WO2026020907A1 - 一种多级多模态融合的oled缺陷检测装置 - Google Patents

一种多级多模态融合的oled缺陷检测装置

Info

Publication number
WO2026020907A1
WO2026020907A1 PCT/CN2025/090784 CN2025090784W WO2026020907A1 WO 2026020907 A1 WO2026020907 A1 WO 2026020907A1 CN 2025090784 W CN2025090784 W CN 2025090784W WO 2026020907 A1 WO2026020907 A1 WO 2026020907A1
Authority
WO
WIPO (PCT)
Prior art keywords
multispectral
light source
oled
polarization
detection
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
PCT/CN2025/090784
Other languages
English (en)
French (fr)
Inventor
林斌
郭睿妮
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.)
Zhejiang University
Original Assignee
Zhejiang University
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 Zhejiang University filed Critical Zhejiang University
Publication of WO2026020907A1 publication Critical patent/WO2026020907A1/zh
Pending legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G01N21/89—Investigating the presence of flaws or contamination in moving material, e.g. running paper or textiles
    • G01N21/892—Investigating the presence of flaws or contamination in moving material, e.g. running paper or textiles characterised by the flaw, defect or object feature 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
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/21—Polarisation-affecting properties
    • 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
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • 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
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/255—Details, e.g. use of specially adapted sources, lighting or optical systems
    • 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
    • G01N21/8806—Specially adapted optical and illumination features
    • 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
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1734—Sequential different kinds of measurements; Combining two or more methods
    • 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
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N2021/1734—Sequential different kinds of measurements; Combining two or more methods
    • G01N2021/1736—Sequential different kinds of measurements; Combining two or more methods with two or more light sources
    • 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
    • G01N21/8806—Specially adapted optical and illumination features
    • G01N2021/8845—Multiple wavelengths of illumination or detection
    • 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
    • G01N21/8806—Specially adapted optical and illumination features
    • G01N2021/8848—Polarisation of light

