WO2020029237A1 - Procédé et système de détection - Google Patents

Procédé et système de détection Download PDF

Info

Publication number
WO2020029237A1
WO2020029237A1 PCT/CN2018/099921 CN2018099921W WO2020029237A1 WO 2020029237 A1 WO2020029237 A1 WO 2020029237A1 CN 2018099921 W CN2018099921 W CN 2018099921W WO 2020029237 A1 WO2020029237 A1 WO 2020029237A1
Authority
WO
WIPO (PCT)
Prior art keywords
coherent light
measured
speckle image
multiple beams
glass
Prior art date
Application number
PCT/CN2018/099921
Other languages
English (en)
Chinese (zh)
Inventor
王星泽
舒远
Original Assignee
合刃科技（深圳）有限公司
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 合刃科技（深圳）有限公司 filed Critical 合刃科技（深圳）有限公司
Priority to CN201880067184.XA priority Critical patent/CN111226110A/zh
Priority to PCT/CN2018/099921 priority patent/WO2020029237A1/fr
Publication of WO2020029237A1 publication Critical patent/WO2020029237A1/fr

Links

Images

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

Definitions

  • the present application relates to the field of electronics, and in particular, to a detection method and system.
  • glass defect detection systems mainly use manual online detection, laser detection, Moore interference principle, and machine vision methods. Manual inspection is susceptible to subjective factors of inspectors, and it is easy to miss inspections of glass defects, especially missing inspections of inclusion defects with small deformation and small distortion. Laser detection is susceptible to external interference, which affects the accuracy of the detection.
  • the Moore interference principle requires relatively high light and dark contrast of the grating due to the thin Moiré fringes in the grating, otherwise the error is large; and the machine vision-based method uses CCD imaging technology and backlight In type lighting, a light source is placed on the back of the glass. The light passes through the glass to be inspected and is transmitted to the camera. When the glass contains impurities, the emitted light will change, so the light detected on the target surface of the CCD camera will change accordingly.
  • the present application provides a detection method and system, which can quickly detect curved glass.
  • a detection system including:
  • Coherent light source for generating coherent light
  • Reflector for adjusting the irradiation angle of part or all of the coherent light
  • a collection detector configured to collect a speckle image formed by the multiple beams of coherent light passing through the object to be measured or the multiple beams of coherent light being reflected by the object to be measured, and determine according to the speckle image Whether the glass is defective.
  • the collection detector is specifically configured to collect a speckle image formed by the multiple beams of coherent light passing through the object to be measured or the multiple beams of coherent light being reflected by the object to be measured and formed in a non-imaging manner .
  • the acquisition detector is further configured to determine whether there is a defect in the object to be measured according to the speckle image and a neural network of deep learning.
  • the de-coherent light includes a first de-coherent light and a second de-coherent light
  • the curved portion includes a first curved portion and a second curved portion, wherein the first de-coherent light is used to illuminate the light.
  • the first curved portion, and the second divided coherent light is used to irradiate the second curved portion.
  • the reflecting mirror includes a first reflecting mirror and a second reflecting mirror, wherein the first reflecting mirror is used to adjust an angle at which the first divided coherent light illuminates the first curved portion, and the first The two reflecting mirrors are used to adjust an angle at which the second divided coherent light illuminates the second bending portion.
  • the system further includes a first concave lens and a second concave lens, the first concave lens is used for diffusing the first divided coherent light, and the second concave lens is used for diffusing the second divided coherent light For diffusion.
  • the system further includes a beam expander, which is configured to diffuse the coherent light that irradiates the flat portion of the object to be measured.
  • a beam expander which is configured to diffuse the coherent light that irradiates the flat portion of the object to be measured.
  • the system further includes a prism mirror, which is configured to expand an irradiation area of the coherent light that irradiates the straight portion of the object to be measured.
  • a prism mirror which is configured to expand an irradiation area of the coherent light that irradiates the straight portion of the object to be measured.
  • the spectrum of the coherent light ranges from 215 to 2000 nanometers.
  • a detection method including:
  • Coherent light source produces coherent light
  • a beam splitter divides the coherent light into multiple beams of coherent light, wherein the multiple beams of coherent light are used to illuminate an object to be measured;
  • the reflector adjusts the irradiation angle of part or all of the coherent light
  • a collection detector collects a speckle image formed by the multiple beams of coherent light passing through the object to be measured or the multiple beams of coherent light are reflected by the object to be measured, and determines the glass according to the speckle image Whether there are defects.
  • a speckle image formed by transmitting the multiple beams of coherent light through the object to be measured or the multiple beams of coherent light being reflected by the object to be measured and formed in a non-imaging manner is collected.
  • the method further comprises: the acquisition detector determines whether there is a defect in the object to be measured according to the speckle image and a deep learning neural network; wherein the neural network uses a large sample of speckles Images for training and learning.
  • the de-coherent light includes a first de-coherent light and a second de-coherent light
  • the curved portion includes a first curved portion and a second curved portion, wherein the first de-coherent light is used to illuminate the light.
  • the first curved portion, and the second divided coherent light is used to irradiate the second curved portion.
  • the reflecting mirror includes a first reflecting mirror and a second reflecting mirror, wherein the first reflecting mirror is used to adjust an angle at which the first divided coherent light illuminates the first curved portion, and the first The two reflecting mirrors are used to adjust an angle at which the second divided coherent light illuminates the second bending portion.
  • the system further includes a first concave lens and a second concave lens, the first concave lens is used for diffusing the first divided coherent light, and the second concave lens is used for diffusing the second divided coherent light For diffusion.
