CN119000709A - Battery piece microdefect automatic identification system based on 3D machine vision - Google Patents

Battery piece microdefect automatic identification system based on 3D machine vision Download PDF

Info

Publication number
CN119000709A
CN119000709A CN202411100534.1A CN202411100534A CN119000709A CN 119000709 A CN119000709 A CN 119000709A CN 202411100534 A CN202411100534 A CN 202411100534A CN 119000709 A CN119000709 A CN 119000709A
Authority
CN
China
Prior art keywords
image
defects
defect
battery cell
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202411100534.1A
Other languages
Chinese (zh)
Inventor
王勇
汪金
韩耀
蒋梦菲
刘玲莉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu Xinyoupeng Technology Co ltd
Original Assignee
Jiangsu Huiming Information Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jiangsu Huiming Information Technology Co ltd filed Critical Jiangsu Huiming Information Technology Co ltd
Priority to CN202411100534.1A priority Critical patent/CN119000709A/en
Publication of CN119000709A publication Critical patent/CN119000709A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/01—Arrangements or apparatus for facilitating the optical investigation
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/01—Arrangements or apparatus for facilitating the optical investigation
    • G01N2021/0106—General arrangement of respective parts
    • G01N2021/0112—Apparatus in one mechanical, optical or electronic block
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8854—Grading and classifying of flaws
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8854—Grading and classifying of flaws
    • G01N2021/8858—Flaw counting
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8854—Grading and classifying of flaws
    • G01N2021/8861—Determining coordinates of flaws
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8883—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges involving the calculation of gauges, generating models
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P70/00—Climate change mitigation technologies in the production process for final industrial or consumer products
    • Y02P70/50—Manufacturing or production processes characterised by the final manufactured product

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Chemical & Material Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Signal Processing (AREA)
  • Image Analysis (AREA)

Abstract

本发明涉及电池片检测技术领域，公开了基于3D机器视觉的电池片微缺陷自动识别系统，系统包括3D视觉传感模块、机械传动模块、3D图像处理模块、机器学习模块、用户交互模块和报告生成模块。通过高精度深度相机全面采集电池片焊接区域的3D图像，并采用深度学习算法构建缺陷识别模型，实现对电池片3D图像的精确缺陷识别，包括缺陷的数量、位置和种类。图形用户界面提供了实时图像显示、缺陷列表展示和参数设置功能，便于操作人员交互。系统最终生成详细的检测报告。

The present invention relates to the field of battery cell detection technology, and discloses a battery cell micro-defect automatic identification system based on 3D machine vision. The system includes a 3D visual sensing module, a mechanical transmission module, a 3D image processing module, a machine learning module, a user interaction module, and a report generation module. The 3D image of the battery cell welding area is fully collected by a high-precision depth camera, and a defect recognition model is constructed using a deep learning algorithm to achieve accurate defect recognition of the battery cell 3D image, including the number, location, and type of defects. The graphical user interface provides real-time image display, defect list display, and parameter setting functions to facilitate operator interaction. The system finally generates a detailed inspection report.