Definitions

  • This invention relates to the field of workpiece manufacturing and processing, specifically to a multi-level, multi-modal fusion OLED defect detection device.
  • Specular reflection interferes with measurement results: Because the surface of the OLED screen is smooth and it is a specular reflective object, the high-resolution image taken vertically often contains the camera reflection, which in turn causes great interference to the detection results. If the camera angle is adjusted to be tilted for shooting, some full field of view resolution will usually be sacrificed, resulting in lower measurement accuracy.
  • OLED screens have a wide variety of defects with complex and varied shapes and characteristics, and a single detection device often cannot detect all types of defects.
  • This invention addresses the shortcomings of existing technologies by designing a multi-level, multi-modal fusion device based on polarization and multispectral analysis for six types of defects in OLED screens: cracks, scratches, foreign objects, fragments, bubbles, and dirt.
  • this paper innovatively introduces polarization and multispectral information into the OLED image acquisition stage to enhance defect features and thereby improve the system's ability to detect minute defects.
  • a dual-point detection system was adopted to perform phased detection of different types of defects, based on the causes and material properties of the defects.
  • a polarization-fusion-based detection point one is used to detect cracks, scratches, foreign objects, and fragments.
  • a multispectral fusion-based device two is used to detect bubbles and dirt defects on the OLED screen.
  • This invention discloses a multi-level, multi-modal fusion OLED defect detection device, the device comprising:
  • Sample transfer control device includes a conveyor belt for automatically feeding, placing and transferring OLED samples
  • Dual-point image acquisition device a polarization camera and a multispectral camera, used for detection based on polarization fusion and multispectral fusion, respectively;
  • Illumination devices coaxial light source and multispectral light source.
  • the coaxial light source is used to emit coaxial light onto the OLED sample and generate reflected light to the field of view of the polarization camera;
  • the multispectral light source is used to emit multispectral ring light onto the OLED sample and generate reflected light to the field of view of the multispectral camera.
  • Light source controller Used to control the switching and brightness of coaxial light source and multispectral light source, as well as to ensure that the emission frequency of multispectral ring light emitted by multispectral light source is consistent with the acquisition frequency of multispectral camera;
  • Industrial control computer used to synchronously control the acquisition speed of the image acquisition device, generate polarization fusion image and multispectral fusion image respectively, execute defect detection algorithm, and perform post-fusion of the two-stage detection results to determine whether the sample is qualified;
  • Polarizing film Attached to the light-emitting surface of a coaxial light source to obtain polarized light
  • the light source controller, conveyor belt, polarization camera, and multispectral camera are all connected to the industrial control computer.
  • the coaxial light source and multispectral light source are both connected to the light source controller.
  • the polarization camera, coaxial light source, and polarization film constitute the first detection point through which the OLED sample passes, and the multispectral light source and multispectral camera constitute the second detection point through which the OLED sample passes.
  • the detection device of the present invention also includes a sample stage, which is placed on a conveyor belt for placing the OLED sample to be detected.
  • the conveyor belt described in this invention is used to stably transport samples through two detection points; the coaxial light source is used to provide uniform illumination; and the multispectral ring light is used to provide illumination over a wide range of spectra.
  • the detection point one described in this invention is used to detect cracks, scratches, foreign objects and fragments on the surface of the OLED screen; the detection point is used to detect defects such as bubbles and dirt on the OLED screen.
  • the light-emitting area of the coaxial light source described in this invention is greater than or equal to 1.5 times the sample size, ensuring that the light emitted by the coaxial light source can uniformly illuminate the sample surface.
  • the two detection points of the present invention also include a fixing device for fixing the camera and light source at the two points to the main body device.
  • the fixing device is connected to the polarization camera and coaxial light source at point one, and the fixing device is connected to the multispectral camera and multispectral ring light at point two, so as to ensure the stability and accuracy of each point during the detection process.
  • This invention effectively removes the interference of camera reflection in the acquired image by using a large-size coaxial light source, solves the problem of reduced measurement accuracy in the usual tilt shooting anti-reflection strategy, and realizes the acquisition of high-quality images under a large vertical field of view;
  • This invention solves the problem of excessive noise in polarization and polarization images acquired under conventional light source acquisition scenarios with low polarization by adding a linear polarization film to the coaxial light source, thus achieving the acquisition of high-quality polarization images;
  • This invention introduces polarization mode information into the image by designing a single-segment polarization fusion device and a two-segment multispectral fusion device, which introduces spectral information into the image.
  • the introduction of these two modes of information enhances the features of minute defects in the acquired image, thereby improving the detection rate of minute defects.
  • This device employs a multi-level, multi-modal fusion architecture, dividing the detection process into a first-stage detection based on a polarization fusion device and a second-stage detection based on a multispectral fusion device.
  • An industrial control computer then performs post-fusion of the two-stage detection results, enabling a more comprehensive and accurate assessment of OLED screen defects.
  • This multi-level detection result fusion method effectively improves the recall rate of the defect detection system, reduces the false detection rate, and provides a more efficient, comprehensive, and reliable OLED screen defect detection technology solution.
  • Figure 1 is a schematic diagram of a multi-level, multi-modal fusion detection device
  • 1 is a polarization camera
  • 2 is a coaxial light source
  • 3 is a polarization film
  • 4 is a sample stage
  • 5 is a conveyor belt
  • 7 is a multispectral light source
  • 8 is a multispectral camera
  • 9 is a light source controller
  • 10 is an industrial control computer.
  • Figure 1 shows the multi-level multimodal fusion detection device designed for this device, which mainly includes a sample transmission control device, a dual-point image acquisition device, an illumination device, a light source controller 9, an industrial control computer 10, and a polarizing film 3, etc.
  • Sample transfer control device including conveyor belt 5 and control motor, using Zhenyu Automation multi-station guide rail assembly, used for automatic feeding, placement and transfer of OLED samples; conveyor belt 5 is used to stably transport the sample through the two detection points for the next test.
  • Dual-point acquisition device including polarization camera 1 and multispectral camera 8.
  • Polarization camera 1 uses a Hikvision MV-CH050-10UP industrial camera and MVL-MF2528M-8MP industrial camera lens group to acquire four-way polarization images;
  • multispectral camera 8 uses a Keyence CA-H500MX industrial camera and CA-LH25 industrial camera lens group to acquire multispectral images.