  • the method further includes: diffusing the coherent light that irradiates the straight portion of the object to be measured through a beam expander.
  • the method further includes: expanding the irradiation area of the coherent light that irradiates the straight portion of the object to be measured by a prism mirror.
  • the spectrum of the coherent light ranges from 215 to 2000 nanometers.
  • coherent light is generated by a coherent light source, and then the coherent light is divided into a plurality of beams of coherent light by a beam splitter, and the angle of the beam of coherent light that irradiates the curved part of the object to be measured is collected by a reflector,
  • the detector collects a speckle image formed by the multiple beams of coherent light passing through the object to be measured or the multiple beams of coherent light are reflected by the object to be measured.
  • the acquisition detector determines the speckle image based on the speckle image. Whether there is a defect in the object to be tested. It is not difficult to see that the above solution can quickly detect curved glass.
  • FIG. 1 is a schematic diagram of changes in light transmission through a curved glass in the prior art
  • FIG. 2 is a schematic structural diagram of a detection system proposed in the present application.
  • FIG. 3 is a comparison diagram of light intensity when there are no air bubbles and when there are air bubbles in the present application;
  • FIG. 4 is a schematic diagram of a neural network proposed by the present application.
  • FIG. 5 is a schematic flowchart of an inspection method proposed in the present application.
  • 3D glass is flat in the middle and gradually curved on both sides.
  • the curved part of 3D glass is caused by heating and bending of flat glass, which is prone to various damages, such as scratches, surface residues, depressions, etc. Therefore, the probability of problems in the curved part is much higher than the probability of problems in the straight part. Need to focus on testing to avoid a large number of bad products.
  • Figure 1 is a comparison between 2D glass and 3D glass after parallel backlight detection.
  • the C (scratch) of 3D glass in a curved place cannot be detected because the backlight penetrates too much glass and the light intensity is attenuated. .
  • the present application proposes a detection system and method, which can quickly detect curved glass, and improve the accuracy of detection of curved glass.
  • the present application provides a schematic structural diagram of a detection system.
  • the detection system of the present application includes a coherent light source 110, a beam splitter 120, a reflector 130, and a collection detector 140.
  • the coherent light source 110 generates coherent light and sends the coherent light to the beam splitter 120.
  • the beam splitter 120 divides the coherent light into a plurality of beams of coherent light, wherein the plurality of beams of coherent light is used to illuminate an object to be measured.
  • the reflecting mirror 130 adjusts the irradiation angle of part or all of the coherent light.
  • the acquisition detector 140 collects a speckle image formed by the multiple beams of coherent light passing through the object to be measured or the multiple beams of coherent light are reflected by the object to be measured, and determines the speckle image according to the speckle image Whether the object to be tested is defective.
  • the object to be measured is described as glass.
  • the coherent light generated by the coherent light source 110 is linearly polarized light having the same frequency and the same vibration direction. Because the non-imaging speckle image is used for monitoring, there is no dispersion problem of the lens of the imaging system in different spectral bands. In this way, the value range of the coherent light spectrum can be relatively large.
  • the value range of the coherent light spectrum can be 215-2000 nm. That is, the spectrum of coherent light can extend from ultraviolet to near-infrared light. It can be understood that the value range of the above-mentioned coherent light spectrum is merely an example, and should not constitute a specific limitation.
  • Coherent speckle defect detection is used to obtain more defect information on the surface of the inspected item, such as intensity information, phase information, and incident angle information after the defect is reflected, so as to identify more defects that cannot be detected by traditional methods, such as fine scratching. Injuries, chipping, internal bubbles, etc.
  • the beam splitter 120 may include one or more beam splitters.
  • the beam splitter When the beam splitter splits the coherent light, it will distribute multiple beams of the coherent light according to the proportion of the optical power.
  • the optical power of each beam of coherent light can be determined by the area of the glass irradiated by each beam of coherent light. For example, if the ratio of the area of the glass illuminated by the first decoherent light and the area of the glass illuminated by the second decoherent light is 2: 1, then the optical power of the first decoherent light and the optical power of the second decoherent light The ratio is also 2: 1. It can be understood that the first and second coherent lights are obtained from the same coherent light. Therefore, it is strictly guaranteed that the frequencies and vibration directions of the first and second coherent lights are consistent.
  • the beam splitter 120 includes a first beam splitter 121 and a second beam splitter 122. It should be understood that the above-mentioned beam splitter 120 is merely an example. In other embodiments, the number of beam splitters 120 may be less or more, which is not specifically limited herein.
  • the reflecting mirror 130 includes a first reflecting mirror 131 and a second reflecting mirror 132. It should be understood that the above-mentioned reflecting mirror 130 is merely an example. In other embodiments, the number of reflecting mirrors 130 may be less or more, which is not specifically limited herein.
  • the detection system further includes a concave lens, wherein the concave lens is used to diffuse the coherent light.
  • the concave lens includes a first concave lens 151 and a second concave lens 152. It should be understood that the foregoing concave lens is merely an example. In other embodiments, the number of the concave lens may be less or more, which is not specifically limited herein.
  • the detection system further includes a beam expander 160, wherein the beam expander 160 is configured to diffuse the coherent light that irradiates the flat portion of the object to be measured.
  • the detection system further includes a prism mirror for expanding an irradiation area of the coherent light that irradiates the flat portion of the object to be measured.
  • the coherent light generated by the coherent light source 110 is incident on the first beam splitter 121.