Description

Battery piece microdefect automatic identification system based on 3D machine vision
Technical Field
The invention relates to the technical field of battery piece detection, in particular to a battery piece microdefect automatic identification system based on 3D machine vision.
Background
In the production process of solar cells, a welding process is a key link for ensuring effective connection of the cells with components such as interconnection bars, bus bars and the like. Due to various complications in the welding process, battery plate welding defects occur, which not only affect the electrical performance of the battery plate, but also may cause safety problems, resulting in reduced overall performance and shortened life of the battery assembly.
The battery piece welding defects mainly comprise welding cracks, incomplete welding, slag inclusion, air holes, explosion points, pinholes, pits, salient points, polarized light, black spots and the like, and the forming reasons of the defects are various and mainly comprise the following aspects:
Welding parameters are inappropriate: parameters such as welding current, voltage, welding speed, arc length and the like are unreasonably set, so that welding energy is insufficient or overlarge, and defects such as incomplete welding, air holes or cracks and the like are generated.
Welding material problems: the welding rod or welding wire has poor quality, has the problems of rust, greasy dirt or dampness, and the like, and is easy to generate gas and impurities during welding to form air holes and slag inclusion.
The welding environment is not good: if the wind speed is too high, the humidity is too high, the cleanliness is insufficient, and the like, the welding quality can be affected, and the generation probability of air holes and cracks can be increased.
Welding equipment failure: welding equipment ages, improper maintenance or mishandling may also lead to welding defects.
Currently, detection of battery tab welding defects is largely dependent on conventional manual visual inspection and automated inspection achieved using image processing techniques. The manual visual inspection is simple and convenient to operate, is greatly influenced by experience of operators, has strong subjectivity, and is difficult to ensure the consistency and accuracy of detection. With the expansion of production scale and the improvement of automation degree, the manual visual inspection has been difficult to meet the requirements of efficient and accurate detection. Although the image processing technology has certain application in weld joint detection, the traditional 2D vision scheme cannot obtain the space coordinate information of an object, and has poor detection effect on defects (such as pits, bumps and the like) related to high degree. Therefore, we propose a battery piece micro-defect automatic identification system based on 3D machine vision.
Disclosure of Invention
The invention aims to provide a battery piece microdefect automatic identification system based on 3D machine vision so as to solve the problems in the background art.
In order to achieve the above purpose, the present invention provides the following technical solutions: battery piece microdefect automatic identification system based on 3D machine vision includes:
the 3D vision sensing module is used for acquiring detailed surface morphology information of the battery piece; the 3D vision sensor module comprises an image shooting unit and a light source control unit, wherein the image shooting unit adopts a high-precision depth camera to comprehensively acquire 3D images of a battery piece welding area; the light source control unit is used for adjusting illumination conditions and reducing the influence of reflection and shadow on image quality;
the mechanical transmission module is used for automatically and stably transmitting the battery piece to the lower part of the image shooting unit for detection;
The 3D image processing module is used for preprocessing the acquired 3D image and optimizing image data; the 3D image processing module comprises an image denoising unit, a registration alignment unit and a geometric correction unit; the image denoising unit adopts a 3D image filtering technology to denoise and smooth the image; the image registration alignment unit is used for aligning the same object in different images to the same coordinate system in the multi-view images; the geometric correction unit is used for correcting geometric distortion in the image, including perspective deformation and distortion of the image;
The machine learning module is used for carrying out defect recognition on the battery piece image output by the 3D image processing module, and the machine learning module is used for constructing a defect recognition model based on a deep learning algorithm, wherein the defect recognition model is used for judging whether the battery piece in the image has defects and the number, the position and the type of the defects;
the user interaction module is used for providing a graphical user interface for the interaction between an operator and the system; the interface design comprises a real-time image display window used for displaying the battery piece image acquired by the 3D vision sensor module; a defect list display area for listing defect information identified by the machine learning module, including the number, location and type of defects; the parameter setting panel is used for an operator to adjust system detection parameters, including illumination conditions, image preprocessing algorithm parameters and defect recognition thresholds;
and the report generation module is used for sorting the data and the results collected in the system detection process into a report form and outputting the report form.
Preferably, the mechanical transmission module comprises a battery piece fixing device, a transmission guide rail and a driving device; the battery piece fixing device is used for firmly clamping the battery piece; the transmission guide rail is used for realizing stable conveying of the battery piece; the driving device is used for providing power and controlling the battery piece fixing device to move along the transmission guide rail;
The driving device starts and drives the transmission guide rail to move, the battery piece is placed on the fixing device and is firmly clamped, and along with the movement of the transmission guide rail, the fixing device carries the battery piece to move along a preset path to the position below the image shooting unit.
Preferably, the image denoising unit removes noise points in the 3D image by using a gaussian filtering method, and obtains a denoised image by convolution of a three-dimensional gaussian kernel and calculation of a three-dimensional gaussian function, wherein a calculation formula of the three-dimensional gaussian function is as follows:
wherein x, y and z represent coordinates of each voxel of the three-dimensional image in a three-dimensional space, G (x, y and z) represents coordinates obtained by calculating the three-dimensional Gaussian function of the voxels, sigma is a standard deviation of the three-dimensional Gaussian function, e is a base of natural logarithm, and pi is a circumference ratio.
Preferably, the method for aligning the same object in the multi-view image to the same coordinate system by the registration alignment unit is as follows:
i) Collecting 3D image data of multiple views including a battery cell welding region;
ii) selecting an image of one view as a reference image and transforming the images of the other views to align with the reference image;
iii) Extracting key feature points from each image by using a feature extraction algorithm, and establishing a corresponding relation between the feature points;
iv) calculating a coordinate transformation matrix for transforming the feature points in the non-reference image into the coordinate system of the reference image according to the correspondence between the feature points;
v) performing coordinate transformation on the non-reference image by using a coordinate transformation matrix, and aligning the same object in the multi-view image to the same coordinate system;
The coordinate transformation matrix is calculated by the following formula:
Wherein T is a coordinate transformation matrix, R is a rotation matrix, representing a rotation relationship between images, and T is a translation vector, representing a translation relationship between images.
