Mobile phone lens window glass missing detection method

Abstract

The invention relates to the technical field of optical detection and image processing, in particular to a mobile phone lens window glass defect detection method. Comprising the following steps: acquiring a high-resolution image of a window glass sample by imaging equipment by adopting a multispectral imaging technology; inputting the acquired image into a pre-trained deep learning model for preprocessing; determining whether defects exist or not according to an output result of the deep learning model, marking the positions and types of the defects, and generating a detection report containing defect geometric parameters; if the defect exists, the severity of the defect is judged according to preset standards and geometric parameters of the defect, and the standards comprise the size, the shape, the depth and the position of the defect; and the detection result and the image data are stored in a cloud database to support subsequent quality analysis and process improvement. By combining the multispectral imaging technology, the deep learning model and the fuzzy logic system, the quality detection level of the window glass is comprehensively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of optical detection and image processing, and in particular to a method for detecting defects in mobile phone lens window glass. Background Art

[0002] With the popularity of smart phones and the advancement of technology, the quality requirements of mobile phone lens window glass are getting higher and higher. The defects of window glass not only affect the appearance of the mobile phone, but more importantly, they will affect the imaging quality of the camera, and thus affect the user experience. Therefore, efficient window glass defect detection methods have become an indispensable part of the production process.

[0003] Traditional methods for detecting defects in mobile phone lens window glass mainly rely on manual visual inspection or single imaging technology. Manual visual inspection methods have problems such as slow detection speed, strong subjectivity, and easy fatigue, which makes it difficult to meet the requirements of large-scale production and high precision. Although single imaging technology (such as visible light imaging) can detect certain types of surface defects, it has limited detection effects on internal defects, thermal stress changes, and transparency changes.

[0004] The existing mobile phone lens window glass defect detection technology has the following shortcomings: (1) Insufficient detection accuracy: Single imaging technology cannot fully detect various defects in window glass, especially when detecting internal defects and thermal stress changes, the accuracy is low. (2) Low detection efficiency: Manual visual inspection is slow and easily affected by subjective factors and fatigue of operators, resulting in unstable detection results. (3) Poor adaptability: Existing detection equipment is difficult to maintain consistency in image quality under different production conditions and requires frequent manual adjustments, which increases the difficulty and cost of operation. Summary of the invention

[0005] The present invention provides an efficient, accurate and adaptive method for detecting defects in mobile phone lens window glass. By combining multispectral imaging technology, deep learning models and fuzzy logic systems, the quality inspection level of window glass is comprehensively improved. At the same time, automatic data management and analysis are achieved through a cloud database to support continuous optimization of the production process.

[0006] The technical solution adopted by the present invention is: a method for detecting defects in a mobile phone lens window glass, comprising the following steps:

[0007] Step 1: The imaging device uses multispectral imaging technology to obtain a high-resolution image of the window glass sample;

[0008] Step 2: Input the acquired image into the pre-trained deep learning model for pre-processing;

[0009] Step 3: Based on the output of the deep learning model, determine whether there is a defect, mark the location and type of the defect, and generate a test report containing the defect geometry parameters;

[0010] Step 4: If there is a defect, determine the severity of the defect based on the preset standards and the geometric parameters of the defect, including the size, shape, depth and location of the defect;

[0011] Step 5: Store the inspection results and image data in a cloud database to support subsequent quality analysis and process improvement.

[0012] As a further improvement of the present invention, the multispectral imaging technology includes visible light imaging, ultraviolet imaging and infrared light imaging. The images obtained by each imaging technology are stored separately, and then a comprehensive image is generated through an image fusion algorithm to improve the multi-spectral information content of the image and the accuracy of defect recognition.

[0013] As a further improvement of the present invention, the visible light imaging is used to detect surface defects; the ultraviolet light imaging is used to detect internal defects; and the infrared light imaging is used to detect thermal stress and transparency changes;

[0014] The resolution of the visible light imaging is not less than 50 million pixels, and the resolution of the ultraviolet imaging and the infrared light imaging is not less than 20 million pixels, so as to ensure that the images under different spectra have sufficient details to identify tiny defects.

[0015] As a further improvement of the present invention, the imaging device is provided with a feedback mechanism, which automatically adjusts the parameters of the imaging device, including exposure time, aperture size, and light source intensity, to ensure that the quality of images obtained under different production conditions is consistent;

[0016] The feedback mechanism includes an adaptive learning module, which adopts a reinforcement learning algorithm. Through continuous learning and optimization, it can automatically optimize the parameters of the imaging equipment according to the historical detection data in the cloud database to adapt to changes in different production environments and improve image quality and detection stability.

