WO2019114453A1 - 一种针对类镜面物体的高光区域自适应匀光的方法 - Google Patents

一种针对类镜面物体的高光区域自适应匀光的方法 Download PDF

Info

Publication number
WO2019114453A1
WO2019114453A1 PCT/CN2018/113245 CN2018113245W WO2019114453A1 WO 2019114453 A1 WO2019114453 A1 WO 2019114453A1 CN 2018113245 W CN2018113245 W CN 2018113245W WO 2019114453 A1 WO2019114453 A1 WO 2019114453A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
block
point
mirror
wavelet decomposition
Prior art date
Application number
PCT/CN2018/113245
Other languages
English (en)
French (fr)
Inventor
杜娟
陈芳
胡跃明
Original Assignee
华南理工大学
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 华南理工大学 filed Critical 华南理工大学
Publication of WO2019114453A1 publication Critical patent/WO2019114453A1/zh

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
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20021—Dividing image into blocks, subimages or windows
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20048—Transform domain processing
    • G06T2207/20064—Wavelet transform [DWT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20212—Image combination
    • G06T2207/20224—Image subtraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection

Definitions

  • the invention relates to the production, detection and related links of precision mirror-like objects, relates to feature region extraction, automatic homogenization processing in high-light regions, and particularly relates to a method for adaptive homogenization of high-light regions of mirror-like objects.
  • Machine vision technology uses digital images as a means of obtaining information, and is closely integrated with computer graphics, automation technology and other related fields, and is very suitable for defect detection on the surface of products.
  • the surface defects of most industrial products mainly include dents, streaks, cracks, cracks, stains and the like.
  • the light source of the general visual inspection system will produce strong reflection on the surface of the product, that is, the highlight portion. This problem has a great influence on the subsequent image links, which will greatly reduce the detection rate of defects.
  • the light source can be driven from the fixed direction by the diffuse reflection and scattering phenomenon of light.
  • Disperse evenly illuminate the surface of the product, thereby eliminating the high light on the surface of the product, and the camera collects a more uniform image of gray scale; from the perspective of the algorithm, for the highlight problem that cannot be eliminated by the hardware, the image is homogenized for the subsequent defect detection. Convenience, which improves the accuracy and efficiency of detection.
  • the method analyzes and processes the acquired image from the perspective of wavelet decomposition and image restoration, and can realize the extraction of the highlight region of the image and the adaptive homogenization of the highlight region.
  • the object of the present invention is a method for adaptively homogenizing a high-light region of a mirror-like object, which has the characteristics of accuracy and robustness compared with the existing defect detection method, and the specific technical solutions are as follows.
  • a method for adaptively homogenizing a highlight region of a mirror-like object comprising the steps of:
  • Wavelet decomposition of the image to be detected 2-layer wavelet decomposition of the source image to be detected, and extracting the approximate matrix CA1, CA2 of the first layer and the second layer after two-layer wavelet decomposition, and the first and second The detail coefficient of the layer wavelet decomposition, and taking out the maximum value MAX of the second layer wavelet decomposition approximation matrix, and the minimum value MIN; the detail coefficient includes a horizontal component, a vertical component, and a diagonal component;
  • the high-light area mark image I1 is obtained: the detail coefficient of the second layer wavelet decomposition is zero-processed and wavelet reconstruction is performed, and a new first-layer wavelet decomposition approximate matrix is obtained, and the first layer wavelet is obtained.
  • the decomposed detail coefficient is zeroed, and the wavelet reconstruction is performed again to obtain a new image; the new image is correspondingly subtracted from the source image, the threshold T is set, the highlight region is extracted, and the green image is displayed on the source image, the new image.
  • Parameter initialization Initialize the direction of the equal line, confidence, data items;