Preferably, the geometric correction unit corrects the image by using a perspective transformation matrix to eliminate perspective deformation and distortion; the perspective transformation matrix is expressed as:
Wherein H is a perspective transformation matrix, H ij is an element in the matrix, and the direct linear transformation DLT algorithm is used for calculation.
Preferably, the types of welding defects include welding cracks, incomplete welding, slag inclusion, air holes, explosion points, pinholes, pits, bumps, polarized light and black spots.
Preferably, the step of constructing and training the defect recognition model by the machine learning module includes:
step 1: constructing a three-dimensional convolutional neural network model, wherein the model comprises an input layer, a plurality of hidden layers and a plurality of output layers; the input layer is used for receiving the preprocessed 3D battery piece image data; the hidden layer is used for extracting image features; the plurality of output layers are respectively used for outputting the number, the position and the type of the defects;
Step 2: preparing a training dataset comprising a plurality of 3D battery slice images labeled with the number, location, and type of defects;
Step 3: training the three-dimensional convolutional neural network model by using a training data set, and adjusting model parameters through forward propagation and backward propagation algorithms to minimize the difference between the number, position and type of defects predicted by the output layer and the real marks;
Step 4: in the training process, a proper loss function is adopted to evaluate the accuracy of model prediction, and model parameters are updated according to the gradient of the loss function;
Step 5: repeating the steps 3 and 4 until the performance of the model on the verification set reaches a predetermined standard or training reaches a predetermined iteration number;
step 6: and storing the trained model for the subsequent battery piece defect identification.
Preferably, the method for evaluating the performance of the defect recognition model comprises the following steps: preparing a test dataset containing images of the battery cells of known defect number, location and type; performing defect recognition on the images in the test data set by using the trained defect recognition model to obtain the number, position and type of defects predicted by the model; comparing the number, position and type of the defects predicted by the model with the known number, position and type of the defects in the test data set, and calculating the accuracy rate, recall rate and F1 score of model identification; and evaluating the performance of the model according to the accuracy, recall and F1 score calculation results.
Preferably, the report generating module designs a report template according to a preset format and structure, and the content comprising the template comprises a battery piece image, a defect identification result, detection parameters and detection time.
Compared with the prior art, the invention has the beneficial effects that:
The system adopts the 3D visual sensing module, can comprehensively and accurately acquire the detailed surface morphology information of the battery piece, and compared with the traditional 2D visual scheme, the detection capability of the battery piece on highly relevant defects (such as pits, bumps and the like) is remarkably improved. The automatic detection flow replaces manual visual inspection, so that the detection efficiency is greatly improved, and the detection error caused by human factors is reduced.
The machine learning module builds a strong defect recognition model based on a deep learning algorithm, can accurately judge whether defects exist in the battery piece, and the number, the position and the type of the defects, so that the comprehensiveness and the accuracy of defect recognition are improved.
Drawings
FIG. 1 is a diagram of the overall structure of the present invention;
FIG. 2 is a diagram of the working steps of the registration alignment unit;
FIG. 3 is a diagram of the working steps of the geometry correction unit;
FIG. 4 is a diagram of training steps for a defect recognition model.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1-4, the present invention provides a technical solution: battery piece microdefect automatic identification system based on 3D machine vision includes:
3D vision sensing module: the 3D vision sensing module adopted by the system comprises an image shooting unit and a light source control unit. The image shooting unit uses a high-precision depth camera, and the camera can acquire a comprehensive 3D image of a battery piece welding area and capture detailed surface morphology information. The light source control unit is responsible for adjusting illumination conditions so as to reduce the influence of reflection and shadow on image quality and ensure that the acquired image is clear and accurate.
And the mechanical transmission module is as follows: the mechanical transmission module is responsible for automatically and stably transmitting the battery piece to the lower part of the image shooting unit for detection. The module adopts a precise mechanical design and a transmission device, ensures the accurate and stable position of the battery piece in the transmission process, and avoids the blurring or distortion of images caused by vibration or displacement.
3D image processing module: the 3D image processing module preprocesses the acquired 3D image to optimize image data. The module comprises an image denoising unit, a registration alignment unit and a geometric correction unit. The image denoising unit adopts a 3D image filtering technology to denoise and smoothen the image, so that the definition and contrast of the image are improved. The registration alignment unit is used for aligning the same object in different images to the same coordinate system in the multi-view images, so as to ensure the accuracy of subsequent processing. The geometric correction unit is used for correcting geometric distortion in the image, including perspective deformation and distortion of the image, so that the image can more truly and accurately reflect the actual shape of the battery piece.
A machine learning module: the machine learning module is responsible for carrying out defect recognition on the battery piece image output by the 3D image processing module. The module builds a defect identification model based on a deep learning algorithm, and the model can judge whether the battery piece in the image has defects and the number, the position and the type of the defects. Through a large amount of data training and algorithm optimization, the model can realize efficient and accurate defect identification.
And a user interaction module: the user interaction module provides a graphical user interface for an operator to interact with the system. The interface design comprises a real-time image display window used for displaying the battery piece image acquired by the 3D vision sensor module; a defect list display area for listing defect information identified by the machine learning module, including the number, location and type of defects; and a parameter setting panel, which allows an operator to adjust system detection parameters, such as illumination conditions, image preprocessing algorithm parameters, defect recognition thresholds and the like, according to actual requirements.
A report generation module: the report generation module is responsible for arranging data and results collected in the system detection process into a report form and outputting the report form. The module can automatically generate a detailed detection report, including detection time, battery piece information, defect details and the like, and is convenient for an operator to analyze and archive detection results.
According to the invention, the 3D image of the welding area of the battery piece is acquired through the high-precision depth camera, and the automatic identification of the defects is realized through the steps of image processing, machine learning and the like, so that the method has the advantages of high efficiency, accuracy, high degree of automation and the like, and can remarkably improve the production quality and efficiency of the battery piece.
The invention is further illustrated in the following in connection with examples 1 to 3:
example 1:
the mechanical transmission module mainly comprises a battery piece fixing device, a transmission guide rail and a driving device.
The battery piece fixing device is used for firmly clamping the battery piece and ensuring that the battery piece cannot displace or shake in the conveying process. For example, the device can adopt a spring clamping mechanism, and the battery pieces with different thicknesses and sizes can be adapted by adjusting the pretightening force of the spring, so that the moderate clamping force is ensured, and the battery pieces cannot be damaged. Meanwhile, soft materials such as rubber pads and the like can be arranged on the clamping mechanism, so that the clamping stability is improved, and the surface of the battery piece is protected from being scratched.
The transmission guide rail is used for realizing stable conveying of the battery piece, the guide rail can be a high-precision linear guide rail, and the surface of the guide rail is subjected to precise grinding and polishing treatment, so that the stability and the accuracy of the battery piece in the conveying process are ensured. The guide rail can be provided with components such as a sliding block, a guide wheel and the like so as to reduce friction resistance and improve conveying stability. In addition, the design of the guide rail also considers the requirements of easy maintenance and replacement, for example, the guide rail is conveniently detached, cleaned and replaced by adopting a modularized design.
The driving device is a power source of the mechanical transmission module and is used for providing power and controlling the battery piece fixing device to move along the transmission guide rail. The driving device can adopt a mode of combining a servo motor and a speed reducer, and the accurate positioning and speed control of the battery piece fixing device can be realized by accurately controlling the rotating speed and the steering of the motor. For example, the servo motor can adopt a closed-loop control system, and the encoder feeds back the position information to realize accurate position control of the battery piece fixing device. The speed reducer can adopt a planetary speed reducer or a worm gear speed reducer and other high-precision speed reducers so as to ensure accurate transmission ratio and high transmission efficiency.
In the actual working process, the driving device is started and drives the transmission guide rail to move. The battery piece is placed on the fixing device by an operator, and the battery piece is firmly clamped by adjusting the pretightening force of the clamping mechanism. Along with the movement of the transmission guide rail, the fixing device carries the battery piece to move along a preset path to the lower part of the image shooting unit. For example, after the transmission speed and the path are set in the operation interface, the driving device is started, and the servo motor starts to rotate and drives the transmission guide rail to move after being decelerated by the speed reducer. The battery piece fixing device stably moves along the guide rail, and the battery piece is conveyed to the lower part of the image shooting unit for image acquisition and defect identification. In the process, the mechanical transmission module can ensure the stability and the accuracy of the battery piece, and provides a reliable basis for subsequent image acquisition and defect identification.
Example 2:
In the 3D image processing module, the image denoising unit is responsible for removing noise points in the 3D image so as to improve the quality and definition of the image. The gaussian filtering method is a widely used image denoising technology, and the denoised image is obtained through convolution of a three-dimensional gaussian kernel and calculation of a three-dimensional gaussian function. The three-dimensional Gaussian function is the core of the method, and the calculation formula is as follows:
wherein x, y and z represent coordinates of each voxel of the three-dimensional image in a three-dimensional space, G (x, y and z) represents coordinates obtained by calculating the three-dimensional Gaussian function of the voxels, sigma is a standard deviation of the three-dimensional Gaussian function, e is a base of natural logarithm, and pi is a circumference ratio. The specific denoising steps are as follows:
Step 1, determining standard deviation sigma: a suitable standard deviation sigma value is selected depending on the noise level of the image and the degree of denoising desired. A larger σ value results in a stronger denoising effect, but may also cause the image to become too blurred. If the noise level in the image is low, a relatively small sigma value may be selected to slightly remove the noise while preserving the details of the image as much as possible. For example, the sigma value may be between 0.5 and 1.0. For images with moderate noise levels, a moderate σ value may be chosen to balance the denoising effect and image detail preservation. For example, the sigma value may be between 1.0 and 2.0.
Step 2, constructing a three-dimensional Gaussian kernel: based on the selected sigma value, a three-dimensional gaussian kernel is constructed. The kernel is a three-dimensional matrix in which each element is calculated by a three-dimensional gaussian function.
Step 3, convolution operation: and carrying out convolution operation on the constructed three-dimensional Gaussian kernel and the original 3D image. This step corresponds to sliding a gaussian kernel across each voxel of the image and weighting the surrounding voxels according to the weights in the kernel, resulting in a denoised voxel value.
Step 4, acquiring a denoised image: the convolution operation is repeated until all voxels in the image have been processed. Finally, a denoised 3D image is obtained, wherein noise points are effectively removed, and important features of the image are preserved.
The registration alignment unit is a key part in multi-view 3D image processing, which is responsible for aligning the same object in images of different view angles under the same coordinate system. The following is a specific implementation procedure of the registration alignment unit:
and step 1, collecting multi-view 3D image data containing the welding area of the battery piece. These data may be acquired by a 3D scanner or other 3D imaging device to ensure that the images for each view contain the desired weld area.
And 2, selecting an image at one view angle as a reference image, and transforming the images at other view angles so as to enable the images to be visually aligned with the reference image.
And 3, extracting key feature points from each image by using a feature extraction algorithm, wherein the key feature points are corner points and edge points in the image. After extracting the feature points, establishing a corresponding relation between the feature points, namely determining which feature points in different images represent the same physical position.
And 4, calculating a coordinate transformation matrix according to the corresponding relation between the characteristic points, wherein the transformation matrix is used for transforming the characteristic points in the non-reference image into the coordinate system of the reference image. The calculation of the coordinate transformation matrix uses the following formula: Wherein T is a coordinate transformation matrix, R is a rotation matrix, representing a rotation relationship between images, and T is a translation vector, representing a translation relationship between images. By solving this matrix, the rotation and translation parameters required to align the non-reference image to the reference image can be obtained.
And 5, carrying out coordinate transformation on the non-reference image by using a coordinate transformation matrix, and updating the coordinates of each pixel in the non-reference image by using the coordinate transformation matrix calculated before so as to align the non-reference image with the reference image. After this step, the same object in the multi-view image will be aligned to the same coordinate system.
The geometric correction unit corrects the image by using a perspective transformation matrix. The following steps are specific:
Step 1, determining a perspective transformation matrix H, wherein the matrix is a 3x3 matrix, and elements in the matrix are obtained through Direct Linear Transformation (DLT) algorithm calculation. The DLT algorithm is an effective mathematical tool that can calculate a perspective transformation matrix from corresponding points in an image.
Before the perspective transformation matrix is calculated, some corresponding points in the image are selected, which are points with a definite position in both the original image and the desired corrected image. For example, four corner points of a rectangle are selected in the original image, and positions to which the four corner points should correspond are specified in the image for which correction is desired.
And 2, calculating a perspective transformation matrix H according to the corresponding points by using a DLT algorithm. The DLT algorithm considers the geometric relationships in the image and calculates a perspective transformation matrix that best maps the corresponding points. The perspective transformation matrix is expressed as:
Wherein H is a perspective transformation matrix, H ij is an element in the matrix, and the direct linear transformation DLT algorithm is used for calculation.
After the perspective transformation matrix H is obtained, it can be used to correct the image, which involves applying the perspective transformation matrix to each pixel in the image to eliminate perspective distortion and warping. Specifically, for each pixel in the image, its new coordinates in the corrected image are calculated from its coordinates and the perspective transformation matrix.
And 3, rearranging pixels in the image according to the calculated new coordinates to obtain a corrected image. This corrected image will eliminate perspective distortion and warping in the original image, making the objects in the image appear more natural and realistic.
Assuming that a 3D image of the battery cell welding area is photographed from obliquely above, the welding lines in the image appear curved due to perspective deformation. To correct the image, a rectangular region of the image is first selected, which contains the solder strips that need to be corrected. The position and shape that this rectangular area should correspond to, i.e. a standard rectangle, is specified in the desired corrected image. The perspective transformation matrix H is calculated from the corresponding points of the two rectangular areas using DLT algorithm. Finally, the original image is corrected by applying the perspective transformation matrix, so that a corrected image without perspective deformation and distortion is obtained, and welding strips in the corrected image are more visual and accurate.
Example 3:
The machine learning module builds and trains the defect recognition model comprising the following steps:
Step 1, constructing a three-dimensional convolutional neural network model: a three-dimensional convolutional neural network model is constructed, which consists of an input layer, a plurality of hidden layers and a plurality of output layers. The input layer is responsible for receiving the preprocessed 3D battery piece image data, and the image data is subjected to registration alignment and denoising processing so as to ensure the quality and consistency of the image. The hidden layer is used to extract features in the image. The plurality of output layers are respectively designed for outputting the number, the position and the type of the defects so as to realize comprehensive defect identification.
Step 2, preparing a training data set: for training a three-dimensional convolutional neural network model, a training dataset containing a large number of 3D battery slice images is prepared, and the images in the training dataset have been marked with the number, location and type of defects to ensure that the model can learn accurate defect recognition capability. The types of welding defects include welding cracks, lack of penetration, slag inclusions, air holes, explosions, pinholes, pits, bumps, polarized light, and black spots.
Step 3, training a three-dimensional convolutional neural network model: the three-dimensional convolutional neural network model is trained using the prepared training data set. In the training process, the output of the model is calculated by adopting a forward propagation algorithm and compared with the real mark. The parameters of the model are then adjusted by a back propagation algorithm to minimize the differences between the number, location and kind of defects predicted by the output layer and the real marks. The above is iterated until the performance of the model meets the production requirements.
Step 4, evaluating the accuracy of model prediction: during the training process, a loss function is used to evaluate the accuracy of model predictions. The loss function can quantify the difference between the model predictions and the true signature and provide a basis for adjusting the model parameters. Model parameters are updated according to the gradient of the loss function to further optimize the performance of the model.
Step 5, iterative training until reaching a preset standard: repeating steps 3 and 4, and continuously training and adjusting the model until the performance of the model on the verification set reaches a predetermined standard or the training reaches a predetermined iteration number.
Step 6, saving the trained model: after the performance of the model on the verification set reaches a preset standard or training reaches a preset iteration number, the trained model is stored.
The method for evaluating the performance of the defect recognition model comprises the following steps: a test dataset containing images of the battery cells of known defect number, location and type is prepared, the test dataset being independent of the training dataset, for objectively evaluating the recognition capabilities of the model. Each image in the data set is accurately marked, so that the accuracy of defect information is ensured. And performing defect recognition on the images in the test data set by using the trained defect recognition model. The model will output the number, location and type of defects predicted in each image.
The number, location and type of defects predicted by the model are compared to the number, location and type of defects known in the test dataset. By comparing the prediction result with the real label, the accuracy rate, recall rate and F1 fraction of model identification can be calculated. The accuracy is the ratio of the number of defects correctly identified by the model to the total number of defects identified, and reflects the accuracy of model identification. The recall rate is the ratio of the number of defects correctly identified by the model to the total number of real defects, and reflects the detection capability of the model to the defects. The F1 score is the harmonic mean of the accuracy and the recall rate, and the F1 score comprehensively considers the accuracy and the detectability of the model, so that the model is a more comprehensive performance index. According to the calculation results of the accuracy, the recall rate and the F1 score, the performance of the model can be objectively evaluated. If the accuracy, recall rate and F1 score of the model are all higher, the model has better performance in the task of identifying the defects of the battery piece, and the number, the position and the types of the defects can be accurately identified. Conversely, if these metrics are low, then further optimization of the model structure or training process is required to improve its recognition capabilities.
By the method, the machine learning module can effectively identify the welding defects in the battery piece so as to improve the quality and reliability of the product.
The report generation module designs a report template according to a preset format and structure, and the content comprising the template comprises a battery piece image, a defect identification result, detection parameters and detection time.
In order to intuitively display the detection object, the template comprises the insertion position of the battery piece image. And when the report is actually generated, the battery piece image acquired during detection is inserted into the corresponding position.
The defect recognition results detail all defects detected by the defect recognition model, including information on the number, location, type, and possibly severity of the defects.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (9)