  • Illumination device includes Dongguan COAX-200X coaxial light source 2 and Keyence CA-DRM10X multispectral light source 7.
  • Coaxial light source 2 is used to provide uniform illumination, emitting coaxial light onto the OLED sample and generating reflected light to the field of view of polarization camera 1.
  • the emitting area of coaxial light source 2 is greater than or equal to 1.5 times the sample size, ensuring that the light emitted by coaxial light source 2 can uniformly illuminate the sample surface.
  • the multispectral light source 7 provides illumination with a wide range of spectra, emitting multispectral ring light onto the OLED sample and generating reflected light into the field of view of the multispectral camera 8.
  • the Keyence CA-DRM10X light source controller 9 is used to control the switching and brightness of the coaxial light source 2 and the multispectral light source 7, as well as to control the emission frequency of the multispectral ring light emitted by the multispectral light source 7 to be consistent with the acquisition frequency of the multispectral camera 8.
  • Industrial PC 10 Using a Dell Inspiron 3020 office unit, it is used to synchronously control the acquisition speed of the image acquisition device, generate polarization fusion images and multispectral fusion images respectively, execute defect detection algorithms, and perform post-fusion of the two-stage detection results to determine whether the sample is qualified.
  • Polarizing film 3 A new material 200mm*200mm linear polarizing film 3 is grown, which matches the coaxial light emitting area and is attached to the coaxial light emitting surface to obtain polarized light.
  • the testing device also includes a sample stage 4, which is placed on a conveyor belt 5 for placing OLED samples to be tested.
  • the light source controller 9, conveyor belt 5, polarization camera 1, and multispectral camera 8 are all connected to the industrial control computer 10.
  • the coaxial light source 2 and multispectral light source 7 are both connected to the light source controller 9.
  • the polarization camera 1, coaxial light source 2, and polarization film 3 constitute the first detection point through which the OLED sample passes, and the multispectral light source 7 and multispectral camera 8 constitute the second detection point through which the OLED sample passes.
  • Detection point one is used to detect cracks, scratches, foreign objects, and fragment defects on the surface of the OLED screen; detection point two is used to detect defects such as bubbles and dirt on the OLED screen.
  • the two detection points also include a fixing device for fixing the camera and light source at the two points to the main device.
  • the fixing device is connected to the polarization camera 1 and coaxial light source 2 at point one, and the fixing device is connected to the multispectral camera 8 and multispectral ring light at point two, to ensure the stability and accuracy of each point during the detection process.
  • the testing process is as follows:
  • the OLED sample is placed on conveyor belt 5 through an automatic feeding system and transported to point 1 for first-stage testing.
  • a coaxial light source 2 is provided at point 1, and a linear polarization film 3 is attached to the light-emitting surface.
  • the emitted coaxial light is uniformly reflected by the OLED sample.
  • the four-directional polarization image of the OLED screen is captured by the polarization camera 1 and input into the industrial control computer 10 for fusion processing to obtain a polarization fused image.
  • the back-end detection algorithm performs a first-stage defect detection.
  • the OLED sample is transferred to point two via conveyor belt 5 for the second stage of testing.
  • the multispectral light source 7 emits light of different spectral bands in a time-division manner under the control of the industrial control computer 10.
  • the multispectral camera 8 acquires multispectral images synchronously under the control of the light source controller 9 and the industrial control computer 10.
  • the multispectral images are fused in the industrial control computer 10 to obtain a multispectral fused image, which is then used for two-stage defect detection by the back-end detection algorithm.
  • the defect detection results of the first stage and the second stage are fused together, and the OLED sample is judged as qualified based on the fusion result.
  • the fusion process is divided into two stages.
  • the first stage involves polarization fusion based on four-directional polarization images and multispectral fusion based on multi-band images.
  • the second stage involves polarization and multispectral image fusion based on multi-scale decomposition.
  • the specific detection steps are as follows:
  • polarization camera 1 At the first detection point, use polarization camera 1 to collect polarization images of the OLED screen in four directions: 0°, 45°, 90° and 135°, respectively, and record them as I0 , I45 , I90 and I135 .
  • the Stokes parameter S is solved using the acquired polarization images in four directions, where...
  • a multispectral camera 8 to acquire images of the OLED screen in seven spectral bands: 405nm, 457nm, 527nm, 600nm, 660nm, 730nm and 860nm.
  • Image Registration Due to differences in acquisition equipment and shooting conditions, it is necessary to ensure that the two images are aligned in spatial position and resolution before fusing polarization-fused and multispectral-fused images. Therefore, image registration based on the SIFT algorithm is required.
  • Feature point detection The SIFT algorithm is applied to the polarization fusion image and the multispectral fusion image respectively to detect key points (i.e., feature points). These key points are scale invariant and rotation invariant.
  • Feature point description For each detected key point, calculate the gradient direction histogram of the surrounding area and generate a feature vector describing the key point.
  • Feature point matching Using Euclidean distance as a measure of similarity between feature vectors, matching feature point pairs are found between polarization fusion images and multispectral fusion images.
  • Transformation model estimation Based on the matched feature point pairs, the multispectral fusion image is mapped to the spatial location and resolution of the polarization fusion image using affine transformation.
  • Image resampling and interpolation The multispectral fused image is resampled and interpolated to generate a new image that is spatially and resolution-aligned with the polarization fused image.
  • Polarization and multispectral image fusion Wavelet transform is used to decompose the polarization fusion and multispectral fusion images into sub-images of different scales, which are then fused separately. Finally, an inverse transform is performed to generate the polarization multispectral fusion image.
  • Image decomposition Two-dimensional discrete wavelet transform (2D-DWT) is used to decompose the registered polarization fusion image and multispectral fusion image into low-frequency coefficient I, high-frequency coefficient I, low-frequency coefficient II, and high-frequency coefficient III respectively.
  • 2D-DWT Two-dimensional discrete wavelet transform
  • model attention mechanism In order to improve the model’s feature extraction and detection classification capabilities in small sample scenarios, the original YOLOV8 model was optimized by adding the CABM attention mechanism and fusing the CABM output with the original features to enhance the model’s generalization ability.
  • Model pre-training Using the transfer learning training strategy, the improved YOLOV8 model is pre-trained on the NEU-DET public dataset to obtain the pre-trained model, and then formally trained on the enhanced OLED dataset to obtain the trained detection model.
  • Model prediction and result output Input the fused image of the OLED to be detected into the model for defect detection. Based on the detection results, determine whether the OLED screen is OK or NG and output the corresponding judgment result.