  • the first beam splitter 121 separates the first divided coherent light from the coherent light.
  • the first concave lens 151 is disposed on the optical path of the first divided coherent light, and the first divided coherent light passes through the axis of the first concave lens 151.
  • the first concave lens 151 diffuses the first divided coherent light to obtain the diffused first divided coherent light.
  • the first reflecting mirror 131 is disposed on the optical path of the diffused first divided coherent light, and the first reflecting mirror 131 reflects the diffused first divided coherent light so that the reflected first divided coherent light is illuminated The first bend (the left wing portion of the 3D glass).
  • the second beam splitter 122 After passing through the first beam splitter 121, the remaining coherent light is incident on the second beam splitter 122.
  • the second beam splitter 122 separates the second divided coherent light from the remaining coherent light.
  • the second concave lens 152 is disposed on the optical path of the second divided coherent light, and the second divided coherent light passes through the axis of the second concave lens 152.
  • the second concave lens 152 diffuses the second divided coherent light to obtain the diffused second divided coherent light.
  • the second reflector 132 is disposed on the optical path of the diffused second divided coherent light, and the second reflector 132 reflects the diffused second divided coherent light so that the reflected second divided coherent light is irradiated
  • the second curved portion (the right wing portion of the 3D glass).
  • the remaining split coherent light After passing through the second beam splitter 122, the remaining split coherent light enters the beam expander 160.
  • the beam expander 160 is configured to diffuse the remaining split coherent light to obtain the diffused split coherent light, and irradiate the flat portion (the middle portion of the 3D glass).
  • the first concave lens, the second concave lens, and the beam expander diffuse the first, second, and remaining split coherent light, respectively, so that the light can be more uniformly irradiated on the first The bent portion, the second bent portion, and the straight portion.
  • the above detection system is only a specific embodiment. In other embodiments, it may further include more reflecting mirrors, concave lenses, beam expanders, and the like. Only the first split coherent light and the second The decoherent light and the remaining decoherent light perform transillumination on the glass of the curved part and the straight part in a vertical manner as much as possible, which is not specifically limited here.
  • the acquisition detector 140 includes one or more photoelectric sensors, wherein the photoelectric sensors are used to acquire speckle images.
  • the positions and numbers of the photoelectric sensors can be set according to actual needs, and are not specifically limited here.
  • the collection detector 140 is specifically configured to collect the multiple beams of coherent light through the object to be measured or the multiple beams of coherent light are reflected by the object to be measured and in a non-imaging manner Speckle image formed.
  • the non-imaging method means that it is not necessary to perform a restoration calculation on the 3D glass according to the speckle image to determine whether there is a defect, but to directly determine whether a defect exists based on the speckle image calculation. It can be understood that when the speckle image is formed in a non-imaging form, the defect information is reflected in the speckle image.
  • Resolving and reconstructing the real image of the object such as phase solving, directly discriminates and detects defects on the speckle image, which can effectively reduce the amount of data calculation and increase the speed of recognition.
  • the speckle image is an image formed when light passes through the optically rough surface of the vibrating object or light is reflected by the optically rough surface of the vibrating object.
  • optically rough surfaces or transmission plates with optically rough transmission
  • the wavelets scattered by irregularly distributed surfaces on these surfaces are superimposed on each other.
  • make the reflected light field (or transmitted light field) have a random spatial light intensity distribution and present a granular structure, which is speckle. For example, as shown in FIG. 3, when a bubble appears inside the curved surface of the glass, the speckle distribution on the image collector changes.
  • the light intensity in the area where bubbles are present in the glass will increase significantly.
  • surface scratches, residues, pits, etc. can cause changes in light intensity.
  • these defects are generally in the range of 10 micrometers to 5 millimeters.
  • the detection accuracy can reach more than 5 micrometers, so it is sufficient to check most defects.
  • the shape of glass defects is complex and changeable. These features cannot well represent the defect target. If the speckle image with a glass defect structure is directly subjected to phase inversion to obtain an image restoration image, the restored image will be greatly distorted, and it is almost impossible to identify whether there is a defect in the glass. Therefore, in this application, the speckle image can be input into a neural network to determine whether a defect exists in the glass.
  • the speckle image initially collected by the acquisition detector 140 is a speckle image full of noise.
  • the useful information in the speckle image is submerged in a large amount of noise. Therefore, the acquisition detector 140 needs to process the speckle image collected initially to obtain a processed speckle image to remove speckle noise and improve fringe contrast.
  • methods for processing an image include a phase shift method, a fringe grayscale method, a fringe centerline method, a Fourier transform method, a sub-pixel search method, and so on. It should be understood that the foregoing processing methods are merely examples, and should not constitute a specific limitation.
  • the speckle image may include a straight speckle image and a curved speckle image, which are not specifically limited herein.
  • the flat speckle image may be one or more.
  • the number of straight speckle images can be one; when the area of the straight portion is relatively small, the number of straight speckle images can be multiple, and different straight speckles The image corresponds to different areas of the straight portion.
  • the number of straight speckle images can be set according to the area of the flat portion of the glass, which is not specifically limited this time.
  • the curved speckle image may be one or more.
  • the number of curved speckle images when the curved portions are concentrated in the same area, the number of curved speckle images may be one; when the curved portions are scattered in multiple regions, the number of curved speckle images may be multiple, and different curved speckle images correspond to Different areas of the bend.
