WO2020051779A1 - 曲面物体的缺陷检测方法及检测系统 - Google Patents

曲面物体的缺陷检测方法及检测系统 Download PDF

Info

Publication number
WO2020051779A1
WO2020051779A1 PCT/CN2018/105110 CN2018105110W WO2020051779A1 WO 2020051779 A1 WO2020051779 A1 WO 2020051779A1 CN 2018105110 W CN2018105110 W CN 2018105110W WO 2020051779 A1 WO2020051779 A1 WO 2020051779A1
Authority
WO
WIPO (PCT)
Prior art keywords
defect
coherent
curved object
speckle image
light source
Prior art date
Application number
PCT/CN2018/105110
Other languages
English (en)
French (fr)
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 PCT/CN2018/105110 priority Critical patent/WO2020051779A1/zh
Priority to CN201880001802.0A priority patent/CN111344553B/zh
Publication of WO2020051779A1 publication Critical patent/WO2020051779A1/zh

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

Definitions

  • the invention relates to the technical field of defect detection, in particular to a defect detection method and a detection system for curved objects.
  • the surface area of highly reflective curved objects can be divided into specular reflection areas and diffuse reflection areas. Because the light intensity of the specular reflection areas exceeds the sensing range of the camera, the image captured by the camera will form gray saturation, thereby losing detailed information on the surface of the object. .
  • the polarized light method is mainly used to reduce the intensity of the specular reflection component and construct a polysaturated area for measurement. However, this method makes the diffuse reflection area with a lower light intensity on the surface of the metal workpiece darker, which makes the diffuse reflection area due to gray Too low to distinguish defects.
  • a defect detection method for curved objects includes the following steps:
  • the step S1) includes controlling the light source to generate a coherent laser, and dividing the coherent laser into a plurality of coherent lasers by a beam splitter.
  • the step S2) includes: connecting the multiple coherent lasers to a plurality of beam expanders with adjustable angles to control the multiple coherent lasers to the curved object from multiple irradiation angles.
  • the detection area is illuminated.
  • the multiple coherent lasers are four coherent lasers.
  • the photoelectric sensor is a CCD sensor or a CMOS sensor.
  • the step S3) further includes adjusting the speckle image using an adaptive speckle image adjustment algorithm to make the grayscale distribution of the speckle image uniform.
  • the adaptive speckle image adjustment algorithm includes the following steps:
  • the related parameters include at least one of an exposure time of the photosensor, a gain value of the photosensor, a light source brightness, a light source splitting ratio, and an irradiation angle.
  • the step S4) includes: analyzing the speckle image by a deep learning method of a neural network to determine whether a defect exists on the surface of the curved object and the type of the defect.
  • a defect detection system for curved objects includes:
  • a light source unit configured to generate a plurality of coherent lasers and control the plurality of coherent lasers to illuminate a detection area of a curved object from a plurality of irradiation angles;
  • a photoelectric sensor for receiving a coherent laser signal reflected or scattered by the curved object and generating a speckle image
  • a detection and judgment unit that is connected to the photoelectric sensor to receive the speckle image and is configured to determine whether the curved object has a defect and a type of the defect according to the speckle image, wherein the light source unit
  • the coherent laser light source includes a coherent laser light source, a beam splitter, and a plurality of beam expanders with adjustable angles.
  • the coherent laser light source is used to generate a coherent laser, and the coherent laser is passed through the beam splitter to generate a plurality of coherent lasers.
  • the plurality of beam expanders are configured to receive the plurality of coherent lasers and control the plurality of coherent lasers to illuminate a detection area of a curved object from a plurality of irradiation angles.
  • multiple coherent lasers such as four coherent lasers, are used to illuminate the detection areas of curved objects in four directions, and the surface illumination areas overlap.
  • the laser receives a coherent laser signal reflected or scattered by the curved object through a photoelectric sensor and generates a speckle image. Due to defects on the surface of the measured curved object, such as scratches, cracks, uneven deformation, surface Dirty, etc. will change the phase information of the coherent light source. After imaging, these defects and bad information will be modulated into the speckle image. Different types and different sizes of defects will change the distribution of light and dark spots in the speckle image.
  • Image can effectively detect defects; At the same time, surface defect detection using multi-angle coherent laser speckle imaging has great advantages, its optical structure is simple, and it can effectively solve the problem of surface depth of field and illumination uniformity, detection accuracy High, for a variety of highly reflective surfaces in industrial inspection The appearance of quality defect detection provides a very simple and practical and feasible solutions.
  • FIG. 1 is a flowchart of a defect detection method for a curved object according to an embodiment of the present invention
  • FIG. 2 is a technical principle diagram of an adaptive speckle image adjustment algorithm for a defect detection method of a curved object according to an embodiment of the present invention
  • FIG. 3 is a schematic diagram of a principle of artificial neural network deep learning to determine a defect in a method for detecting a defect of a curved object according to an embodiment of the present invention.
  • FIG. 4 is a schematic diagram of a defect detection system for a curved object according to an embodiment of the present invention.
  • the present invention provides a defect detection method for curved objects, which includes the following steps:
  • the inventors of the present invention have discovered through research that when detecting surface defects of curved objects, strong reflective bright spots appearing on highly reflective curved surfaces easily cause image saturation, and the imaging of reflecting surrounding objects overwhelms the defect information, so the influence of the light source on the detection results It is very large.
  • the illumination light source needs to be uniformly irradiated to various locations on the curved surface in the form of scattered light to avoid the impact of uneven illumination on subsequent recognition.
  • multiple beams of coherent laser light are used to illuminate the detection area of a curved object from multiple irradiation angles.
  • Speckle images irradiated from multiple angles can obtain speckle images with uniform and appropriate brightness, and further Defect detection on the surface of highly reflective curved objects.
  • the present invention uses multi-angle coherent laser speckle interference imaging to perform defect detection on the surface of a highly reflective curved object.
  • the surface defect detection using multi-angle coherent laser speckle imaging has great advantages, its optical structure is simple, and it can effectively solve the problem.
  • the problem of surface depth of field and illumination uniformity, high detection accuracy provides a very simple and practical solution for the appearance defect detection of a variety of highly reflective surface materials in industrial inspection.
  • the step S1) includes: controlling the light source to generate a coherent laser, and dividing the coherent laser into a plurality of coherent lasers by a beam splitter, that is, in this embodiment, the beam splitter
  • the coherent laser is divided into a plurality of coherent lasers.
  • the step S2) includes: connecting the multiple coherent lasers to multiple beam expanders with adjustable angles to control the multiple coherent lasers to a curved object from multiple irradiation angles
  • the detection area is illuminated. That is, in this embodiment, after generating a plurality of coherent laser beams, an irradiation angle of each coherent laser beam is adjusted by a beam expander to realize illumination of multiple irradiation angles on a detection area of a curved object.