  • Best matching block search the block with the highest priority is taken as the padding target block according to the calculated priority, and the search is performed in the 600*800 area block centered on the center point of the block with the highest priority. Find the matching block with the smallest gray level difference from the block with the highest priority; (8) Fill and update confidence: assign the searched fast matching value to the block with the highest priority, and then recalculate the confidence.
  • step (1) the image of the image to be detected which is weakly reflected is obtained by performing image capturing in a dark field environment by changing the optical path a plurality of times.
  • the stretching coefficient in the step (3) is set to 1.5.
  • step (4) the threshold T is set to 150, and the difference image E(a, b) obtained by subtracting the reconstructed image from the source image is compared with a threshold T, and (a, b) refers to a pixel point.
  • step (6) sets a mask of 9*9, performs priority calculation in units of masks of 9*9 size, selects the block with the highest priority, and searches for the most matching fast.
  • step (7) performs different settings according to different detection objects, and the search center center point is the center point of the block with the highest priority, and searches in the search area to find the optimal matching block.
  • the present invention has the following advantages and beneficial effects:
  • the present invention improves the accuracy of illumination unevenness detection by using a method of adaptive homogenization for a specular object in a high-light region to improve the accuracy of illumination unevenness detection, and the process is simple and operable. Strong.
  • the present invention has strong robustness, and can achieve good effects for uneven illumination of different intensities.
  • the invention also has strong applicability, and can be applied not only to the detection of mirror-like object defects, but also to the detection of other objects with strong contrast of gray scale.
  • Figure 1 is a schematic view of an image mark of the present invention
  • Figure 3 is a flow chart of the area to be repaired of the present invention.
  • Figure 4 is a flow chart of the best match search.
  • Figure 5 is a graph showing the results of an implementation of a method for adaptively homogenizing a highlight region of a mirror-like object in an example.
  • the invention provides a method for adaptively homogenizing a high-light area of a mirror-like object, comprising the following steps:
  • the high-light area mark image I1 is obtained: the detail coefficient of the second layer wavelet decomposition is zero-processed and wavelet reconstruction is performed, and a new first-layer wavelet decomposition approximate matrix is obtained, and the first layer wavelet is obtained.
  • the decomposed detail coefficient is zeroed, and the wavelet reconstruction is performed again to obtain a new image.
  • Best matching block search the block with the highest priority is taken as the padding target block according to the calculated priority, and the search is performed in the 600*800 area block centered on the center point of the block with the highest priority. Find the matching block with the smallest gray level difference from the block with the highest priority.
  • FIG. 5 is a source image, in which a region with a higher luminance is a highlight region, and there is a significant difference in color and pixel values from the neighboring neighborhood; 2-layer wavelet decomposition, stretching, and Reconstruction, detecting and extracting the position of the highlight region in the source image, as shown in (b) of FIG. 5, the green marker region is the extracted highlight region; and then performing the steps of searching and filling in the algorithm of the present invention The highlight area is automatically filled, and the result is finally obtained, as shown in (c) of FIG.
  • the brightness of the highlight region in (c) of Figure 5 is significantly reduced; from the numerical point of view, the maximum of the gray values of all the pixels in (a) of Figure 5 is 180, and the minimum is 112.
  • the gray value of all the pixel value points in (c) of FIG. 5 after the processing of the algorithm in this paper has a maximum of 140 and a minimum value of 112, and the contrast of the image (the difference between the maximum value and the minimum value: the minimum value) is also 60.7. % dropped to 25% and the effect was significant.

Landscapes

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

Abstract

本发明提供了一种针对类镜面物体的高光区域自适应匀光的方法，通过对采集的图像进行两层小波分解与重构，提取出图像的高光区域。然后再运用局部搜索图像修复算法实现高光区域的自适应匀光处理。这种方法可以消除检测过程中高光区域对缺陷检测的影响且无需人工干涉，自动根据源图像的信息进行均匀高光区域信息。相比于其他匀光算法，该方法可以得到更均匀的处理结果，且具有更好的适应性。