  • the number of curved speckle images can be set according to the area where the curved portion of the glass is scattered, which is not specifically limited this time.
  • the acquisition detector 140 is configured to determine whether the glass has a defect according to the speckle image and a deep learning neural network.
  • the input of the deep learning neural network includes a first curved speckle image, a second curved speckle image, a first flat speckle image, and a second flat speckle image.
  • the example of the input of the neural network is merely an example. In actual applications, the speckle image input by the neural network may be more or less, which is not specifically limited herein.
  • the output results of the deep learning neural network may include: no defects, scratches, bubbles, dirt, etc.
  • the output results may also be expressed using fewer or more levels. . It can be understood that the above-mentioned level division is only used as an example, and the more the level division is, the more accurate the output result is represented.
  • the deep learning neural network includes multiple training models.
  • the training model may include a defect-free training model, a bubble model, a dirty model, and so on.
  • the above training model is merely an example, and in actual applications, more or less training models may be included, which is not specifically limited herein.
  • the neural network can be BP neural network, Hopfield network, ART network, Kohonen network, Long Short-Term Memory (LSTM), Residual Network (ResNet), Recurrent Neural Networks , RNN), etc., are not specifically limited here.
  • the output when the input includes a first curved speckle image, a second curved speckle image, a first flat speckle image, and a second flat speckle image, the output includes no defects, scratches, and bubbles.
  • the dirty, deep learning neural network can be shown in Figure 4.
  • the deep learning neural network may be trained and learned using a large sample of speckle images. For example, a large number of speckle images are collected according to different defect samples in advance, and the deep neural network is used to classify and train these speckle images that can indirectly reflect the fine structure of glass to obtain the correct neural network. During recognition, the collected speckle images are input to the trained neural network to obtain the recognition results.
  • coherent light is generated by a coherent light source, and then the coherent light is divided into multiple beams of coherent light by a beam splitter, and the irradiation angle of part or all of the coherent light is adjusted by a mirror, and the acquisition detector collects the multi A beam of decoherent light passes through the object to be measured or a speckle image formed by the reflection of the multi-beam of decoherent light by the object to be measured, and finally, the acquisition detector determines whether the object to be measured is based on the speckle image Flawed. It is not difficult to see that the above solution can quickly detect curved glass.
  • FIG. 5 is a schematic flowchart of a detection method provided by the present application.
  • the object to be measured is described as glass.
  • the detection method in this embodiment includes the following steps:
  • the coherent light source generates coherent light.
  • the coherent light generated by the coherent light source is linearly polarized light having the same frequency and the same vibration direction. Because the non-imaging speckle image is used for monitoring, there is no dispersion problem of the lens of the imaging system in different spectral bands. In this way, the value range of the coherent light spectrum can be relatively large. 215-2000 nm. That is, the spectrum of coherent light can extend from ultraviolet to near-infrared light. It can be understood that the value range of the above-mentioned coherent light spectrum is merely an example, and should not constitute a specific limitation.
  • Coherent speckle defect detection is used to obtain more defect information on the surface of the inspected item, such as intensity information, phase information, and incident angle information after the defect is reflected, so as to identify more defects that cannot be detected by traditional methods, such as fine scratching. Injuries, chipping, internal bubbles, etc.
  • a beam splitter divides the coherent light into multiple beams of coherent light, wherein the multiple beams of coherent light are used to illuminate an object to be measured.
  • the beam splitter may include one or more beam splitters.
  • the optical splitter splits coherent light, it will distribute multiple beams of split coherent light according to the proportion of the optical power.
  • the optical power of each beam of coherent light can be determined by the area of the glass irradiated by each beam of coherent light. For example, if the ratio of the area of the glass illuminated by the first decoherent light and the area of the glass illuminated by the second decoherent light is 2: 1, then the optical power of the first decoherent light and the optical power of the second decoherent light The ratio is also 2: 1. It can be understood that the first and second coherent lights are obtained from the same coherent light. Therefore, it is strictly guaranteed that the frequencies and vibration directions of the first and second coherent lights are consistent.
  • the beam splitter includes a first beam splitter and a second beam splitter. It should be understood that the above-mentioned beam splitter is merely an example. In other embodiments, the number of beam splitters may be less or more, which is not specifically limited herein.
  • the reflector adjusts the irradiation angle of part or all of the coherent light.
  • the reflecting mirror includes a first reflecting mirror and a second reflecting mirror. It should be understood that the above-mentioned reflecting mirror is merely an example, and in other embodiments, the number of reflecting mirrors may be less or more, which is not specifically limited herein.
  • the detection system further includes a concave lens, wherein the concave lens is used to diffuse the coherent light.
  • the concave lens includes a first concave lens and a second concave lens. It should be understood that the foregoing concave lens is merely an example. In other embodiments, the number of the concave lens may be less or more, which is not specifically limited herein.
  • the detection system further includes a beam expander, wherein the beam expander is configured to diffuse the coherent light that irradiates the flat portion of the object to be measured.
  • the detection system further includes a prism mirror for expanding an irradiation area of the coherent light that irradiates the flat portion of the object to be measured.
  • the coherent light generated by the coherent light source is incident on the first beam splitter.