[0017] As a further improvement of the present invention, the deep learning model is a multi-task convolutional neural network (MT-CNN), which is trained with a large amount of annotated comprehensive image data. It can not only identify the defect type in the image, but also simultaneously estimate the geometric parameters and location information of the defects, including size, shape, and depth.

[0018] As a further improvement of the present invention, the MT-CNN model includes a trunk network and multiple branch networks, the trunk network is used to extract common features of the image, and the branch networks are used for defect classification, geometric parameter estimation and position positioning respectively;

[0019] The backbone network includes multiple convolutional layers, pooling layers and activation layers, the branch network includes multiple fully connected layers and output layers, and the feature sharing mechanism between the backbone network and the branch network is used to improve the generalization ability and detection efficiency of the model.

[0020] As a further improvement of the present invention, the inspection report includes an image, position coordinates, type description, size, shape and depth of the defect.

[0021] As a further improvement of the present invention, the defect severity assessment adopts a fuzzy logic system to comprehensively consider multiple parameters and generate a score to more accurately assess the impact of the defect.

[0022] As a further improvement of the present invention, the fuzzy logic system includes a fuzzification module, multiple fuzzy rules and a defuzzification module. The fuzzification module converts the input defect geometric parameters into fuzzy values, which represent a fuzzy set of defect parameters; the multiple fuzzy rules define the fuzzy relationship between the defect parameters and the defect severity based on expert knowledge; the defuzzification module converts the output of the fuzzy rules into a specific severity score, which reflects the degree of influence of the defect on the quality of the window glass.

[0023] As a further improvement of the present invention, the cloud database includes a data cleaning module for automatically removing invalid data and noise, improving the quality and availability of data, and supporting batch processing and analysis of data;

[0024] The cloud database supports encrypted transmission and storage of data to ensure data security and privacy protection, and provides data backup and recovery functions to prevent data loss.

[0025] Beneficial effects of the present invention: (1) Comprehensively improve detection accuracy: The present invention, by combining multi-spectral imaging technology, can realize comprehensive detection of various defects in mobile phone lens window glass, including surface defects, internal defects, thermal stress changes and transparency changes, which significantly improves the accuracy and reliability of detection.

[0026] (2) Significantly improve detection efficiency: The present invention utilizes deep learning models, especially multi-task convolutional neural networks (MT-CNN), which can automatically identify and classify defects and estimate the geometric parameters and location information of the defects, thereby greatly reducing manual intervention and improving detection efficiency.

[0027] (3) Enhanced adaptive capability: The imaging device of the present invention is equipped with a feedback mechanism and an adaptive learning module, which enables the imaging parameters to be automatically adjusted according to different production environments, thereby ensuring the consistency of image quality and improving the stability and adaptability of detection.

[0028] (4) Supporting data management and process optimization: The present invention realizes automatic management and analysis of data through a cloud database, which not only improves the security and availability of data, but also supports the continuous optimization of production processes and provides strong data support for production decisions. DETAILED DESCRIPTION

[0029] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with the embodiments. It should be understood that the embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0030] The present invention provides a method for detecting a defect in a mobile phone lens window glass, comprising the following steps:

[0031] Step 1: The imaging device uses multispectral imaging technology to obtain a high-resolution image of the window glass sample;

[0032] Multispectral imaging technology includes visible light imaging, ultraviolet imaging and infrared imaging. The images obtained by each imaging technology are stored separately, and then a comprehensive image is generated through an image fusion algorithm to improve the image's multi-dimensional information content and the accuracy of defect recognition. Visible light imaging is used to detect surface defects; ultraviolet imaging is used to detect internal defects; infrared imaging is used to detect thermal stress and transparency changes; the resolution of visible light imaging is not less than 50 million pixels, and the resolution of ultraviolet imaging and infrared imaging is not less than 20 million pixels to ensure that images under different spectra have sufficient details to identify tiny defects.

[0033] The imaging equipment is equipped with a feedback mechanism, which automatically adjusts the parameters of the imaging equipment, including exposure time, aperture size, and light source intensity, to ensure consistent image quality under different production conditions. The feedback mechanism includes an adaptive learning module, which uses a reinforcement learning algorithm. Through continuous learning and optimization, it can automatically optimize the parameters of the imaging equipment based on the historical detection data in the cloud database to adapt to changes in different production environments and improve image quality and detection stability.

[0034] Step 2: Input the acquired image into the pre-trained deep learning model for pre-processing;

[0035] The deep learning model is a multi-task convolutional neural network (MT-CNN), which is trained with a large amount of annotated comprehensive image data. It can not only identify the defect type in the image, but also simultaneously estimate the geometric parameters and location information of the defects, including size, shape, and depth.