  • the structure and working principle of the beam expander are well known to those skilled in the art, which will not be described in detail in the present invention.
  • various methods can be used to connect the coherent laser to the beam expander, which is not particularly limited in the present invention. For example, each coherent laser can be connected to the beam expander through an optical fiber.
  • the invention uses multiple beams of coherent laser light to illuminate the detection area of a curved object from different irradiation angles, the illumination is more uniform, and the high light effect of the curved object can be effectively eliminated, and the detection accuracy can be improved.
  • multi-beam in the present invention should be understood in a broad sense, and it can be two beams, three beams, or more, as long as the coherent laser beams perform detection on curved surface objects from different irradiation angles. Lighting, so that the light is even.
  • the present invention has no special limitation on the irradiation angle of each coherent laser beam, and it can be based on the surface of the curved object to be detected. To adjust the actual situation, as long as the illumination light source can be uniformly irradiated to each position on the curved surface, and the influence of uneven illumination on subsequent recognition can be avoided.
  • four beams of coherent laser light are used to illuminate the detection area of the curved object, that is, four beams of coherent laser light are used to illuminate the detection area of the curved object from four directions, and the surface illumination areas may overlap.
  • Defects of curved objects are mainly distributed on the surface of the product.
  • the size of these defects is generally different from more than ten microns to more than 1 mm.
  • Coherent lasers with different wavelengths can detect the deformation of object surfaces with different accuracy. For example, for a wavelength of For a 650nm coherent laser, the accuracy of surface deformations that can be detected can reach 1um.
  • the invention has no special limitation on the wavelength of the coherent laser, and it can be adjusted according to actual needs to meet different detection accuracy requirements.
  • a coherent laser When a coherent laser is irradiated on the surface of a curved object, there is scattered light at each point on the surface. These scattered lights are coherent light, but their amplitude and phase are different and they are randomly distributed. After the scattered light is superimposed, a photoelectric sensor (such as a CCD sensor or a CMOS sensor) can form a granular structure with better contrast, that is, speckle.
  • a plurality of coherent laser beams irradiate the detected area from different angles. These scattered lights do not pass through a conventional lens, but directly enter a photoelectric sensor for A / D digital conversion, thereby obtaining a speckle image.
  • the detection method provided by the present invention uses non-imaging speckle image detection, and the scattered light reflected from the measured object is directly projected onto the photoelectric sensor.
  • This non-traditional lens imaging method can avoid the need to change the distance of the measured object.
  • Adjusted focal plane imaging is clear, especially suitable for object detection of curved structure.
  • multi-angle coherent laser light is used to illuminate curved objects at multiple angles.
  • Direct lens imaging is used to obtain speckle images, which can avoid image blurring caused by the depth of field of 3D surface imaging with different curvature changes, and then obtain clear speckle images. .
  • the photoelectric sensor may be various commonly used photoelectric sensors in the art, and the present invention has no special restrictions on this.
  • the photoelectric sensor is a CCD sensor or a CMOS sensor.
  • the structure and working principle of a CCD sensor or a CMOS sensor are well known to those skilled in the art, which will not be described in detail in the present invention.
  • step S3) further includes adjusting the speckle image using an adaptive speckle image adjustment algorithm to make the grayscale distribution of the speckle image uniform.
  • the adaptive speckle image adjustment algorithm may include the following steps: S31) analyze the grayscale histogram distribution of the speckle image to calculate the grayscale distribution of the speckle image; and S32) when overexposure occurs After the number of gray levels exceeds the threshold, the related parameters are adjusted to obtain a uniform, non-overexposed speckle image.
  • the related parameters may be various commonly used parameters, such as at least one of an exposure time of a photoelectric sensor, a gain value of the photoelectric sensor, a light source brightness, a light source spectral ratio, and an irradiation angle.
  • using an adaptive speckle image adjustment algorithm can effectively ensure that highly reflective surfaces with different reflection coefficients and different curvatures can form a uniformly distributed speckle image in grayscale, avoiding overexposure of the image caused by high exposure in some areas The problem of missing surface defects of the object under test occurs.
  • the gray histogram is to count all the pixels in the image according to the size of the gray value, and count the frequency of their appearance. It is a function of the gray level distribution in the image and a statistics of the gray level distribution in the image. It can be understood that various methods can be used to analyze the grayscale histogram distribution of the speckle image to calculate the grayscale distribution of the speckle image. The present invention has no special restrictions on this. For example, in some implementations of the present invention, In the example, the gray histogram distribution of the speckle image is analyzed using an automatic exposure algorithm, and the gray distribution of the speckle image is calculated.
  • a speckle image is obtained by a photoelectric sensor (for example, a CCD sensor or a CMOS sensor)
  • a photoelectric sensor for example, a CCD sensor or a CMOS sensor
  • the gray histogram distribution of the speckle image is analyzed by an automatic exposure algorithm to statistically calculate the speckle.
  • the gray distribution of the spot image when the number of over-exposed gray levels exceeds the threshold, feedback adjustments are made on the exposure time of the photosensor, the gain value of the photosensor, the light source brightness, the light source splitting ratio, and the irradiation angle, so as to obtain Even and overexposed speckle image.
  • the principle of the automatic exposure algorithm is well known to those skilled in the art, which will not be described in detail in the present invention.
  • the method for adjusting the exposure time and gain value of the photoelectric sensor can adopt various existing adjustment methods. There are no special restrictions. Of course, you can also use computer software to automatically adjust parameters such as the exposure time of the photoelectric sensor, the gain value of the photoelectric sensor, the light source brightness, the light source splitting ratio, and the irradiation angle, for example, the photoelectric sensor, light source, beam splitter, and beam expander.
  • the mirror is connected to a computer.
  • the computer software can control the photoelectric sensor, light source, beam splitter, and beam expander to automatically Adjust the exposure time of the photoelectric sensor, the gain value of the photoelectric sensor, the light source brightness, the light source splitting ratio, and the irradiation angle and other parameters.
  • adaptive brightness and angle adjustment is performed according to the collected speckle image to achieve uniform distribution of the speckle image in the measured area, reduce the brightness of the highly reflective area, and adjust the image gray level in the shadow area.
  • the speckle image with defect information can be effectively analyzed to determine whether there is a defect and the type of the defect.
