claim 2, characterized in that, A comprehensive mirror testing database was established based on historical mirror production information, including: Historical mirror reflection information and historical foreign object reflection information are extracted based on historical mirror production information. The historical mirror reflection information includes mirror film reflection information of several mirror detection environments, and the historical foreign object reflection information includes foreign object reflection information of several types of foreign objects. The mirror film reflection information is mapped to environmental detection conditions, foreign object type and foreign object reflection information respectively; A combined index relationship is created based on the environmental detection conditions and foreign object types, and the environmental detection conditions and foreign object types are classified and managed according to the combined index relationship. A comprehensive mirror detection database is constructed based on the mapping relationship and the combined index relationship.

6. The method for detecting defects in a car rearview mirror according to claim 1, characterized in that, Based on the aforementioned mirror detection environment, reflection response correction is performed on each sampling point to obtain several calibration reflection datasets, including: Based on the empty rearview mirror state, multi-band spectral baseline data and polarization baseline data at different polarization angles were collected; Spectral correction coefficients are calculated based on the spectral baseline data, and a polarization correction mapping matrix is ​​constructed based on the polarization baseline data; The specular reflection signal is spectrally and polarized based on the spectral correction coefficients and polarization correction mapping matrix to form a calibration reflection dataset.

7. The method for detecting defects in a car rearview mirror according to claim 6, characterized in that, The specular reflection signal is spectrally and polarized based on the spectral correction coefficients and polarization correction mapping matrix to form a calibrated reflection dataset, including: The specular reflection signal is weighted based on the spectral correction coefficients corresponding to the wavelength to obtain the spectral correction signal; For each sampling point and each sampling wavelength, the spectral correction signals with different polarization angles are combined to obtain a polarization signal vector; The polarization signal vector is linearly transformed with the polarization correction mapping matrix corresponding to the wavelength to obtain the polarization correction signal; Several polarization correction signals are obtained according to several mirror detection environments, and the several polarization correction signals are aggregated to obtain a calibration reflection dataset.

8. The method for detecting defects in a car rearview mirror according to claim 1, characterized in that, Anomaly determinations are made for the foreign object reflection feature map and the mirror film reflection deviation map, respectively. Based on the determination results, the rearview mirror defect detection results are obtained, including: Based on the foreign object reflection feature map and the mirror film reflection deviation map, the foreign object feature parameters and film thickness deviation values ​​are extracted, and compared with their respective preset thresholds to determine the foreign object contamination area and the film thickness abnormal area. Based on the foreign matter contamination area and the film thickness anomaly area, respectively obtain the contamination anomaly mask and the film thickness anomaly mask, and then perform spatial pixel superposition of the contamination anomaly mask and the film thickness anomaly mask; The abnormal defect type is identified based on the spatial pixel overlay result, and the rearview mirror defect detection result is obtained.

9. A defect detection system for automotive rearview mirrors, characterized in that, The system includes: The signal acquisition module collects the mirror reflection signals under multispectral polarization conditions at each sampling point on the mirror surface, and obtains the reflection response information of each sampling point based on the mirror reflection signals. The response correction module acquires the mirror detection environment, performs reflection response correction on each sampling point based on the mirror detection environment, and acquires several calibration reflection datasets. The reflection decomposition module establishes a specular reflection decomposition network, and performs reflection information stripping on several calibration reflection datasets based on the specular reflection decomposition network to obtain foreign object reflection feature maps and mirror film reflection deviation maps. The anomaly detection module performs anomaly detection on the foreign object reflection feature map and the mirror film reflection deviation map, and obtains the defect detection results of the car rearview mirror based on the detection results.

10. The defect detection system for automotive rearview mirrors according to claim 9, characterized in that, The response correction module includes: The data acquisition unit, based on the empty rearview mirror state, acquires multi-band spectral baseline data and polarization baseline data at different polarization angles; The mapping matrix unit calculates spectral correction coefficients based on spectral baseline data and constructs a polarization correction mapping matrix based on polarization baseline data; The data generation unit performs spectral and polarization correction on the specular reflection signal based on the spectral correction coefficient and polarization correction mapping matrix to form a calibrated reflection dataset.

Images