[0036] The MT-CNN model consists of a backbone network and multiple branch networks. The backbone network is used to extract common features of the image, and the branch networks are used for defect classification, geometric parameter estimation, and position positioning, respectively. The backbone network includes multiple convolutional layers, pooling layers, and activation layers, and the branch networks include multiple fully connected layers and output layers. The feature sharing mechanism between the backbone network and the branch networks is used to improve the generalization ability and detection efficiency of the model.

[0037] Step 3: Based on the output of the deep learning model, determine whether there is a defect, mark the location and type of the defect, and generate a test report containing the defect geometry parameters;

[0038] The inspection report includes the defect's image, location coordinates, type description, size, shape and depth.

[0039] Step 4: If there is a defect, determine the severity of the defect based on the preset standards and the geometric parameters of the defect, including the size, shape, depth and location of the defect;

[0040] The defect severity assessment uses a fuzzy logic system, which comprehensively considers multiple parameters to generate a score to more accurately assess the impact of the defect. The fuzzy logic system includes a fuzzification module, multiple fuzzy rules, and a defuzzification module. The fuzzification module converts the input defect geometric parameters into fuzzy values, which represent the fuzzy set of defect parameters; multiple fuzzy rules define the fuzzy relationship between defect parameters and defect severity based on expert knowledge; the defuzzification module converts the output of the fuzzy rules into a specific severity score, which reflects the degree of impact of the defect on the quality of the window glass.

[0041] Step 5: Store the test results and image data in a cloud database to support subsequent quality analysis and process improvement.

[0042] The cloud database includes a data cleaning module, which is used to automatically remove invalid data and noise, improve data quality and availability, and support batch processing and analysis of data; the cloud database supports encrypted transmission and storage of data to ensure data security and privacy protection, and provides data backup and recovery functions to prevent data loss.

[0043] Example

[0044] Object: Mobile phone lens window glass samples;

[0045] Equipment: multispectral imaging system, deep learning computing platform, cloud data management system;

[0046] Software tools: deep learning framework (TensorFlow or PyTorch), image processing software, fuzzy logic system building tools, database management and analysis tools;

[0047] The implementation steps are as follows:

[0048] 1. High-resolution multispectral imaging

[0049] Use a multispectral imaging device equipped with three light sources: visible light, ultraviolet light, and infrared light to photograph the window glass of the mobile phone lens. According to the characteristics of different light sources: (1) Visible light imaging: Captures external defects such as tiny scratches and stains on the surface of the window glass. (2) Ultraviolet light imaging: Penetrates the window glass to detect possible internal defects such as cracks and bubbles. (3) Infrared light imaging: Evaluates the distribution of thermal stress and uneven transparency in the window glass.

[0050] The imaging equipment uses machine vision and automatic control technology to ensure that the parameters of each shot (such as exposure time, aperture size, and light source intensity) are optimal and adaptable to different production environments. These parameters are adjusted through the device's built-in adaptive learning module, which uses a reinforcement learning algorithm to continuously optimize parameter settings based on previous inspection experience.

[0051] (II) Image preprocessing

[0052] The acquired high-resolution images (including images under different spectra) are transmitted to the deep learning computing platform, and then preprocessed by denoising and enhancement to remove irrelevant information in the image and highlight the defective area. These preprocessing steps can be implemented based on existing image processing technology.

[0053] (III) Defect identification and classification

[0054] The preprocessed images are fed into a multi-task convolutional neural network (MT-CNN) that has been trained with a large amount of labeled data. The MT-CNN model automatically identifies defects in the image and classifies and measures the defect type, size, shape, depth, and location. The backbone network of the model is responsible for extracting common features from the image, while the branch networks focus on different tasks, such as defect classification, geometric parameter estimation, and location localization.

[0055] (IV) Defect severity assessment

[0056] Once the specific information of the defect is determined, the system inputs this information into the fuzzy logic system to assess the severity of the defect. The fuzzification module in the fuzzy logic system converts the geometric parameters of the defect into fuzzy values; a preliminary assessment of the severity of the defect is obtained based on the established fuzzy rules; finally, the defuzzification module converts the output of the fuzzy rules into an exact score to help operators quickly understand the impact of the defect on product quality.

[0057] (V) Data management and analysis

[0058] All test results, including image data and defect analysis reports, will be securely uploaded to the cloud database. This database not only supports data encryption transmission and storage to prevent data leakage, but also has a data cleaning function that can automatically process invalid data and noise to improve data quality. By analyzing historical test data stored in the cloud, continuous optimization of production processes can be achieved. For example, the types of defects in high-incidence areas can be identified and production parameters can be adjusted to reduce the probability of such defects.

[0059] Conclusion: Through the above implementation steps, the mobile phone lens window glass defect detection method provided by the present invention can realize fast and accurate defect detection on the production line, greatly improving product quality and production efficiency. At the same time, with the technical support of fuzzy logic system and cloud database, it is also possible to make a more accurate assessment of the impact of defects and provide empirical evidence for the improvement of production process.