  • Defects on the surface of the measured curved object such as scratches, cracks, bump deformation, surface dirt, etc., will change the phase information of the coherent light source. After imaging, these defect bad information will be modulated into speckle images, different types and different sizes. Defects in the speckle image will change the distribution of the light and dark spots in the speckle image. Furthermore, by analyzing the speckle image, it is possible to determine whether the curved object has a defect and the type of the defect.
  • the speckle image is analyzed by a deep learning method of a neural network to determine whether there is a defect on the surface of the measured curved object and the type of the defect.
  • deep learning artificial neural network is used for large sample speckle data training and learning. More than 100 images are collected according to different defect samples, and the speckles that can indirectly reflect the surface microstructure are classified and trained to obtain the measured A neural network model of speckle images on curved object surfaces. Through this neural network model, a variety of different output states can be defined, such as: defect-free (OK), dirt, scratches, deformation, etc., and further, speckle images are generated. Then, the neural network model can be used to determine whether the curved object has a defect and the type of the defect according to the speckle image.
  • the artificial neural network deep training and learning method generally includes an input layer, a hidden layer, and an output layer.
  • the present invention first, collect hundreds or more speckle images of curved objects without defects on the surface, and train these speckle images through a deep learning neural network to obtain defect-free nerves.
  • the neural network model can be used to obtain a neural network model with no defects on the surface or various defects on the surface. Furthermore, when determining whether there is a defect on the surface of the surface to be tested and the type of the defect, the input layer input of the neural network learning The speckle image of the surface object to be measured is compared with various neural network models of the hidden layer and calculated.
  • the speckle image corresponding to an output state such as: a defect-free surface (the OK), the presence of surface dirt, surface scratches present, the presence of surface modification and the like.
  • the present invention has no special restrictions on the method for deep training and learning of artificial neural networks, and it can be various commonly used methods for deep training and learning of artificial neural networks.
  • Deep neural network is used to directly modulate surface defects to coherent light speckle interference patterns for defect learning and detection. Compared with traditional image preprocessing and adaptive segmentation detection methods, this method is more adaptable, simple to use, and not Requires complex light source design and parameter settings.
  • the detection method provided by the present invention can be applied to different occasions. For example, it can be applied to the detection of surface defects on complex free-form surfaces, and the color caused by chipping, deformation, and foreign pollution on the surface will cause the corresponding speckle pattern to change and be identified; it can also be applied to highly reflective planes and 3D surfaces
  • the speckle signal on the surface of a normal object, such as glass and transparent products is also different from the speckle signal caused by various defects, so it can also be identified by the detection method provided by the present invention.
  • the detection method provided by the present invention can also be applied to some specific materials, such as translucent plastic materials, and defects on the surface of objects mixed with different materials, such as shallow bubbles and pores, which interfere with the coherent laser. At this time, these lights will penetrate into the product, and thus, the internal information of the measured object will also be displayed in the speckle image, so as to be identified.
  • the illumination light path of the detection method provided by the present invention can be extended to one angle, multiple angles, and adjustable angles.
  • One photoelectric sensor or multiple photoelectric sensors can be used to obtain speckle images for identification, and can also be adapted. Spectral images taken with a lens.
  • the illumination optical design and the photoelectric sensor used in the detection method provided by the present invention can be variously modified to adapt to different application occasions and achieve different detection resolution requirements.
  • the detection method provided by the present invention can perform complex structure and large-scale surface detection on the measured curved object.
  • the measured object can be moved and the position of the optical system can be changed to realize the complex structure and large-scale surface detection of the measured curved object.
  • the position of the optical vision system is fixed, and the way of measuring the object is changed by moving to realize the complex structure and large-scale surface detection of the measured curved object.
  • the present invention also provides a defect detection system for curved objects, which includes a light source unit, a photoelectric sensor 2, and a detection and judgment unit 3.
  • the light source unit is used to generate multiple coherent lasers and control the multiple coherent lasers to illuminate the detection area of the curved object 4 from multiple irradiation angles;
  • the photoelectric sensor 2 is used to receive the coherent laser light reflected or scattered by the curved object 4 The signal is generated and a speckle image 5 is generated;
  • the detection and judgment unit 3 is connected to the photoelectric sensor 2 to receive the speckle image 5, and is configured to determine whether the curved object 4 has a defect and a type of the defect according to the speckle image 5.
  • the light source unit includes a coherent laser light source 11, a beam splitter 12, and a plurality of beam expanders 13 with adjustable angles.
  • the coherent laser light source 11 is used to generate a coherent laser
  • the coherent laser is
  • the beam splitter 12 is used to generate multiple coherent lasers.
  • the multiple beam expanders 12 are used to receive the multiple coherent lasers and control the multiple coherent lasers to detect the curved object 4 from multiple irradiation angles. The area is illuminated.
  • the beam splitter 12 and the beam expander 13 are well known to those skilled in the art, and the present invention will not repeat them here. It can be understood that various methods can be used to connect the coherent laser beams to the beam expander mirror 13, which is not particularly limited in the present invention.
  • the coherent laser beams can be connected to the beam expander mirror 13 through the optical fiber 14.
  • the detection and judgment unit 3 may be various commonly used detection and judgment units, such as a computer.
  • the computer obtains an accurate neural network model through a deep learning method of a neural network, and then compares and judges the current speckle image through the neural network model to determine that the speckle image corresponds to Whether there is a defect on the surface of the curved object 4.
  • multiple coherent lasers such as four coherent lasers, are used to illuminate the detection areas of curved objects in four directions, and the surface illumination areas overlap.
  • the laser receives a coherent laser signal reflected or scattered by the curved object through a photoelectric sensor and generates a speckle image. Due to defects on the surface of the measured curved object, such as scratches, cracks, uneven deformation, surface Dirty, etc. will change the phase information of the coherent light source. After imaging, these defects and bad information will be modulated into the speckle image. Different types and different sizes of defects will change the distribution of light and dark spots in the speckle image.
  • Image can realize effective detection of defects.
  • the use of multi-angle coherent laser speckle imaging for surface defect detection has great advantages. Its optical structure is simple, and it can effectively solve the problem of surface depth of field and illumination uniformity. The detection accuracy is high, and it is highly reflective for industrial inspection. The appearance defect detection of curved materials provides a very simple and practical solution.