  • the first beam splitter separates the first divided coherent light from the coherent light.
  • the first concave lens is disposed on the optical path of the first divided coherent light, and the first divided coherent light passes through the axis of the first concave lens.
  • the first concave lens diffuses the first divided coherent light to obtain the diffused first divided coherent light.
  • the first reflector is disposed on the optical path of the diffused first divided coherent light, and the first reflector reflects the diffused first divided coherent light so that the reflected first divided coherent light irradiates the first Bend (left wing part of 3D glass). After passing through the first beam splitter, the remaining coherent light is incident on the second beam splitter.
  • the second beam splitter separates the second coherent light from the remaining coherent light.
  • the second concave lens is disposed on the optical path of the second divided coherent light, and the second divided coherent light passes through the axis of the second concave lens.
  • the second concave lens diffuses the second divided coherent light to obtain a diffused second divided coherent light.
  • the second reflector is disposed on the optical path of the diffused second divided coherent light, and the second reflector reflects the diffused second divided coherent light so that the reflected second divided coherent light irradiates the second Bend (right wing part of 3D glass). After passing through the second beam splitter, the remaining split coherent light enters the beam expander.
  • the beam expander is configured to diffuse the remaining split coherent light to obtain the diffused split coherent light, and irradiate the flat portion (the middle portion of the 3D glass).
  • the first concave lens, the second concave lens, and the beam expander diffuse the first, second, and remaining split coherent light, respectively, so that the light can be more uniformly irradiated on the first The bent portion, the second bent portion, and the straight portion.
  • the above detection system is only a specific embodiment. In other embodiments, it may further include more reflecting mirrors, concave lenses, beam expanders, and the like. Only the first split coherent light and the second The decoherent light and the remaining decoherent light perform transillumination on the glass of the curved part and the straight part in a vertical manner as much as possible, which is not specifically limited here.
  • a collection detector collects a speckle image formed by the multiple beams of coherent light passing through the object to be measured or the multiple beams of coherent light are reflected by the object to be measured, and determines the speckle image based on the speckle image. Describes whether the glass is defective.
  • the acquisition detector includes one or more photoelectric sensors, wherein the photoelectric sensors are used to acquire speckle images.
  • the positions and numbers of the photoelectric sensors can be set according to actual needs, and are not specifically limited here.
  • the acquisition detector collects the speckles that are transmitted by the multi-beam decoherent light through the object or the multi-beam decoherent light is reflected by the object and is formed in a non-imaging manner. image.
  • the non-imaging method means that it is not necessary to perform a restoration calculation on the 3D glass according to the speckle image to determine whether there is a defect, but to directly determine whether a defect exists based on the speckle image calculation. It can be understood that when the speckle image is formed in a non-imaging form, the defect information is reflected in the speckle image.
  • Resolving and reconstructing the real image of the object such as phase solving, directly discriminates and detects defects on the speckle image, which can effectively reduce the amount of data calculation and increase the speed of recognition.
  • the speckle image is an image formed when light passes through the optically rough surface of the vibrating object or light is reflected by the optically rough surface of the vibrating object.
  • optically rough surfaces or transmission plates with optically rough transmission
  • the wavelets scattered by irregularly distributed surfaces on these surfaces are superimposed on each other.
  • make the reflected light field (or transmitted light field) have a random spatial light intensity distribution and present a granular structure, which is speckle. For example, as shown in FIG. 3, when a bubble appears inside the curved surface of the glass, the speckle distribution on the image collector changes.
  • the light intensity in the area where bubbles are present in the glass will increase significantly.
  • surface scratches, residues, pits, etc. can cause changes in light intensity.
  • these defects are generally in the range of 10 micrometers to 5 millimeters.
  • the detection accuracy can reach more than 5 micrometers, so it is sufficient to check most defects.
  • the shape of glass defects is complex and changeable. These features cannot well represent the defect target. If the speckle image with a glass defect structure is directly subjected to phase inversion to obtain an image restoration image, the restored image will be greatly distorted, and it is almost impossible to identify whether there is a defect in the glass. Therefore, in this application, the speckle image can be input into a neural network to determine whether a defect exists in the glass.
  • the speckle image initially collected by the acquisition detector is a speckle image full of noise.
  • the useful information in the speckle image is drowned in a lot of noise. Therefore, the acquisition detector needs to process the speckle image originally collected to obtain a processed speckle image in order to remove speckle noise and improve fringe contrast.
  • methods for processing an image include a phase shift method, a fringe grayscale method, a fringe centerline method, a Fourier transform method, a sub-pixel search method, and so on. It should be understood that the foregoing processing methods are merely examples, and should not constitute a specific limitation.
  • the speckle image may include a straight speckle image and a curved speckle image, which are not specifically limited herein.
  • the flat speckle image may be one or more.
  • the number of straight speckle images can be one; when the area of the straight portion is relatively small, the number of straight speckle images can be multiple, and different straight speckles The image corresponds to different areas of the straight portion.
  • the number of straight speckle images can be set according to the area of the flat portion of the glass, which is not specifically limited this time.
  • the curved speckle image may be one or more.
  • the number of curved speckle images when the curved portions are concentrated in the same area, the number of curved speckle images may be one; when the curved portions are scattered in multiple regions, the number of curved speckle images may be multiple, and different curved speckle images correspond to Different areas of the bend.