[0060] In summary, the present invention provides an efficient, accurate and adaptive method for detecting defects in mobile phone lens window glass, which solves the problems of insufficient precision, low efficiency and poor adaptability in traditional detection methods. By combining multispectral imaging technology, deep learning models (specifically multi-task convolutional neural networks, MT-CNN) and fuzzy logic systems, the present invention realizes comprehensive detection of various defects in window glass (including surface defects, internal defects, thermal stress changes and transparency changes), and can automatically evaluate the severity of the defects. At the same time, automatic management and analysis of data are realized through cloud databases to support continuous optimization of production processes.

[0061] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting defects in mobile phone lens window glass, characterized in that: The following steps are involved: Step 1: The imaging device uses multispectral imaging technology to obtain a high-resolution image of the window glass sample; Step 2: Input the acquired image into the pre-trained deep learning model for pre-processing; Step 3: Based on the output of the deep learning model, determine whether there is a defect, mark the location and type of the defect, and generate a test report containing the defect geometry parameters; Step 4: If there is a defect, determine the severity of the defect based on the preset standards and the geometric parameters of the defect, including the size, shape, depth and location of the defect; Step 5: Store the inspection results and image data in a cloud database to support subsequent quality analysis and process improvement.

2. The method for detecting defects in mobile phone lens window glass according to claim 1, characterized in that: The multispectral imaging technology includes visible light imaging, ultraviolet imaging and infrared light imaging. The images obtained by each imaging technology are stored separately, and then a comprehensive image is generated through an image fusion algorithm to improve the multi-spectral information content of the image and the accuracy of defect recognition.

3. The method for detecting defects in mobile phone lens window glass according to claim 2, characterized in that: The visible light imaging is used to detect surface defects; the ultraviolet light imaging is used to detect internal defects; and the infrared light imaging is used to detect thermal stress and transparency changes; The resolution of the visible light imaging is not less than 50 million pixels, and the resolution of the ultraviolet imaging and the infrared light imaging is not less than 20 million pixels, so as to ensure that the images under different spectra have sufficient details to identify tiny defects.

4. The method for detecting defects in mobile phone lens window glass according to claim 1, characterized in that: The imaging device is provided with a feedback mechanism, which automatically adjusts the parameters of the imaging device, including exposure time, aperture size, and light source intensity, to ensure consistent image quality under different production conditions; The feedback mechanism includes an adaptive learning module, which adopts a reinforcement learning algorithm. Through continuous learning and optimization, it can automatically optimize the parameters of the imaging equipment according to the historical detection data in the cloud database to adapt to changes in different production environments and improve image quality and detection stability.

5. The method for detecting defects in mobile phone lens window glass according to claim 1, characterized in that: The deep learning model is a multi-task convolutional neural network (MT-CNN), which is trained with a large amount of annotated comprehensive image data. It can not only identify the defect type in the image, but also simultaneously estimate the geometric parameters and location information of the defects, including size, shape, and depth.

6. The method for detecting defects in mobile phone lens window glass according to claim 5, characterized in that: The MT-CNN model includes a backbone network and multiple branch networks, wherein the backbone network is used to extract common features of the image, and the branch networks are used for defect classification, geometric parameter estimation, and position positioning respectively; The backbone network includes multiple convolutional layers, pooling layers and activation layers, the branch network includes multiple fully connected layers and output layers, and the feature sharing mechanism between the backbone network and the branch network is used to improve the generalization ability and detection efficiency of the model.

7. The method for detecting defects in mobile phone lens window glass according to claim 1, characterized in that: The inspection report includes the image, location coordinates, type description, size, shape and depth of the defect.

8. The method for detecting defects in mobile phone lens window glass according to claim 1, characterized in that: The defect severity assessment uses a fuzzy logic system to comprehensively consider multiple parameters and generate a score to more accurately assess the impact of the defect.

9. The method for detecting defects in mobile phone lens window glass according to claim 8, characterized in that: The fuzzy logic system includes a fuzzification module, multiple fuzzy rules and a defuzzification module. The fuzzification module converts the input defect geometric parameters into fuzzy values, which represent the fuzzy set of defect parameters; the multiple fuzzy rules define the fuzzy relationship between the defect parameters and the defect severity based on expert knowledge; the defuzzification module converts the output of the fuzzy rules into a specific severity score, which reflects the degree of influence of the defect on the quality of the window glass.

10. The method for detecting defects in mobile phone lens window glass according to claim 1, characterized in that: The cloud database includes a data cleaning module for automatically removing invalid data and noise, improving the quality and availability of data, and supporting batch processing and analysis of data; The cloud database supports encrypted transmission and storage of data to ensure data security and privacy protection, and provides data backup and recovery functions to prevent data loss.