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

一种曲面物体(4)的缺陷检测方法及检测系统，其包括以下步骤：S1)控制光源(11)生成多束相干激光；S2)控制多束相干激光从多个照射角度对曲面物体(4)的检测区域进行照明；S3)通过光电传感器(2)接收曲面物体(4)反射或散射的相干激光信号并生成散斑图像(5)；以及S4)根据散斑图像(5)判断曲面物体(4)是否存在缺陷及缺陷的类型。该检测方法及检测系统能有效解决曲面景深问题和光照均匀性问题，有效识别曲面物体(4)表面的缺陷，提高检测精度。

Description

曲面物体的缺陷检测方法及检测系统 技术领域
本发明涉及缺陷检测技术领域，特别涉及一种曲面物体的缺陷检测方法及检测系统。
 背景技术
随着制造技术不断提高和加工工艺的改进，市场竞争对产品的质量提出了更高的要求，尤其是工件表面的微小缺陷。对于平面物体的表面缺陷检测，相关技术已经具有很好的检测效果，但是目前对于具有低纹理高反光的曲面物体来说，表面缺陷的检测效果不佳。
对于曲面物体，比如BGA锡球，高亮金属球，手机金属外壳等，其具备如下两种特点：一、表面纹理特征单一，甚至缺失；二、表面光滑造成在进行照明时具有极强的反光特性，从而容易产生过亮光斑。上述两个特点的存在使得其在其生产及后期处理过程中，可能会产生划痕、擦伤等表面缺陷，严重影响产品的使用性能和寿命。
高反光曲面物体的表面区域可以分为镜面反射区域和漫反射区域，镜面反射区域由于其光强超过相机的感应范围，使得相机拍摄的图像上会形成灰度饱和，从而丢失物体表面的细节信息。目前主要通过偏振光法减少镜面反射分量光强，构建多饱和区域进行测量，然而这种方法使金属工件表面原来光强较低的漫反射区域变得更加黑暗，从而使得漫反射区域由于灰度太低而无法区分出缺陷。
现有技术中有另一种采用基于主动结构光投影的三维重构方法对高反光曲面物体的表面区域进行缺陷检测，但是，由于大面积耀光会影响光栅条纹的提取，从而导致无法获得准确深度信息，会出现大面积的数据空洞。因此，现有技术中常用的光学三维扫描和二维成像方法都很难对高反光金属表面进行缺陷检测。当下的检测手段大多采用传统的人工目视灯检方式，但是该检测手段对于缺陷的识别有效性不足80%，而且检测成本极高。
因此如何在排除高光影响的基础上有效地识别缺陷目标，对曲面物体表面的缺陷的识别具有重要意义。
 技术问题
基于此，为解决现有技术中由于存在高光影响，无法有效的识别曲面物体表面的缺陷的技术问题，特提出了一种曲面物体的缺陷检测方法。
 技术解决方案
一种曲面物体的缺陷的检测方法，其包括以下步骤：
S1）控制光源生成多束相干激光；
S2）控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明；
S3）通过光电传感器接收所述曲面物体反射或散射的相干激光信号并生成散斑图像；以及
S4）根据所述散斑图像判断所述曲面物体是否存在缺陷及缺陷的类型。
在其中的一个实施例中，所述步骤S1）包括：控制光源生成相干激光，通过分束器将所述相干激光分成多束相干激光。
在其中的一个实施例中，所述步骤S2）包括：将所述多束相干激光连接到角度可调节的多个扩束镜以控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明。
在其中的一个实施例中，所述多束相干激光为四束相干激光。
在其中的一个实施例中，所述光电传感器为CCD传感器或CMOS传感器。
在其中的一个实施例中，所述步骤S3）中还包括采用自适应的散斑图像调节算法对所述散斑图像进行调节以使所述散斑图像的灰度分布均匀。
在其中的一个实施例中，所述自适应的散斑图像调节算法包括以下步骤：
S31）分析所述散斑图像的灰度直方图分布，以统计出所述散斑图像的灰度分布；以及
S32）当出现过曝的灰度数量超过阈值后，通过调节相关参数，以获得均匀无过曝的散斑图像。
在其中的一个实施例中，所述相关参数包括光电传感器的曝光时间、光电传感器的增益值、光源亮度、光源分光比及照射角度中的至少一种。
在其中的一个实施例中，所述步骤S4）包括：通过神经网络深度学习的方法对所述散斑图像进行分析以判断所述曲面物体的表面是否存在缺陷以及缺陷的类型。
 