1.基于3D机器视觉的电池片微缺陷自动识别系统，其特征在于，包括：1. A battery cell micro-defect automatic identification system based on 3D machine vision, characterized by comprising: 3D视觉传感模块，用于获取电池片的详细表面形貌信息；所述3D视觉传感器模块包括图像拍摄单元和光源控制单元，图像拍摄单元采用高精度深度相机，对电池片焊接区域进行全面3D图像采集；光源控制单元用于调节光照条件，减少反光和阴影对图像质量的影响；3D visual sensor module, used to obtain detailed surface morphology information of the battery cell; the 3D visual sensor module includes an image capture unit and a light source control unit. The image capture unit uses a high-precision depth camera to perform comprehensive 3D image acquisition of the battery cell welding area; the light source control unit is used to adjust the lighting conditions to reduce the impact of reflections and shadows on image quality; 机械传动模块，用于将电池片自动、稳定地传送到图像拍摄单元下方进行检测；A mechanical transmission module is used to automatically and stably transfer the battery cells to the bottom of the image capture unit for inspection; 3D图像处理模块，用于对采集到的3D图像进行预处理，优化图像数据；所述3D图像处理模块包括图像去噪单元、配准对齐单元和几何校正单元；所述图像去噪单元采用3D图像滤波技术，对图像进行去噪和平滑处理；所述图像配准对齐单元用于在多视角图像中，将不同图像中的同一对象对齐到同一坐标系下；所述几何校正单元用于纠正图像中的几何畸变，包括图像的透视变形和扭曲；A 3D image processing module is used to pre-process the collected 3D images and optimize the image data; the 3D image processing module includes an image denoising unit, a registration and alignment unit and a geometric correction unit; the image denoising unit uses 3D image filtering technology to denoise and smooth the image; the image registration and alignment unit is used to align the same object in different images to the same coordinate system in multi-view images; the geometric correction unit is used to correct geometric distortion in the image, including perspective deformation and distortion of the image; 机器学习模块，用于对3D图像处理模块输出的电池片图像进行缺陷识别，所述机器学习模块基于深度学习算法构建一个缺陷识别模型，所述缺陷识别模型用于判断图像中的电池片是否存在缺陷以及存在的缺陷数量、位置和种类；A machine learning module, used to identify defects in the cell images output by the 3D image processing module, wherein the machine learning module constructs a defect recognition model based on a deep learning algorithm, and the defect recognition model is used to determine whether there are defects in the cell in the image and the number, location and type of defects that exist; 用户交互模块，提供一个图形用户界面，用于操作人员与系统进行交互；界面设计包括实时图像显示窗口，用于展示3D视觉传感器模块采集的电池片图像；缺陷列表展示区，用于列出机器学习模块识别的缺陷信息，包括缺陷数量、位置和种类；参数设置面板，用于操作人员调整系统检测参数，包括光照条件、图像预处理算法参数和缺陷识别阈值；The user interaction module provides a graphical user interface for operators to interact with the system. The interface design includes a real-time image display window for displaying cell images collected by the 3D vision sensor module; a defect list display area for listing defect information identified by the machine learning module, including the number, location and type of defects; and a parameter setting panel for operators to adjust system detection parameters, including lighting conditions, image preprocessing algorithm parameters and defect recognition thresholds. 报告生成模块，用于将系统检测过程中收集的数据和结果整理成报告形式输出。The report generation module is used to organize the data and results collected during the system detection process into a report format for output. 2.根据权利要求1所述的基于3D机器视觉的电池片微缺陷自动识别系统，其特征在于：所述机械传动模块包括电池片固定装置、传动导轨和驱动装置；电池片固定装置用于稳固地夹持电池片；传动导轨用于实现电池片的平稳传送；驱动装置用于提供动力，控制电池片固定装置沿传动导轨移动；2. According to the 3D machine vision-based battery cell micro-defect automatic identification system of claim 1, it is characterized in that: the mechanical transmission module includes a battery cell fixing device, a transmission guide rail and a driving device; the battery cell fixing device is used to firmly clamp the battery cell; the transmission guide rail is used to achieve smooth transmission of the battery cell; the driving device is used to provide power to control the battery cell fixing device to move along the transmission guide rail; 驱动装置启动并带动传动导轨运动，电池片被放置在固定装置上并被稳固夹持，随着传动导轨的运动，固定装置携带电池片沿预定路径向图像拍摄单元下方移动。The driving device starts and drives the transmission guide rail to move, the battery sheet is placed on the fixing device and is firmly clamped, and as the transmission guide rail moves, the fixing device carries the battery sheet and moves along a predetermined path to the bottom of the image capture unit. 3.根据权利要求2所述的基于3D机器视觉的电池片微缺陷自动识别系统，其特征在于，所述图像去噪单元使用高斯滤波法去除3D图像中的噪点，通过三维高斯核的卷积和三维高斯函数的计算获取去噪后的图像，三维高斯函数的计算公式为：3. According to the 3D machine vision-based battery cell micro-defect automatic identification system of claim 2, it is characterized in that the image denoising unit uses Gaussian filtering to remove noise points in the 3D image, and obtains the denoised image by convolution of a three-dimensional Gaussian kernel and calculation of a three-dimensional Gaussian function, and the calculation formula of the three-dimensional Gaussian function is: 其中，x、y、z表示三维图像每个体素在三维空间中的坐标，G(x,y,z)表示体素经过三维高斯函数计算后获得的坐标，σ是三维高斯函数的标准差，e是自然对数的底数，π是圆周率。Among them, x, y, z represent the coordinates of each voxel of the three-dimensional image in the three-dimensional space, G(x, y, z) represents the coordinates of the voxel after the three-dimensional Gaussian function is calculated, σ is the standard deviation of the three-dimensional Gaussian function, e is the base of the natural logarithm, and π is the number of pi. 4.