Landscapes

  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Engineering & Computer Science (AREA)
  • Textile Engineering (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

一种多级多模态融合的OLED缺陷检测装置，针对传统的基于图像的检测方案精度受限的问题，在OLED图像采集阶段引入偏振和多光谱信息，增强缺陷特征，进而提升系统对于细微缺陷的检测能力，针对OLED屏体缺陷种类丰富，形态复杂多变，单一装置无法完全检出的问题，从缺陷的成因和材料特性出发，采用了双点位检测系统对不同种类的缺陷进行分阶段检测。利用偏振信息对待测物表面的粗糙度和形貌特征较为敏感的特性，使用基于偏振融合的检测点位一检测裂纹、划痕、异物和破片缺陷；利用气泡和脏污对光谱特性敏感的特点，使用基于多光谱融合的装置二检测OLED屏体上的气泡和脏污缺陷。

Description

一种多级多模态融合的OLED缺陷检测装置 技术领域
本发明涉及工件制造加工领域，具体地说，是一种多级多模态融合的OLED缺陷检测装置。
背景技术
OLED柔性屏的缺陷检测是其制造过程中至关重要的环节，对成品质量和制造过程分析具有重要意义。传统的目视检测方案人力成本高、效率低且易受主观因素影响。当下广泛应用的自动光学检测技术(Automatic Optic Inspection，AOI)多采用基于强度图像的机器视觉方案，这些检测方案大多存在以下问题：
1.镜面反射对测量结果带来干扰：由于OLED屏体表面光滑，为镜面反射物体，垂直拍摄的高分辨率图像中常常存在相机倒影，进而对检测结果带来较大干扰，若调整相机角度倾斜拍摄通常会牺牲一定的全视场分辨率，使测量精度偏低。
2.检测精度受限：由于OLED的材料特性，细微的缺陷都可能导致屏幕以较快的速度氧化损毁，而普通工业相机采集到的强度图像中的细微缺陷往往对比度较低，难以分辨，无法满足实际工业检测的需求。
3.检出种类不全：于OLED屏体的缺陷种类丰富，形态与特性复杂多变，单一检测装置往往无法将所有种类的缺陷检出。
发明内容
本发明针对现有技术的不足之处做出改进，对OLED屏体的六类缺陷：裂纹、划痕、异物、破片、气泡和脏污，设计了一种基于偏振和多光谱的多级多模态融合装置。
针对传统的基于图像的检测方案精度受限的问题，创新性地在OLED图像采集阶段引入偏振和多光谱信息，增强缺陷特征，进而提升系统对于细微缺陷的检测能力。
针对OLED屏体缺陷种类丰富，形态复杂多变，单一装置无法完全检出的问题，从缺陷的成因和材料特性出发，采用了双点位检测系统对不同种类的缺陷进行分阶段检测。利用偏振信息对待测物表面的粗糙度和形貌特征较为敏感的特性，使用基于偏振融合的检测点位一检测裂纹、划痕、异物和破片缺陷；利用气泡和脏污对光谱特性敏感的特点，使用基于多光谱融合的装置二检测OLED屏体上的气泡和脏污缺陷。
本发明是通过以下技术方案来实现的：
本发明公开了一种多级多模态融合的OLED缺陷检测装置，装置包括:
样品传输控制装置：包括传送带，用于自动上料、放置并传输OLED样品；
双点位图像采集装置：偏振相机和多光谱相机，分别用于基于偏振融合和多光谱融合的检测；
照明装置：同轴光源和多光谱光源，同轴光源用于发射同轴光至OLED样品上且产生反射光至偏振相机的视场范围；多光谱光源用于发射多光谱环形光至OLED样品上且且产生反射光至多光谱相机的视场范围；
光源控制器：用于控制同轴光源和多光谱光源的开关和亮度，以及控制多光谱光源发射的多光谱环形光发射频率与多光谱相机采集频率一致；
工控机：用于同步控制图像采集装置的采集速度，分别生成偏振融合图像和多光谱融合图像，执行缺陷检测算法，并将两阶段检测结果进行后融合，以判定样品是否合格；
偏振膜：贴附于同轴光源发光面，用于获取偏振光；
光源控制器、传送带、偏振相机和多光谱相机均与工控机相连，同轴光源和多光谱光源均与光源控制器相连，偏振相机、同轴光源和偏振膜构成OLED样品经过的检测第一点位，多光谱光源、多光谱相机构成OLED样品经过的检测第二点位。
作为进一步地改进，本发明所述的检测装置还包括样品台，所属的样品台置于传送带上，用于放置待检测的OLED样品。
作为进一步地改进，本发明所述的传送带用于稳定地输送样品通过两个检测点位；同轴光源用于提供均匀照明；多光谱环光用于提供宽范围光谱的照明。
作为进一步地改进，本发明所述的检测点位一用于检测OLED屏体表面的裂纹、划痕、异物和破片缺陷；检测点位用于检测OLED的屏体上的气泡和脏污等缺陷。