  • the number of curved speckle images can be set according to the area where the curved portion of the glass is scattered, which is not specifically limited this time.
  • an acquisition detector is used to determine whether the glass has a defect according to the speckle image and a deep learning neural network.
  • the input of the deep learning neural network includes a first curved speckle image, a second curved speckle image, a first flat speckle image, and a second flat speckle image.
  • the example of the input of the neural network is merely an example. In actual applications, the speckle image input by the neural network may be more or less, which is not specifically limited herein.
  • the output results of the deep learning neural network may include: no defects, scratches, bubbles, dirt, etc.
  • the output results may also be expressed using fewer or more levels. . It can be understood that the above-mentioned level division is only used as an example, and the more the level division is, the more accurate the output result is represented.
  • the deep learning neural network includes multiple training models.
  • the training model may include a defect-free training model, a bubble model, a dirty model, and so on.
  • the foregoing training model is merely an example, and in actual applications, more or fewer training models may be included, which is not specifically limited herein.
  • the neural network can be BP neural network, Hopfield network, ART network, Kohonen network, Long Short-Term Memory (LSTM), Residual Network (ResNet), Recurrent Neural Networks , RNN), etc., are not specifically limited here.
  • the output when the input includes a first curved speckle image, a second curved speckle image, a first flat speckle image, and a second flat speckle image, the output includes no defects, scratches, and bubbles.
  • the dirty, deep learning neural network can be shown in Figure 4.
  • the deep learning neural network may be trained and learned using a large sample of speckle images. For example, a large number of speckle images are collected according to different defect samples in advance, and the deep neural network is used to classify and train these speckle images that can indirectly reflect the fine structure of glass to obtain the correct neural network. During recognition, the collected speckle images are input to the trained neural network to obtain the recognition results.
  • coherent light is generated by a coherent light source, and then the coherent light is divided into a plurality of beams of coherent light by a beam splitter, and the angle of the beam of coherent light that irradiates the curved portion of the object to be measured is collected by a reflector to collect
  • the detector collects a speckle image formed by the multiple beams of coherent light passing through the object to be measured or the multiple beams of coherent light are reflected by the object to be measured.
  • the acquisition detector determines the speckle image based on the speckle image. Whether there is a defect in the object to be tested. It is not difficult to see that the above solution can quickly detect curved glass.
  • the above detection method can also be applied to other curved glass surfaces, semi-transparent plastic, frosted glass, etc., and even other objects that are not transparent, etc. It is not specifically limited here. The following is a description with reference to several specific embodiments.
  • the detection system of this embodiment is used to implement defect detection on the appearance of curved glass. Place the curved glass in the inspection system, and then adjust the mirror so that the reflected coherent light illuminates the curved portion of the curved glass vertically. Defects such as scratches, dirt, and bubbles inside the curved glass will cause corresponding defects. The speckle image changes and is identified.
  • the detection system of this embodiment is used to implement defect detection of large-format flat glass, such as display screens of tablet computers and liquid crystal displays, and television glass. Place the large-format flat glass in the detection system, and then adjust the angle of the reflector to ensure that coherent light can cover the large-format flat glass.
  • the large-format glass is scratched, dirty, and air bubbles inside the large-format glass. Such defects will cause the corresponding speckle image to change and be identified.
  • the detection system of this embodiment is used to implement defect detection of a complex curved surface having an arbitrary curved surface shape. Change the illumination angle of the mirror of the inspection system and the position and angle distribution of the photoelectric sensor, so as to realize the defect detection of the complex curved surface with any curved surface shape.
  • the detection system of this embodiment is used to implement defect detection of opaque objects. Changing the position of the photoelectric sensor of the inspection system enables the photoelectric sensor to collect coherent light reflected by opaque objects, thereby achieving defect detection of opaque objects.
  • the detection system of this embodiment is used to implement continuous online detection.
  • a plurality of objects to be tested are placed on the electric platform, and the plurality of objects to be tested are dragged by the electric platform through the detection system, and the detection system performs defect detection on the plurality of objects to be tested in turn, thereby achieving continuous online detection.
  • the disclosed system, terminal, and method may be implemented in other ways.
  • the device embodiments described above are only schematic.
  • the division of the unit is only a logical function division.
  • multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may also be electrical, mechanical or other forms of connection.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, may be located in one place, or may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments of the present invention.
  • each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist separately physically, or two or more units may be integrated into one unit.
  • the above integrated unit may be implemented in the form of hardware or in the form of software functional unit.
  • the integrated unit When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium.
  • the technical solution of the present invention is essentially a part that contributes to the existing technology, or all or part of the technical solution may be embodied in the form of a software product, which is stored in a storage medium
  • Included are several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in the embodiments of the present invention.
  • the foregoing storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical disks and other media that can store program codes .