此外，为解决现有技术中由于存在高光影响，无法有效的识别曲面物体表面的缺陷的技术问题，特提出了一种曲面物体的缺陷检测系统。
一种曲面物体的缺陷的检测系统，其包括：
光源单元，用于生成多束相干激光并控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明；
光电传感器，用于接收所述曲面物体反射或散射的相干激光信号并生成散斑图像；以及
检测判断单元，所述检测判断单元与所述光电传感器连接以接收所述散斑图像，并用于根据所述散斑图像判断所述曲面物体是否存在缺陷及缺陷的类型，其中，所述光源单元包括相干激光光源，分束器及角度可调节的多个扩束镜，其中，相干激光光源用于生成相干激光，并将所述相干激光通过所述分束器以生成多束相干激光，所述多个扩束镜用于接收所述多束相干激光并控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明。
 有益效果
实施本发明实施例，将具有如下有益效果：
根据本发明提供的曲面物体的缺陷的检测方法及检测系统，其采用多束相干激光，例如四束相干激光，在四个方位对曲面物体的检测区域进行照明，表面照明区域有重叠，这些相干激光经过曲面物体的多角度反射后，通过光电传感器接收所述曲面物体反射或散射的相干激光信号并生成散斑图像，由于被测曲面物体表面的缺陷，例如划伤、裂痕、凹凸变形、表面脏污等都会改变相干光源的相位信息，经成像后，这些缺陷不良信息会调制到散斑图像，不同类型和不同大小的缺陷会改变散斑图像中的明暗光斑的分布，进而通过分析散斑图像，可以实现对缺陷的有效检测；同时，采用多角度相干激光散斑成像的表面缺陷检测具有非常大的优势，其光学结构简单，而且能有效解决曲面景深问题和光照均匀性问题，检测精度高，对于工业检测中多种高反光曲面材质的外观缺陷检测提供了一种非常简单实用可行的解决方案。
 附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
其中：
图1为本发明一个实施例提供的曲面物体的缺陷的检测方法的流程图；
图2为本发明一个实施例提供的曲面物体的缺陷的检测方法的自适应的散斑图像调节算法的技术原理图；
图3为本发明一个实施例提供的曲面物体的缺陷的检测方法的人工神经网络深度学习判断缺陷的原理示意图；以及
图4为本发明一个实施例提供的曲面物体的缺陷的检测系统的示意图。
 本发明的最佳实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
为解决现有技术中由于存在高光影响，无法有效的识别曲面物体表面的缺陷的技术问题，特提出了一种曲面物体的缺陷检测方法及检测系统。
参照图1所示，本发明提供一种曲面物体的缺陷的检测方法，其包括以下步骤：
S1）控制光源生成多束相干激光；
S2）控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明；
S3）通过光电传感器接收所述曲面物体反射或散射的相干激光信号并生成散斑图像；以及
S4）根据所述散斑图像判断所述曲面物体是否存在缺陷及缺陷的类型。
本发明的发明人通过研究发现：在检测曲面物体的表面缺陷时，高反光曲面出现的强反光亮斑容易引起图像饱和，而且反射周围物体的成像淹没了缺陷信息，因此光源对检测结果的影响很大，为了使得拍摄得到的图像能显著分割出缺陷区域与正常区域，需要照明光源均匀地以散射光形式照射到曲面各个位置，避免不均匀光照对后续识别的影响。
因此，本发明提供的检测方法中，采用多束相干激光从多个照射角度对曲面物体的检测区域进行照明，多个角度照射的散斑都能以均匀合适的亮度得到散斑图像，进而对高反光曲面物体的表面进行缺陷检测。即本发明通过多角度相干激光散斑干涉成像对高反光曲面物体的表面进行缺陷检测，采用多角度相干激光散斑成像的表面缺陷检测具有非常大的优势，其光学结构简单，而且能有效解决曲面景深问题和光照均匀性问题，检测精度高，对于工业检测中多种高反光曲面材质的外观缺陷检测提供了一种非常简单实用可行的解决方案。
可以理解的是，可以采用各种方法获得多束相干激光，本发明对此没有特殊限制。在本发明的一个实施例中，所述步骤S1）包括：控制光源生成相干激光，通过分束器将所述相干激光分成多束相干激光，即，在此实施例中，通过分束器将相干激光分成多束相干激光。可以理解的是，分束器的结构及工作原理为本领域技术人员所公知，本发明对此不再进行赘述。
同样的，可以采用各种方法将多束相干激光以不同的角度对曲面物体的检测区域进行照明，本发明对此没有特殊限制。在本发明的一个实施例中，所述步骤S2）包括：将所述多束相干激光连接到角度可调节的多个扩束镜以控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明。即在此实施例中，在产生多束相干激光后，通过扩束镜调节各相干激光的照射角度，实现对曲面物体的检测区域的多个照射角度的照明。对于扩束镜的结构及工作原理为本领域技术人员所公知，本发明对此不再进行赘述。同样的，可以采用各种方法将相干激光连接到扩束镜，本发明对此没有特殊限定，例如，可以通过光纤将各相干激光连接到扩束镜。
本发明通过采用多束相干激光从不同的照射角度对曲面物体的检测区域进行照明，光照更加均匀，进而能有效排除曲面物体的高光影响，提高检测精度。可以理解的是，本发明的术语“多束”应做广义理解，其可以为两束，也可以为三束及更多，只要各束相干激光从不同的照射角度对曲面物体的检测区域进行照明，使得光照均匀即可。
同样可以理解的是，由于不同的曲面物体的表面的曲度、外形及面积不尽相同，因此，本发明对于各束相干激光的照射角度没有特殊限制，其可以根据待检测的曲面物体的表面的实际情况进行调整，只要能使得照明光源均匀地照射到曲面各个位置，避免不均匀光照对后续识别的影响即可。
在本发明的一个实施例中，采用四束相干激光对曲面物体的检测区域进行照明，即采用四束相干激光从四个方位对曲面物体的检测区域进行照明，表面照明区域可以有重叠。
曲面物体的缺陷主要分布于产品表面，这些缺陷大小一般都是从十几微米到1毫米以上的不同分布，不同波长的相干激光，其能检测的物体表面变形的精度不同，例如，对于波长为650nm的相干激光而言，能检测的物体表面变形精度可以达到1um。本发明对于相干激光的波长没有特殊限制，其可以根据实际需求进行调整，以适应不同检测精度需求。
当相干激光照射在曲面物体表面时，其表面上的每一点都有散射光，这些散射光是相干光，仅仅是其振幅与位相不同，而且随机分布。这些散射光经过叠加后，在光电传感器（例如CCD传感器或CMOS传感器）可以形成对比度较好的颗粒状结构，即散斑。在本发明中，多束相干激光从不同角度对被检测区域进行照射，这些散射光没有经过传统的透镜，而是直接进入到光电传感器进行A/D数字转换，进而得到散斑图像。
在很多行业中，都需要对高反光曲面物体的表面进行缺陷检测，传统的基于可见光或者结构光投影的方法都无法解决光曝和三维曲面清晰成像的问题。本发明提供的检测方法采用的是非成像的散斑图像检测，从被测物体反射回来的散射光直接投射到光电传感器上，这种非传统透镜成像的方法能够避免由于被测物体距离改变而需要调节焦平面成像清晰，特别适合曲面结构的物体检测。同时，通过多束相干激光对曲面物体进行多角度照明，采用非透镜的直接成像方法获取散斑图像，可以避免不同曲率变化的三维表面成像景深造成的图像模糊问题，进而得到清晰的散斑图像。
可以理解的是，光电传感器可以为本领域各种常用的光电传感器，本发明对此没有特殊限制，例如，在本发明的一些实施例中，光电传感器为CCD传感器或CMOS传感器。对于CCD传感器或CMOS传感器的结构及工作原理为本领域技术人员所公知，本发明对此不再进行赘述。
进一步的，由于被测曲面物体曲率分布不同，多方向的相干照射可能在某个方向传感器过曝而出现饱和现象，因此，为获得灰度分布更加均匀的散斑图像，在本发明的一个实施例中，步骤S3）中还包括采用自适应的散斑图像调节算法对所述散斑图像进行调节以使所述散斑图像的灰度分布均匀。
例如，自适应的散斑图像调节算法可以包括以下步骤：S31）分析所述散斑图像的灰度直方图分布，以统计出所述散斑图像的灰度分布；以及S32）当出现过曝的灰度数量超过阈值后，通过调节相关参数，以获得均匀无过曝的散斑图像。可以理解的，所述相关参数可以为各种常用的参数，例如光电传感器的曝光时间、光电传感器的增益值、光源亮度、光源分光比及照射角度中的至少一种。