作为进一步地改进，本发明所述的同轴光源的发光面积大于等于样品尺寸的1.5倍，确保同轴光源发出的光线能够均匀的照射在样品表面。
作为进一步地改进，本发明所述的两个检测点位还包括固定装置，用于将两点位中的相机和光源固定在主体装置上，其中固定装置与点位一中的偏振相机和同轴光源相连，固定装置与点位二中的多光谱相机和多光谱环光相连，确保检测过程中各点位的稳定性和精度。
本发明的有益效果如下：
1.垂直大视场下高质量图像的获取：本发明通过使用大尺寸同轴光源，有效去除了采集到图像中相机倒影的干扰，解决了通常倾斜拍摄去反光策略中测量精度降低的问题，实现了垂直大视场下高质量图像的获取；
2.高质量偏振图像的获取：本发明通过在同轴光源上加装线偏振膜，解决了在偏振度较低的常规光源采集场景下，获取到偏振度和偏振觉图像中存在较多噪声的问题，实现了高质量偏振图像的获取；
3.提高了对细微缺陷的检测能力：本发明通过设计点位一段偏振融合装置，在图像中引入了偏振模态信息；通过设计点位二段多光谱融合装置，在图像中引入了光谱信息，两种模态信息的引入使得采集到图像中细微缺陷的特征得到增强，从而提升了细微缺陷的检出率；
4.实现高效、全面、可靠的OLED缺陷检测：本装置采用多级多模态融合装置，将检测过程分为基于偏振融合装置的一阶段检测和基于多光谱融合装置的的二阶段检测，并使用工控机对两阶段检测结果进行后融合，对OLED屏体的缺陷进行更为全面、准确的评估。这种多级检测结果融合的方法，有效提升了缺陷检测系统的召回率，降低了误检率，提供了更高效、全面、可靠的OLED屏体缺陷检测技术方案。
附图说明
图1是多级多模态融合检测装置示意图；
图中，1是偏振相机，2是同轴光源，3是偏振膜，4是样品台，5是传送带，7是多光谱光源，8是多光谱相机，9是光源控制器，10是工控机。
具体实施方式
如图1所示为本装置设计的多级多模态融合检测装置，主要包括样品传输控制装置、双点位图像采集装置、照明装置、光源控制器9、工控机10和偏振膜3等。
样品传输控制装置：包括传送带5和控制电机，采用臻域自动化多工位导轨线组，用于自动上料、放置并传输OLED样品；传送带5用于稳定地输送样品通过接下来检测的两个检测点位。
双点位采集装置：包括偏振相机1和多光谱相机8。偏振相机1采用海康MV-CH050-10UP工业相机和MVL-MF2528M-8MP工业相机镜头组，用来采集四向偏振图像；多光谱相机8采用基恩士CA-H500MX工业相机和CA-LH25工业相机镜头组，用来进行多谱段图像采集。
照明装置：包括东冠COAX-200X同轴光源2和基恩士CA-DRM10X多光谱光源7。同轴光源2用于提供均匀照明，发射同轴光至OLED样品上且产生反射光至偏振相机1的视场范围，同轴光源2的发光面积大于等于样品尺寸的1.5倍，确保同轴光源2发出的光线能够均匀的照射在样品表面。
多光谱光源7提供宽范围光谱的照明，发射多光谱环形光至OLED样品上且且产生反射光至多光谱相机8的视场范围。
光源控制器9：使用基恩士CA-DRM10X光源控制器9，用于控制同轴光源2和多光谱光源7的开关和亮度，以及控制多光谱光源7发射的多光谱环形光发射频率与多光谱相机8采集频率一致；
工控机10：使用戴尔灵越3020办公机组，用于同步控制图像采集装置的采集速度，分别生成偏振融合图像和多光谱融合图像，执行缺陷检测算法，并将两阶段检测结果进行后融合，以判定样品是否合格。
偏振膜3：长成新材料200mm*200mm线偏振膜3，与同轴光发光面积匹配，贴附于同轴光发光面，用于获取偏振光。
检测装置还包括样品台4，样品台4置于传送带5上，用于放置待检测的OLED样品。
光源控制器9、传送带5、偏振相机1和多光谱相机8均与工控机10相连，同轴光源2和多光谱光源7均与光源控制器9相连，偏振相机1、同轴光源2和偏振膜3构成OLED样品经过的检测第一点位，多光谱光源7、多光谱相机8构成OLED样品经过的检测第二点位。检测点位一用于检测OLED屏体表面的裂纹、划痕、异物和破片缺陷；检测点位用于检测OLED的屏体上的气泡和脏污等缺陷。
两个检测点位还包括固定装置，用于将两点位中的相机和光源固定在主体装置上，其中固定装置与点位一中的偏振相机1和同轴光源2相连，固定装置与点位二中的多光谱相机8和多光谱环光相连，确保检测过程中各点位的稳定性和精度。
检测流程如下：
1OLED样品通过自动上料系统放置在传送带5上，并传输至点位一进行一阶段检测。
2点位一中设有同轴光源2，并在发光面上贴有线偏振膜3，发射出的同轴光均匀地被OLED样品反射，通过偏振相机1捕获OLED屏幕的四向偏振图像，并输入工控机10中进行融合处理得到偏振融合图像，经后端检测算法进行一阶段缺陷检测。