根据权利要求3所述的基于3D机器视觉的电池片微缺陷自动识别系统，其特征在于，所述配准对齐单元将多视角图像中的同一对象对齐到同一坐标系下的方法为：4. The automatic identification system of cell micro-defects based on 3D machine vision according to claim 3 is characterized in that the registration and alignment unit aligns the same object in the multi-view images to the same coordinate system by: i)收集包含电池片焊接区域的多视角的3D图像数据；i) Collecting 3D image data from multiple perspectives including the battery cell welding area; ii)选择一个视角的图像作为基准图像，并对其他视角的图像进行变换，以使其与基准图像对齐；ii) Select an image from one perspective as a reference image, and transform the images from other perspectives to align them with the reference image; iii)利用特征提取算法，从每个图像中提取关键特征点，并建立特征点之间的对应关系；iii) Using feature extraction algorithm, extract key feature points from each image and establish the correspondence between feature points; iv)根据特征点之间的对应关系，计算坐标变换矩阵，该坐标变换矩阵用于将非基准图像中的特征点变换到基准图像的坐标系下；iv) calculating a coordinate transformation matrix according to the correspondence between the feature points, where the coordinate transformation matrix is used to transform the feature points in the non-reference image into the coordinate system of the reference image; v)应用坐标变换矩阵对非基准图像进行坐标变换，将多视角图像中的同一对象对齐到同一坐标系下；v) applying a coordinate transformation matrix to transform the coordinates of the non-reference image, so as to align the same object in the multi-view images to the same coordinate system; 所述坐标变换矩阵的计算采用以下公式：The coordinate transformation matrix is calculated using the following formula: 其中，T是坐标变换矩阵，R是旋转矩阵，表示图像之间的旋转关系，t是平移向量，表示图像之间的平移关系。Among them, T is the coordinate transformation matrix, R is the rotation matrix, which represents the rotation relationship between images, and t is the translation vector, which represents the translation relationship between images. 5.根据权利要求4所述的基于3D机器视觉的电池片微缺陷自动识别系统，其特征在于，所述几何校正单元利用透视变换矩阵对图像进行校正，以消除透视变形和扭曲；所述透视变换矩阵表示为：5. The cell micro-defect automatic identification system based on 3D machine vision according to claim 4 is characterized in that the geometric correction unit corrects the image using a perspective transformation matrix to eliminate perspective deformation and distortion; the perspective transformation matrix is expressed as: 其中，H是透视变换矩阵，hij是矩阵中的元素，通过直接线性变换DLT算法计算获得。Where H is the perspective transformation matrix, and h ij is the element in the matrix, which is calculated by the direct linear transformation DLT algorithm. 6.根据权利要求1所述的基于3D机器视觉的电池片微缺陷自动识别系统，其特征在于：所述焊接缺陷的种类包括焊接裂纹、未焊透、夹渣、气孔、爆点、针孔、凹坑、凸点、偏光和黑点。6. According to the 3D machine vision-based automatic identification system for battery cell micro-defects as described in claim 1, it is characterized in that the types of welding defects include welding cracks, incomplete penetration, slag inclusions, pores, explosion points, pinholes, pits, convex points, polarization and black spots. 7.根据权利要求6所述的基于3D机器视觉的电池片微缺陷自动识别系统，其特征在于，所述机器学习模块构建和训练缺陷识别模型的步骤包括：7. The automatic cell micro-defect recognition system based on 3D machine vision according to claim 6, characterized in that the step of constructing and training the defect recognition model by the machine learning module comprises: 步骤1：构建三维卷积神经网络模型，该模型包含输入层、多个隐藏层和多个输出层；输入层用于接收预处理后的3D电池片图像数据；隐藏层用于提取图像特征；多个输出层分别用于输出缺陷的数量、位置和种类；Step 1: Construct a 3D convolutional neural network model, which includes an input layer, multiple hidden layers, and multiple output layers; the input layer is used to receive the preprocessed 3D cell image data; the hidden layer is used to extract image features; and the multiple output layers are used to output the number, location, and type of defects respectively; 步骤2：准备训练数据集，该数据集包含大量标记了缺陷数量、位置和种类的3D电池片图像；Step 2: Prepare a training dataset containing a large number of 3D cell images labeled with the number, location, and type of defects; 步骤3：使用训练数据集对三维卷积神经网络模型进行训练，通过前向传播和反向传播算法调整模型参数，以最小化输出层预测的缺陷数量、位置和种类与真实标记之间的差异；Step 3: Use the training data set to train the 3D convolutional neural network model, and adjust the model parameters through forward propagation and back propagation algorithms to minimize the difference between the number, location, and type of defects predicted by the output layer and the actual labels; 步骤4：在训练过程中，采用适当的损失函数来评估模型预测的准确性，并根据损失函数的梯度更新模型参数；Step 4: During the training process, an appropriate loss function is used to evaluate the accuracy of the model prediction and the model parameters are updated according to the gradient of the loss function. 步骤5：重复步骤3和4，直到模型在验证集上的性能达到预定标准或训练达到预定的迭代次数；Step 5: Repeat steps 3 and 4 until the performance of the model on the validation set reaches the predetermined standard or the training reaches the predetermined number of iterations; 步骤6：保存训练好的模型，以供后续的电池片缺陷识别使用。Step 6: Save the trained model for subsequent cell defect identification. 8.根据权利要求7所述的基于3D机器视觉的电池片微缺陷自动识别系统，其特征在于，评估缺陷识别模型性能的方法为：准备一个包含已知缺陷数量、位置和种类的电池片图像的测试数据集；使用训练好的缺陷识别模型对测试数据集中的图像进行缺陷识别，得到模型预测的缺陷数量、位置和种类；将模型预测的缺陷数量、位置和种类与测试数据集中已知的缺陷数量、位置和种类进行比较，计算模型识别的准确率、召回率和F1分数；根据准确率、召回率和F1分数的计算结果，评估模型的性能。8. According to the 3D machine vision-based automatic identification system for battery cell micro-defects according to claim 7, it is characterized in that the method for evaluating the performance of the defect recognition model is: preparing a test data set containing battery cell images with known defect quantity, location and type; using the trained defect recognition model to perform defect recognition on the images in the test data set to obtain the number, location and type of defects predicted by the model; comparing the number, location and type of defects predicted by the model with the number, location and type of defects known in the test data set, and calculating the accuracy, recall and F1 score of the model recognition; and evaluating the performance of the model based on the calculation results of the accuracy, recall and F1 score. 9.根据权利要求1所述的基于3D机器视觉的电池片微缺陷自动识别系统，其特征在于：所述报告生成模块按照预定的格式和结构设计报告模版，所述包括模版的内容包括电池片图像、缺陷识别结果、检测参数和检测时间。9. The battery cell micro-defect automatic identification system based on 3D machine vision according to claim 1 is characterized in that: the report generation module designs a report template according to a predetermined format and structure, and the content of the template includes the battery cell image, defect identification results, detection parameters and detection time.
CN202411100534.1A 2024-08-12 2024-08-12 Battery piece microdefect automatic identification system based on 3D machine vision Pending CN119000709A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202411100534.1A CN119000709A (en) 2024-08-12 2024-08-12 Battery piece microdefect automatic identification system based on 3D machine vision