在此实施例中，采用自适应的散斑图像调节算法能有效保证对于不同反射系数和不同曲率的高反射表面都能成灰度均匀分布的散斑图像，避免部分区域的高曝造成图像过亮而丢失被测物体的表面缺陷的问题发生。
灰度直方图是将图像中的所有像素，按照灰度值的大小，统计其出现的频率，其是图像中关于灰度级分布的函数，是对图像中灰度级分布的统计。可以理解的是，可以采用各种方法分析散斑图像的灰度直方图分布，以统计出所述散斑图像的灰度分布，本发明对此没有特殊限制，例如，在本发明的一些实施例中，采用自动曝光算法分析散斑图像的灰度直方图分布，统计出散斑图像的灰度分布。
即，如图2所示，在此实施例中，在光电传感器（例如CCD传感器或CMOS传感器）获得散斑图像后，首先通过自动曝光算法分析散斑图像的灰度直方图分布，统计出散斑图像的灰度分布；当出现过曝的灰度数量超过阈值后，对光电传感器的曝光时间、光电传感器的增益值、光源亮度、光源分光比及照射角度等参数进行反馈调节，进而可以获得均匀无过曝的散斑图像。自动曝光算法的原理为本领域技术人员所公知，本发明对此不再进行赘述。
可以理解的是光电传感器的曝光时间及增益值的调节方法，光源亮度的调节方法，光源分光比的调节方法以及光源的照射角度的调节方法可以采用现有的各种调节方法，本发明对此没有特殊限制。当然，还可以通过计算机软件自动对光电传感器的曝光时间、光电传感器的增益值、光源亮度、光源分光比及照射角度等参数进行反馈调节，例如，将光电传感器、光源、分光镜、及扩束镜与计算机连接，在计算机软件统计出散斑图像的灰度分布后，判断出现过曝的灰度数量超过阈值后，计算机软件可以控制光电传感器、光源、分光镜、及扩束镜，以便自动调节光电传感器的曝光时间、光电传感器的增益值、光源亮度、光源分光比及照射角度等参数。
进而，在此实施例中，根据采集到的散斑图像进行自适应的亮度和角度调节，实现被测区域的散斑图像均匀分布，减少高反射区域的亮度，调节阴影区域的图像灰度，使得带有缺陷信息的散斑图像能有效的被分析判断是否存在缺陷及缺陷的类型。
由于被测曲面物体表面的缺陷，例如划伤、裂痕、凹凸变形、表面脏污等都会改变相干光源的相位信息，经成像后，这些缺陷不良信息会调制到散斑图像，不同类型和不同大小的缺陷会改变散斑图像中的明暗光斑的分布，进而通过分析散斑图像，可以判断所述曲面物体是否存在缺陷及缺陷的类型。
进一步的，在本发明的一些实施例中，通过神经网络深度学习的方法对所述散斑图像进行分析以判断被测曲面物体的表面是否存在缺陷以及缺陷的类型。例如，利用深度学习人工神经网络进行大样本散斑数据训练和学习，按照不同的缺陷样品分别采集100幅以上的图像，对这些能够间接反映表面微细结构的散斑进行分类训练，得到该被测曲面物体表面散斑图像的神经网络模型，通过该神经网络模型可以定义出多种不同的输出状态，例如：无缺陷（OK），脏污，划伤，变形等，进而，在生成散斑图像后，可以通过神经网络模型，根据散斑图像判断所述曲面物体是否存在缺陷及缺陷的类型。
例如，如图3所示，人工神经网络深度训练和学习法一般包括一个输入层、一个隐藏层以及一个输出层。其中，在本发明中，首先，采集数百幅或更多的表面不存在缺陷的曲面物体的散斑图像，通过深度学习的神经网络，对这些散斑图像进行训练，可以得到无缺陷的神经网络模型；采集数百幅或更多的缺陷为“表面脏污”的曲面物体的散斑图像，通过深度学习的神经网络，对这些散斑图像进行训练，可以得到缺陷为“表面脏污”的神经网络模型；而采集数百幅或更多的缺陷为“划伤”的曲面物体的散斑图像，通过深度学习的神经网络，对这些散斑图像进行训练，可以得到缺陷为“划伤”的神经网络模型，依次类推，可以获得表面无缺陷或表面存在各种缺陷的神经网络模型，进而，在判断待测曲面物体表面是否存在缺陷以及缺陷的类型时，神经网络学习的输入层输入的是待测曲面物体的散斑图像，经过隐藏层的各种神经网络模型比对和计算后，可以输出该散斑图像对应的输出状态，如：表面无缺陷（OK）、表面存在脏污、表面存在划伤、表面存在变形等。
可以理解的是，本发明对于人工神经网络深度训练和学习的方法没有特殊限制，其可以为各种常用的人工神经网络深度训练和学习方法。
采用深度神经网络直接对表面缺陷调制到相干光散斑干涉图案进行缺陷学习和检测，与传统的采用图像预处理和自适应分割进行检测方法相比，该方法适应性更强，使用简单，不需要进行复杂的光源设计和参数设置。
根据本发明提供的检测方法可以应用在不同的场合。例如，可以应用在复杂自由曲面的表面缺陷检测，其表面的崩裂、变形及外来污染带来的颜色都会引起对应的散斑图形改变从而被判别出来；也可以应用在高反光的平面和3D曲面，以及玻璃透明等被测产品，正常物体表面的散斑信号也和各种缺陷带来的散斑信号不一样，从而也可以通过本发明提供的检测方法而被识别出。而且，本发明提供的检测方法还可以应用在一些特定材料的被测物体，例如半透明的塑料材质，不同材料混叠的物体表面的缺陷，比如浅层气泡、气孔等，由于相干激光进行干涉时，这些光会穿透到产品内部，由此，被测物体的内部信息同样会显示在散斑图像中，从而被识别出来。
可以理解的是，本发明提供的检测方法的照明光路可以扩展到1个角度，多个角度，以及角度可调节，可以采用一个光电传感器或者多个光电传感器得到散斑图像进行识别，也可以适应于采用透镜成像的散斑图像拍摄。同样的，可以对本发明提供的检测方法中所采用的照明光学设计和光电传感器进行多种改造，以适应于不同应用场合，实现不同检测分辨率的需求。
可以理解的是，可以采用各种方法改变被测曲面物体与光学系统之间的相对位置，进而，本发明提供的检测方法可以对被测曲面物体进行复杂结构和大范围的表面检测。例如，通过与机器人或者单轴、多轴运动机构进行配合，可以采用被测物体不动，改变光学系统的位置的方式，实现对被测曲面物体进行复杂结构和大范围的表面检测；也可以采用光学视觉系统位置固定，通过移动改变被测物体的方式，实现对被测曲面物体进行复杂结构和大范围的表面检测。
此外，参照图4所示，本发明还提出了一种曲面物体的缺陷检测系统，其包括：光源单元，光电传感器2，及检测判断单元3。其中光源单元用于生成多束相干激光并控制所述多束相干激光从多个照射角度对曲面物体4的检测区域进行照明；光电传感器2用于接收所述曲面物体4反射或散射的相干激光信号并生成散斑图像5；检测判断单元3与光电传感器2连接以接收所述散斑图像5，并用于根据所述散斑图像5判断所述曲面物体4是否存在缺陷及缺陷的类型。
在其中的一个实施例中，光源单元包括相干激光光源11，分束器12及角度可调节的多个扩束镜13，其中，相干激光光源11用于生成相干激光，并将所述相干激光通过所述分束器12以生成多束相干激光，所述多个扩束镜12用于接收所述多束相干激光并控制所述多束相干激光从多个照射角度对曲面物体4的检测区域进行照明。
对于分束器12及扩束镜13的结构及工作原理为本领域技术人员所公知，本发明对此不再进行赘述。可以理解的是，可以采用各种方法将各束相干激光连接到扩束镜13，本发明对此没有特殊限定，例如，可以通过光纤14将各相干激光连接到扩束镜13。
可以理解的是，检测判断单元3可以为各种常用的检测判断单元，例如计算机。例如，在本发明的一个实施例中，计算机通过神经网络深度学习的方法获得准确的神经网络模型，进而通过神经网络模型对当前的散斑图像进行比对判断，即可确定该散斑图像对应的曲面物体4的表面是否存在缺陷。
实施本发明实施例，将具有如下有益效果：
根据本发明提供的曲面物体的缺陷的检测方法及检测系统，其采用多束相干激光，例如四束相干激光，在四个方位对曲面物体的检测区域进行照明，表面照明区域有重叠，这些相干激光经过曲面物体的多角度反射后，通过光电传感器接收所述曲面物体反射或散射的相干激光信号并生成散斑图像，由于被测曲面物体表面的缺陷，例如划伤、裂痕、凹凸变形、表面脏污等都会改变相干光源的相位信息，经成像后，这些缺陷不良信息会调制到散斑图像，不同类型和不同大小的缺陷会改变散斑图像中的明暗光斑的分布，进而通过分析散斑图像，可以实现对缺陷的有效检测。同时，采用多角度相干激光散斑成像的表面缺陷检测具有非常大的优势，其光学结构简单，而且能有效解决曲面景深问题和光照均匀性问题，检测精度高，对于工业检测中多种高反光曲面材质的外观缺陷检测提供了一种非常简单实用可行的解决方案。
以上所揭露的仅为本发明较佳实施例而已，当然不能以此来限定本发明之权利范围，因此依本发明权利要求所作的等同变化，仍属本发明所涵盖的范围。