3一阶段检测完成后，OLED样品通过传送带5继续传输至点位二进行二阶段检测。
4点位二中的多光谱光源7在工控机10的控制下分时发出不同谱段的光，多光谱相机8通过光源控制器9和工控机10的控制下与同步采集得到多谱段图像，多谱段图像在工控机10进行融合处理得到多光谱融合图像，经后端检测算法进行二阶段缺陷检测。
5在工控机10中将一阶段和二阶段的缺陷检测结果进行后融合，并根据融合结果判定OLED样品是否合格。
具体来说主要分为两阶段的融合，第一阶段为基于四向偏振图像的偏振融合和基于多谱段图像的多光谱融合，第二阶段为基于多尺度分解的偏振与多光谱图像融合。具体检测步骤如下：
1.偏振图像采集与处理：
(1)在检测第一点位，使用偏振相机1，分别采集OLED屏幕在0°，45°，90°和135°四个方向的偏振图像，记为I0，I45，I90，I135；
(2)根据Stokes参量的计算公式，利用采集到的四个方向的偏振图像求解出Stokes参量S，其中
(3)由Stokes参量计算出线偏振度(Degree of Linear Polarization，DoLP)和偏振角(Angle of Polarization，AoP)，计算公式分别为
(4)将S0、DoLP和AoP映射到HSV颜色空间，从而得到偏振融合图像。
2.多光谱图像采集与处理：
(5)在检测第二点位，使用多光谱相机8采集405nm、457nm、527nm、600nm、660nm、730nm、和860nm七个谱段下OLED屏体的图像；
(6)对采集到的多光谱图像进行加权平均处理，生成多光谱融合图像。加权平均的计算公式为其中wi为各谱段图像的权重，Pi(x,y)为各谱段图像在位置(x,y)的像素值。
3.图像配准：因采集设备和拍摄条件的不同，在偏振融合图像和多光谱融合图像进行融合之前，需要确保两幅图像在空间位置和分辨率上是对齐的。因此，需要基于SIFT算法进行图像配准。
(7)特征点检测：在偏振融合图像和多光谱融合图像上分别应用SIFT算法，检测关键点(即特征点)。这些关键点具有尺度不变性和旋转不变性。
(8)特征点描述：对于每个检测到的关键点，计算其周围区域的梯度方向直方图，并生成一个描述该关键点的特征向量。
(9)特征点匹配：使用欧氏距离作为特征向量之间的相似度的度量，在偏振融合图像和多光谱融合图像之间找到匹配的特征点对。
(10)变换模型估计：基于匹配的特征点对，利用仿射变换将多光谱融合图像映射到偏振融合图像的空间位置和分辨率上。
(11)图像重采样与插值：对多光谱融合图像进行重采样和插值处理，以生成与偏振融合图像在空间和分辨率上对齐的新图像。
(12)配准优化与验证：使用RANSAC(Random Sample Consensus)算法剔除错误的匹配点对，并优化变换模型。
4.偏振与多光谱图像融合：使用小波变换对偏振融合和多光谱融合图像进行多尺度分解，得到不同尺度的子图像，分别进行融合处理，最后再逆变换生成偏振多光谱融合图像。
(13)图像分解：使用二维离散小波变换(2D-DWT)分别对配准后的偏振融合图像和多光谱融合图像进行多尺度分解，分解为低频系数I、高频系数I、低频系数II,高频系数III；
(14)融合操作：对于近似分量(低频信息)，由于它包含了图像的主要结构和轮廓信息，因此可以采用加权平均进行融合，得到融合后的低频系数；对于细节分量(高频信息)，它包含了图像的纹理和边缘信息，因此采用基于局部对比度进行融合，得到融合后的高频系数；
(15)逆变换与重构：将融合后的近似分量和细节分量通过逆二维离散小波变换(2D-IDWT)进行重构，生成最终的偏振多光谱融合图像。
5.缺陷检测与合格判定：
(16)数据增强：基于生成的偏振多光谱融合图像，平移、旋转、调整亮度、添加随机噪声、随机裁剪等五种数据增强策略，将采集到的OLED缺陷数据集扩充至五倍；
(17)模型注意力机制引进：为了提升模型在小样本场景下的特征提取和检测分类能力，在YoloV8原有模型基础上进行了优化，加入CABM注意力机制，并将CABM的输出与原始特征进行融合，以增强模型的泛化能力；
(18)模型预训练：使用迁移学习的训练策略，在NEU-DET公开数据集上对改进的YoloV8模型进行预训练，得到pre-train模型，然后再在增强后的OLED数据集上正式训练，得到训练好的检测模型。
(19)模型预测与结果输出：将待检测的OLED融合图像输入模型中进行缺陷检测，根据检测结果，判断OLED屏幕为OK还是NG件，并输出相应的判定结果。
以上仅为本发明的较佳实施例而已，并不用以限制本发明，凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等，均应包含在本发明的保护范围之内。