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202411100534.1A CN119000709A (en) 2024-08-12 2024-08-12 Battery piece microdefect automatic identification system based on 3D machine vision

Publications (1)

Publication Number Publication Date
CN119000709A true CN119000709A (en) 2024-11-22

Family

ID=93476023

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202411100534.1A Pending CN119000709A (en) 2024-08-12 2024-08-12 Battery piece microdefect automatic identification system based on 3D machine vision

Country Status (1)

Country Link
CN (1) CN119000709A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119198754A (en) * 2024-11-27 2024-12-27 浙江毅力汽车空调有限公司 A visual inspection system for appearance defects of pressure sensors
CN119915841A (en) * 2025-04-01 2025-05-02 深圳市时纬自动化有限公司 A method and system for online detection of coating quality of automobile sunroof glass

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021207318A1 (en) * 2020-04-08 2021-10-14 BWXT Advanced Technologies LLC In-situ inspection method based on digital data model of weld
CN115272271A (en) * 2022-08-09 2022-11-01 郑州芯视道机器人技术有限公司 Pipeline defect detecting and positioning ranging system based on binocular stereo vision
CN116840242A (en) * 2023-08-29 2023-10-03 易启科技(吉林省)有限公司 Defect detection system and method for automobile stamping part
CN117095005A (en) * 2023-10-20 2023-11-21 山东龙拓新材料有限公司 Plastic master batch quality inspection method and system based on machine vision
CN117788400A (en) * 2023-12-22 2024-03-29 无锡尚德太阳能电力有限公司 A photovoltaic module defect detection method and system based on binocular vision
KR20240045396A (en) * 2022-09-29 2024-04-08 (주)나래컴즈 Deep learning CNF-based 3D detection and classification system for electrode coating defects of secondary batteries
CN118470005A (en) * 2024-07-10 2024-08-09 贯文信息技术(苏州)有限公司 Defect detection system based on visual recognition technology

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021207318A1 (en) * 2020-04-08 2021-10-14 BWXT Advanced Technologies LLC In-situ inspection method based on digital data model of weld
CN115272271A (en) * 2022-08-09 2022-11-01 郑州芯视道机器人技术有限公司 Pipeline defect detecting and positioning ranging system based on binocular stereo vision
KR20240045396A (en) * 2022-09-29 2024-04-08 (주)나래컴즈 Deep learning CNF-based 3D detection and classification system for electrode coating defects of secondary batteries
CN116840242A (en) * 2023-08-29 2023-10-03 易启科技(吉林省)有限公司 Defect detection system and method for automobile stamping part
CN117095005A (en) * 2023-10-20 2023-11-21 山东龙拓新材料有限公司 Plastic master batch quality inspection method and system based on machine vision
CN117788400A (en) * 2023-12-22 2024-03-29 无锡尚德太阳能电力有限公司 A photovoltaic module defect detection method and system based on binocular vision
CN118470005A (en) * 2024-07-10 2024-08-09 贯文信息技术(苏州)有限公司 Defect detection system based on visual recognition technology

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
中国机械行业卓越工程师教育联盟 编著: "《中国机械行业卓越工程师教育联盟第五届毕业设计大赛优秀作品案例集》", 30 November 2022, 江苏大学出版社, pages: 390 - 393 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119198754A (en) * 2024-11-27 2024-12-27 浙江毅力汽车空调有限公司 A visual inspection system for appearance defects of pressure sensors
CN119915841A (en) * 2025-04-01 2025-05-02 深圳市时纬自动化有限公司 A method and system for online detection of coating quality of automobile sunroof glass

Similar Documents

Publication Publication Date Title
CN119000709A (en) Battery piece microdefect automatic identification system based on 3D machine vision
WO2025129935A1 (en) Photovoltaic module defect detection method and system based on binocular vision
CN108355981B (en) Battery connector quality detection method based on machine vision
CN104063873B (en) A Online Detection Method for Surface Defects of Bushing Parts Based on Compressive Sensing
CN113111828B (en) A kind of bearing three-dimensional defect detection method and system
CN102680478A (en) Detection method and device of surface defect of mechanical part based on machine vision
CN115131268A (en) Automatic welding system based on image feature extraction and three-dimensional model matching
CN111982921A (en) Hole defect detection method and device, conveying platform and storage medium
CN116465335B (en) Automatic thickness measurement method and system based on point cloud matching
CN108872243B (en) A bearing roller surface defect detection method, system and device
CN120144799B (en) Railway track ballast cover plate abnormality detection method
CN111784645B (en) Filling pipeline crack detection method
CN117078674A (en) Intelligent PCBA appearance defect detection method and detection system
CN117825406A (en) PCB apparent micro defect detection system
CN119574565A (en) A lamination defect detection and grading device and method
CN119693841B (en) Endoscope pipeline inner wall crack automatic identification system based on video deep learning
CN119180820B (en) Detection method and terminal for printing defects of surfaces and trademarks of automobile glass
CN118287883B (en) A method, medium and robot for detecting weld seam on surface of atmospheric pressure tank
CN118002888B (en) Robust real-time weld joint tracking method based on time sequence information fusion
CN119963553A (en) Needle mark detection method and system based on deep learning and image processing
CN107123105A (en) Images match defect inspection method based on FAST algorithms
CN114998279B (en) A method for identifying and locating pits and cracks on the surface of a stone slab
CN116519700A (en) Three-dimensional defect detection system and method for bearing
CN116188373A (en) A Solder Joint Defect Detection Method Based on VGG Deep Convolutional Neural Network Model
CN119904679B (en) Glass bead defect detection method based on improvement Yolov11

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
TA01 Transfer of patent application right

Effective date of registration: 20250115

Address after: Room 103, Building 13, Jianghuai Science and Technology Park, No. 193 Huaihe East Road, Xindukou Street, Huaiyin District, Huai'an City, Jiangsu Province 223302

Applicant after: Jiangsu Xinyoupeng Technology Co.,Ltd.

Country or region after: China

Address before: 214000 No. 88, Renmin Road, Luoshe Town, Huishan District, Wuxi City, Jiangsu Province

Applicant before: Jiangsu Huiming Information Technology Co.,Ltd.

Country or region before: China

TA01 Transfer of patent application right