Claims (10)

  1. 一种曲面物体的缺陷的检测方法，其特征在于，包括以下步骤：
    S1）控制光源生成多束相干激光；
    S2）控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明；
    S3）通过光电传感器接收所述曲面物体反射或散射的相干激光信号并生成散斑图像；以及
    S4）根据所述散斑图像判断所述曲面物体是否存在缺陷及缺陷的类型。
  2. 根据权利要求1所述的曲面物体的缺陷的检测方法，其特征在于，所述步骤S1）包括：控制光源生成相干激光，通过分束器将所述相干激光分成多束相干激光。
  3. 根据权利要求2所述的曲面物体的缺陷的检测方法，其特征在于，所述步骤S2）包括：将所述多束相干激光连接到角度可调节的多个扩束镜以控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明。
  4. 根据权利要求1-3任一所述的曲面物体的缺陷的检测方法，其特征在于，所述多束相干激光为四束相干激光。
  5. 根据权利要求1所述的曲面物体的缺陷的检测方法，其特征在于，所述光电传感器为CCD传感器或CMOS传感器。
  6. 根据权利要求5所述的曲面物体的缺陷的检测方法，其特征在于，所述步骤S3）中还包括采用自适应的散斑图像调节算法对所述散斑图像进行调节以使所述散斑图像的灰度分布均匀。
  7. 根据权利要求6所述的曲面物体的缺陷的检测方法，其特征在于，所述自适应的散斑图像调节算法包括以下步骤：
    S31）分析所述散斑图像的灰度直方图分布，以统计出所述散斑图像的灰度分布；以及
    S32）当出现过曝的灰度数量超过阈值后，通过调节相关参数以获得均匀无过曝的散斑图像。
  8. 根据权利要求7所述的曲面物体的缺陷的检测方法，其特征在于，所述相关参数包括光电传感器的曝光时间、光电传感器的增益值、光源亮度、光源分光比及照射角度中的至少一种。
  9. 根据权利要求1所述的曲面物体的缺陷的检测方法，其特征在于，所述步骤S4）包括：通过神经网络深度学习的方法对所述散斑图像进行分析以判断所述曲面物体的表面是否存在缺陷以及缺陷的类型。
  10. 一种曲面物体的缺陷的检测系统，其特征在于，包括：
    光源单元，用于生成多束相干激光并控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明；
    光电传感器，用于接收所述曲面物体反射或散射的相干激光信号并生成散斑图像；以及
    检测判断单元，所述检测判断单元与所述光电传感器连接以接收所述散斑图像，并用于根据所述散斑图像判断所述曲面物体是否存在缺陷及缺陷的类型；
    其中，所述光源单元包括相干激光光源，分束器及角度可调节的多个扩束镜，其中，相干激光光源用于生成相干激光，并将所述相干激光通过所述分束器以生成多束相干激光，所述多个扩束镜用于接收所述多束相干激光并控制所述多束相干激光从多个照射角度对曲面物体的检测区域进行照明。
PCT/CN2018/105110 2018-09-11 2018-09-11 曲面物体的缺陷检测方法及检测系统 WO2020051779A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2018/105110 WO2020051779A1 (zh) 2018-09-11 2018-09-11 曲面物体的缺陷检测方法及检测系统
CN201880001802.0A CN111344553B (zh) 2018-09-11 2018-09-11 曲面物体的缺陷检测方法及检测系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/105110 WO2020051779A1 (zh) 2018-09-11 2018-09-11 曲面物体的缺陷检测方法及检测系统