Claims (6)

  1. 一种多级多模态融合的OLED缺陷检测装置，其特征在于，所述装置包括:
    样品传输控制装置：包括传送带(5)，用于自动上料、放置并传输OLED样品；
    双点位图像采集装置：偏振相机(1)和多光谱相机(8)，分别用于基于偏振融合和多光谱融合的检测；
    照明装置：同轴光源(2)和多光谱光源(7)，所述的同轴光源用于发射同轴光至OLED样品上且产生反射光至偏振相机的视场范围；所述的多光谱光源用于发射多光谱环形光至OLED样品上且且产生反射光至多光谱相机的视场范围；
    光源控制器(9)：用于控制同轴光源(2)和多光谱光源(7)的开关和亮度，以及控制多光谱光源发射的多光谱环形光发射频率与多光谱相机采集频率一致；
    工控机(10)：用于同步控制图像采集装置的采集速度，分别生成偏振融合图像和多光谱融合图像，执行缺陷检测算法，并将两阶段检测结果进行后融合，以判定样品是否合格；
    偏振膜(3)：贴附于同轴光源(2)发光面，用于获取偏振光；
    所述的光源控制器(9)、传送带(5)、偏振相机和多光谱相机均与工控机相连，所述的同轴光源和多光谱光源均与光源控制器(9)相连，所述的偏振相机、同轴光源和偏振膜构成OLED样品经过的检测第一点位，所述的多光谱光源、多光谱相机构成OLED样品经过的检测第二点位。
  2. 根据权利要求1所述的OLED缺陷检测装置，其特征在于，所述的检测装置还包括样品台，所述的样品台置于传送带上，用于放置待检测的OLED样品。
  3. 根据权利要求1或2所述的OLED缺陷检测装置，其特征在于，所述的传送带用于稳定地输送样品通过两个检测点位；所述的同轴光源用于提供均匀照明；所述的多光谱环光(7)用于提供宽范围光谱的照明。
  4. 根据权利要求3所述的OLED缺陷检测装置，其特征在于，所述检测点位一用于检测OLED屏体表面的裂纹、划痕、异物和破片缺陷；所述的检测点位用于检测OLED的屏体上的气泡和脏污等缺陷。
  5. 根据权利要求1或2或4所述的OLED缺陷检测装置，其特征在于，所述的同轴光源的发光面积大于等于样品尺寸的1.5倍，确保同轴光源发出的光线能够均匀的照射在样品表面。
  6. 根据权利要求5所述的OLED缺陷检测装置，其特征在于，所述的两个检测点位还包括固定装置(11)和(12)，用于将两点位中的相机和光源固定在主体装置上，其中固定装置(11)与点位一中的偏振相机(1)和同轴光源(2)相连，固定装置(12)与点位二中的多光谱相机(8)和多光谱环光(7)相连，确保检测过程中各点位的稳定性和精度。
PCT/CN2025/090784 2024-07-21 2025-04-24 一种多级多模态融合的oled缺陷检测装置 Pending WO2026020907A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202410976724.3 2024-07-21
CN202410976724.3A CN118914221A (zh) 2024-07-21 2024-07-21 一种多级多模态融合的oled缺陷检测装置

Publications (1)

Publication Number Publication Date
WO2026020907A1 true WO2026020907A1 (zh) 2026-01-29

Family

ID=93297093

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2025/090784 Pending WO2026020907A1 (zh) 2024-07-21 2025-04-24 一种多级多模态融合的oled缺陷检测装置

Country Status (2)

Country Link
CN (1) CN118914221A (zh)
WO (1) WO2026020907A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118914221A (zh) * 2024-07-21 2024-11-08 浙江大学 一种多级多模态融合的oled缺陷检测装置
CN119951764A (zh) * 2025-03-03 2025-05-09 杭州国辰机器人科技有限公司 一种基于铝板尺寸及外观表面缺陷的视觉检测系统及方法
CN119780107B (zh) * 2025-03-10 2025-05-09 湖南汇成鑫科技有限公司 一种基于光学-磁学多模态融合的oled屏幕缺陷检测方法
CN119880802B (zh) * 2025-03-26 2025-07-04 登景(天津)科技有限公司 应用于显示面板内部玻璃基板的缺陷检测方法及装置

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110441312A (zh) * 2019-07-30 2019-11-12 上海深视信息科技有限公司 一种基于多光谱成像的产品表面缺陷检测系统
WO2020215199A1 (zh) * 2019-04-23 2020-10-29 合刃科技（深圳）有限公司 一种表面缺陷检测系统及方法
US20220214286A1 (en) * 2021-01-03 2022-07-07 Araz Yacoubian Surface Inspection Sensor
CN115598059A (zh) * 2022-08-26 2023-01-13 宁波永新光学股份有限公司（Cn） 用于半导体缺陷检测的图像信息的获取装置及方法
CN115731456A (zh) * 2022-11-15 2023-03-03 福州大学 基于快照式光谱偏振相机的目标检测方法
CN116448769A (zh) * 2023-05-10 2023-07-18 南京林业大学 一种多模态信息融合的板材缺陷检测系统及其检测方法
CN117191816A (zh) * 2023-11-08 2023-12-08 广东工业大学 基于多光谱融合的电子元器件表面缺陷检测方法和装置
CN118914221A (zh) * 2024-07-21 2024-11-08 浙江大学 一种多级多模态融合的oled缺陷检测装置
CN118982500A (zh) * 2024-07-21 2024-11-19 浙江大学 一种偏振多光谱融合的oled缺陷检测方法