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)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

La présente invention concerne un système et un procédé de détection. Le système comprend : une source de lumière cohérente pour générer de la lumière cohérente ; un séparateur de faisceau pour séparer la lumière cohérente en une pluralité de faisceaux de lumière cohérente séparée, la pluralité de faisceaux de lumière cohérente séparée étant utilisée pour éclairer un objet à détecter ; un miroir réfléchissant pour ajuster les angles de certains ou de l'ensemble des faisceaux de lumière cohérente séparée ; et un détecteur d'acquisition pour acquérir une image tavelée générée par le passage de la pluralité de faisceaux de lumière cohérente séparée à travers l'objet à détecter ou par la réflexion de la pluralité de faisceaux de lumière cohérente séparée par l'objet à détecter, et déterminer, sur la base de l'image tavelée, s'il existe un défaut dans un panneau de verre. La solution peut détecter rapidement un panneau de verre déformé.
PCT/CN2018/099921 2018-08-10 2018-08-10 Procédé et système de détection WO2020029237A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201880067184.XA CN111226110A (zh) 2018-08-10 2018-08-10 检测方法和系统
PCT/CN2018/099921 WO2020029237A1 (fr) 2018-08-10 2018-08-10 Procédé et système de détection

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/099921 WO2020029237A1 (fr) 2018-08-10 2018-08-10 Procédé et système de détection

Publications (1)

Publication Number Publication Date
WO2020029237A1 true WO2020029237A1 (fr) 2020-02-13

Family

ID=69414404

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/099921 WO2020029237A1 (fr) 2018-08-10 2018-08-10 Procédé et système de détection

Country Status (2)

Country Link
CN (1) CN111226110A (fr)
WO (1) WO2020029237A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111650201A (zh) * 2020-07-15 2020-09-11 Oppo（重庆）智能科技有限公司 检测装置及检测方法

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114638467B (zh) * 2022-01-31 2023-04-07 南通鑫鑫医药药材有限公司 一种医用玻璃容器生产质量检测系统
CN114998216A (zh) * 2022-05-06 2022-09-02 湖北文理学院 透明件表面缺陷快速检测方法及装置