Publications (1)

Publication Number Publication Date
WO2020051779A1 true WO2020051779A1 (zh) 2020-03-19

Family

ID=69777309

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/105110 WO2020051779A1 (zh) 2018-09-11 2018-09-11 曲面物体的缺陷检测方法及检测系统

Country Status (2)

Country Link
CN (1) CN111344553B (zh)
WO (1) WO2020051779A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117237355A (zh) * 2023-11-15 2023-12-15 宁波快马加鞭科技有限公司 基于图像特征的保温杯表面缺陷检测方法

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117015121B (zh) * 2023-10-07 2023-12-15 南通医疗器械有限公司 一种无影灯的故障告警管理方法及系统

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7221444B1 (en) * 2005-10-14 2007-05-22 3I Systems Inc. Method and system for improved defect sensitivity for inspecting surfaces
CN101699272A (zh) * 2009-11-19 2010-04-28 西北工业大学 一种利用数字剪切散斑干涉进行无损检测的方法及其装置
US20140268105A1 (en) * 2013-03-15 2014-09-18 Zygo Corporation Optical defect inspection system
CN105548199A (zh) * 2015-12-16 2016-05-04 清华大学 一种测量含轴向裂纹圆柱壳裂尖应力强度因子的方法
CN106770373A (zh) * 2017-02-08 2017-05-31 西安工业大学 一种表面缺陷检测方法
CN108106556A (zh) * 2018-02-02 2018-06-01 上海交通大学 基于数字散斑干涉的曲面物体离面形变测量方法及装置
CN108280824A (zh) * 2018-01-18 2018-07-13 电子科技大学 基于图像配准及融合的激光剪切散斑干涉缺陷检测系统

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8259198B2 (en) * 2009-10-20 2012-09-04 Apple Inc. System and method for detecting and correcting defective pixels in an image sensor
CN103810708B (zh) * 2014-02-13 2016-11-02 西安交通大学 一种激光散斑图像深度感知方法及装置
JP2017116982A (ja) * 2015-12-21 2017-06-29 ソニー株式会社 画像解析装置、画像解析方法及び画像解析システム
CN105578009A (zh) * 2015-12-28 2016-05-11 苏州中启维盛机器人科技有限公司 光斑成像装置
CN106296716A (zh) * 2016-08-24 2017-01-04 深圳奥比中光科技有限公司 光源的功率调整方法、深度测量方法及装置
CN106979749B (zh) * 2016-11-02 2019-05-21 北京信息科技大学 一种光条图像成像参数的模糊自适应调整方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7221444B1 (en) * 2005-10-14 2007-05-22 3I Systems Inc. Method and system for improved defect sensitivity for inspecting surfaces
CN101699272A (zh) * 2009-11-19 2010-04-28 西北工业大学 一种利用数字剪切散斑干涉进行无损检测的方法及其装置
US20140268105A1 (en) * 2013-03-15 2014-09-18 Zygo Corporation Optical defect inspection system
CN105548199A (zh) * 2015-12-16 2016-05-04 清华大学 一种测量含轴向裂纹圆柱壳裂尖应力强度因子的方法
CN106770373A (zh) * 2017-02-08 2017-05-31 西安工业大学 一种表面缺陷检测方法
CN108280824A (zh) * 2018-01-18 2018-07-13 电子科技大学 基于图像配准及融合的激光剪切散斑干涉缺陷检测系统
CN108106556A (zh) * 2018-02-02 2018-06-01 上海交通大学 基于数字散斑干涉的曲面物体离面形变测量方法及装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117237355A (zh) * 2023-11-15 2023-12-15 宁波快马加鞭科技有限公司 基于图像特征的保温杯表面缺陷检测方法
CN117237355B (zh) * 2023-11-15 2024-02-02 宁波快马加鞭科技有限公司 基于图像特征的保温杯表面缺陷检测方法

Also Published As

Publication number Publication date
CN111344553A (zh) 2020-06-26
CN111344553B (zh) 2023-09-12

Similar Documents

Publication Publication Date Title
CN108445007B (zh) 一种基于图像融合的检测方法及其检测装置
CN107735674B (zh) 表面缺陷检测装置、表面缺陷检测方法及钢材的制造方法
JP3560694B2 (ja) レンズ検査のシステムと方法
CN111344559B (zh) 缺陷检测方法及缺陷检测系统
JP6394514B2 (ja) 表面欠陥検出方法、表面欠陥検出装置、及び鋼材の製造方法
US20200018707A1 (en) Appearance inspection device, lighting device, and imaging lighting device
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
RU2011130930A (ru) Способ неразрушающего контроля механической детали
CN111899215B (zh) 一种光学元件体缺陷的提取方法
KR20180009791A (ko) 표면 결함 검출 방법, 표면 결함 검출 장치 및, 강재의 제조 방법
TW201937225A (zh) 用於光學成像系統中之自動聚焦之範圍區別器
WO2020051779A1 (zh) 曲面物体的缺陷检测方法及检测系统
CN110402386A (zh) 圆筒体表面检查装置及圆筒体表面检查方法
JP2011208941A (ja) 欠陥検査装置およびその方法
KR20220110291A (ko) 레이저 기반 내포물 검출 시스템 및 방법
CN112505057A (zh) 一种滚动面缺陷检测系统和方法
JPH0373831A (ja) 欠陥検査装置
US11657495B2 (en) Non-lambertian surface inspection system using images scanned in different directions
CN111179248B (zh) 一种透明平滑曲面缺陷识别方法及检测装置
KR102368707B1 (ko) 라인 스캔용 논-램버시안 표면 검사 시스템
JP2955686B2 (ja) 表面欠陥検査装置
WO2020051780A1 (zh) 图像传感器表面缺陷检测方法及检测系统
JP7015078B1 (ja) 検査方法および検査システム
Li et al. An intelligent system for real time automatic defect inspection on specular coated surfaces
JPH09138201A (ja) 表面検査装置

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: 18933573

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 17/08/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18933573

Country of ref document: EP

Kind code of ref document: A1