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6690469B1 (en) * 1998-09-18 2004-02-10 Hitachi, Ltd. Method and apparatus for observing and inspecting defects
CN101819165B (zh) * 2009-02-27 2013-08-07 圣戈本玻璃法国公司 用于检测图案化基板的缺陷的方法及系统
CN101587079B (zh) * 2009-05-08 2011-01-12 清华大学 纺织原料中异性杂质的多光谱色偏振检测方法及其装置
CN104075998A (zh) * 2014-07-03 2014-10-01 西北工业大学 一种近红外偏振隐形防伪元件辨识系统
CN104122209B (zh) * 2014-07-25 2017-02-15 南京信息工程大学 微晶硅薄膜生长过程的可视化观测系统及测量方法
CN214374364U (zh) * 2021-01-27 2021-10-08 武汉精立电子技术有限公司 一种基于光学成像的缺陷检测装置

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020215199A1 (zh) * 2019-04-23 2020-10-29 合刃科技（深圳）有限公司 一种表面缺陷检测系统及方法
CN110441312A (zh) * 2019-07-30 2019-11-12 上海深视信息科技有限公司 一种基于多光谱成像的产品表面缺陷检测系统
US20220214286A1 (en) * 2021-01-03 2022-07-07 Araz Yacoubian Surface Inspection Sensor
CN115598059A (zh) * 2022-08-26 2023-01-13 宁波永新光学股份有限公司（Cn） 用于半导体缺陷检测的图像信息的获取装置及方法
CN115731456A (zh) * 2022-11-15 2023-03-03 福州大学 基于快照式光谱偏振相机的目标检测方法
CN116448769A (zh) * 2023-05-10 2023-07-18 南京林业大学 一种多模态信息融合的板材缺陷检测系统及其检测方法
CN117191816A (zh) * 2023-11-08 2023-12-08 广东工业大学 基于多光谱融合的电子元器件表面缺陷检测方法和装置
CN118914221A (zh) * 2024-07-21 2024-11-08 浙江大学 一种多级多模态融合的oled缺陷检测装置
CN118982500A (zh) * 2024-07-21 2024-11-19 浙江大学 一种偏振多光谱融合的oled缺陷检测方法

Also Published As

Publication number Publication date
CN118914221A (zh) 2024-11-08

Similar Documents

Publication Publication Date Title
CN118982500B (zh) 一种偏振多光谱融合的oled缺陷检测方法
CN111612737B (zh) 一种人造板表面瑕疵检测装置及检测方法
CN118914221A (zh) 一种多级多模态融合的oled缺陷检测装置
WO2023077404A1 (zh) 缺陷检测方法、装置和系统
CN108490000A (zh) 一种棒线材表面缺陷在线检测装置和方法
Wei et al. Surface defects detection of cylindrical high-precision industrial parts based on deep learning algorithms: a review
CN113608378A (zh) 一种基于lcd制程的全自动缺陷检测方法和系统
CN119887788B (zh) 一种纺织品状态检测分析系统
CN111551350A (zh) 一种基于U_Net网络的光学镜片表面划痕检测方法
CN118837379A (zh) 一种智能眼镜的高精度缺陷检测方法及系统
CN108205210B (zh) 基于傅里叶梅林及特征匹配的lcd缺陷检测系统及方法
CN120253841A (zh) 一种基于机器视觉的键盘按键缺陷检测方法及系统
CN119693841B (zh) 基于视频深度学习的内镜管道内壁裂隙自动识别系统
CN120912621B (zh) 一种柔性显示模组表面缺陷图像识别方法
CN118882520B (zh) 一种大口径曲面光学元件表面缺陷三维检测装置及方法
CN115272256A (zh) 亚像素级传感光纤路径高斯提取方法及系统
CN111739003B (zh) 一种用于外观检测的机器视觉方法
CN120339701B (zh) 一种半导体封装中的表面贴装检测系统
CN119723200A (zh) 缺陷类型的检测方法和装置、电子设备及存储介质
CN117934453B (zh) 一种手机屏背光异物缺陷诊断方法及系统
CN119784741A (zh) 一种用于5g柔体线路板缺陷检测模型的训练方法及系统
CN120894360A (zh) 一种基于显微复检的oled面板检测方法
CN110186929A (zh) 一种实时的产品瑕疵定位方法
CN110412056A (zh) 一种车载玻璃模组自动光学检测方法及装置
CN120471914B (zh) 基于面阵相机移动设备的屏幕弧边区域缺陷检测系统及方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 25843297

Country of ref document: EP

Kind code of ref document: A1