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN85101640A (zh) * 1985-04-01 1987-01-17 欧文斯-伊利诺衣公司 透明物体中折光缺陷的检测系统
CN101175986A (zh) * 2005-04-06 2008-05-07 康宁股份有限公司 玻璃检测系统及其使用方法
JP2008170207A (ja) * 2007-01-10 2008-07-24 Olympus Corp 目視検査装置用のマクロ照明装置、目視検査装置及びガラス基板の目視検査方法
CN104482875A (zh) * 2014-12-19 2015-04-01 合肥工业大学 单狭缝空间载波剪切散斑干涉测量系统及测量方法

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7679059B2 (en) * 2006-04-19 2010-03-16 Spectrasensors, Inc. Measuring water vapor in hydrocarbons
CN101221133A (zh) * 2008-01-25 2008-07-16 天津商业大学 蜂窝纸板粘接缺陷检测方法及检测系统
CN201184867Y (zh) * 2008-04-09 2009-01-21 苏州大学 基于4f相位相干成像系统测量材料的光学非线性的装置
CN201555808U (zh) * 2009-11-19 2010-08-18 西北工业大学 一种无损检测平面物品表面的装置
CN201876417U (zh) * 2010-11-05 2011-06-22 深圳清华大学研究院 一种免疫比浊测定装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN85101640A (zh) * 1985-04-01 1987-01-17 欧文斯-伊利诺衣公司 透明物体中折光缺陷的检测系统
CN101175986A (zh) * 2005-04-06 2008-05-07 康宁股份有限公司 玻璃检测系统及其使用方法
JP2008170207A (ja) * 2007-01-10 2008-07-24 Olympus Corp 目視検査装置用のマクロ照明装置、目視検査装置及びガラス基板の目視検査方法
CN104482875A (zh) * 2014-12-19 2015-04-01 合肥工业大学 单狭缝空间载波剪切散斑干涉测量系统及测量方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111650201A (zh) * 2020-07-15 2020-09-11 Oppo（重庆）智能科技有限公司 检测装置及检测方法
CN111650201B (zh) * 2020-07-15 2023-06-06 Oppo（重庆）智能科技有限公司 检测装置及检测方法

Also Published As

Publication number Publication date
CN111226110A (zh) 2020-06-02

Similar Documents

Publication Publication Date Title
US7471383B2 (en) Method of automated quantitative analysis of distortion in shaped vehicle glass by reflected optical imaging
TWI664586B (zh) 透過瑕疵複數光場之透明基板瑕疵檢測方法
US9891168B2 (en) Device and method for sensing at least one partially specular surface with column-by-column analysis of the global or local intensity maximum
WO2020029237A1 (fr) Procédé et système de détection
BR102016028266A2 (pt) Método e sistema para a inspeção automática de qualidade de materiais
CN108332689A (zh) 一种检测表面粗糙度和表面损伤的光学测量系统及方法
FR2869408A1 (fr) Procede et dispositif pour determiner la forme et les normales locales de surfaces reflechissantes
CN110057841A (zh) 一种基于透射结构光的缺陷检测方法
CN111344559B (zh) 缺陷检测方法及缺陷检测系统
CN110108223B (zh) 散斑干涉与剪切散斑干涉的测量系统及测量方法
CN107561007B (zh) 一种薄膜测量装置和方法
CA2703360A1 (fr) Methode et systeme d'inspection avec analyse des dephasages a images multiples
TWI394946B (zh) Method and device for measuring object defect
CN115950890B (zh) 用于工业检测的谱域光学相干层析成像检测方法及系统
CN111579564B (zh) 一种透明平板缺陷和屈光度变化检测系统及方法
EP1605241A1 (fr) Appareil pour le controle des pièces transparentes ou réflechissantes
JP2000146554A (ja) 透明板の表面凹凸検査方法および装置
WO2020051779A1 (fr) Procédé de détection de défauts et système de détection pour objet à surface incurvée
US10598604B1 (en) Normal incidence phase-shifted deflectometry sensor, system, and method for inspecting a surface of a specimen
CN111316086B (zh) 表面缺陷光学检测方法及相关装置
EP0756152B1 (fr) Procédé et dispositif de détection de l'état de surface de pièces à surface réfléchissante, applicable au contrôle de rugosité de pièces polies
JP5557586B2 (ja) 表面性状測定装置および表面性状測定方法
JP6196841B2 (ja) 透過波面計測装置及び透過波面計測方法
CN209327240U (zh) 一种球面光学元件表面疵病检测装置
FR3018603A1 (fr) Dispositif d'analyse et d'imagerie d'une surface par deflectometrie et procede correspondant

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: 18929591

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 15.06.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18929591

Country of ref document: EP

Kind code of ref document: A1