Description

一种针对类镜面物体的高光区域自适应匀光的方法 技术领域
本发明涉及精密类镜面物体的生产和检测及相关环节，涉及特征区域提取，高光区域的自动匀光处理，特别涉及一种针对类镜面物体的高光区域自适应匀光的方法。
背景技术
制造业的飞速发展，对产品的检测任务提出了更高的要求，传统的人工手段已经无法满足大批量产品的高精度、快速甚至实时动态的检测要求。机器视觉技术以数字图像作为获取信息的手段，与计算机图形学、自动化技术等相关领域紧密结合，很适合对产品表面进行缺陷检测。
目前，大部分工业产品表面缺陷主要包括凹痕、条痕、裂纹、龟裂、色斑等。一般的视觉检测系统的光源会在产品表面产生强烈的反光，即高光部分，这个问题对后续的图像环节有较大影响，会使得缺陷的检测率大大降低。为了消除表面光滑的工业产品产生的高光，提高整个检测过程的工作效率与检测精度，从硬件角度可以通过光的漫反射和散射现象，将光源从固定方向发出的、具有高度方向性的光打散，使其均匀照射在产品表面，从而消除产品表面的高光，摄像机采集灰度较为均匀图像；从算法角度，对于硬件无法消除的高光问题，通过算法进行匀光处理，为后续的缺陷检测提供了便利，从而提高了检测的精度和效率。
本方法从小波分解与图像修复的角度对采集的图像分析处理，能 实现图像高光区域的提取与高光区域自适应匀光处理。
发明内容
本发明的目的在一种针对类镜面物体的高光区域自适应匀光的方法，相于现有的缺陷检测方法具有准确性和鲁棒性的特点，具体技术方案如下。
一种针对类镜面物体的高光区域自适应匀光的方法，其包括以下步骤：
(1)获取待检测的源图像I：将类镜面物体放入一个暗室中，暗室内表面粗糙，光源产生的光线经由暗室内部粗糙的表面的第一次反射照在磨砂半球体表面，再经由磨砂半球体表面的二次透光照在待测物体上，由视觉系统采集而得到待测物体源图像I；经由两次光路变化来将能量集中的光分散来减少图像上高光的强度与区域；
(2)对待检测源图像进行小波分解：对待检测的源图像进行2层小波分解，提取出经过2层小波分解后的第一层、第二层的近似矩阵CA1、CA2，和第一、二层小波分解的细节系数，同时取出第二层小波分解近似矩阵的最大值MAX，和最小值MIN；所述细节系数包含水平分量、垂直分量、对角分量；
(3)对第二层小波分解近似矩阵进行自适应拉伸处理：取第二层小波分解近似矩阵CA2中的任意一个像素点x的灰度值依次与源图像的平均灰度值average进行比较，如果小于average，则
Figure PCTCN2018113245-appb-000001
如果大于average，则
Figure PCTCN2018113245-appb-000002
其中A(x)是x点经拉伸后的灰度值，G是拉伸系数，X MAX是源图像的最大灰度值，X MIN是源图像的最小灰度值，X是近似矩阵CA2上x点的灰度值；
(4)进行小波重构后得到高光区域标记图像I1：对第二层小波分解的细节系数置零处理后进行小波重构，得到新的第一层小波分解的近似矩阵，对第一层小波分解的细节系数置零处理，再次进行小波重构得到新图像；新图像与源图像进行对应相减，设定阈值T,提取出高光区域，并以绿色标记显示在源图像上，该新图像标记为I2；(5)参数初始化：初始化等照线方向，置信度、数据项；
(6)计算优先权：根据I2，寻找绿色标记的高光区域，即为待填充区域，并提取出其边缘，其他区域标记为已知区域；设定以边缘点为中心点的9*9的掩模，计算掩模的优先权P (p)＝C (p)*D (p),C (p)是该掩模的置信度，D (p)是该掩模的数据项；
(7)最佳匹配块搜索：根据计算出的优先权取出优先权最大的块为填充目标块，在以该优先权最大的块的中心点为中心的600*800的区域块内进行搜索，找出与优先权最大的块的灰度差值最小的匹配块；(8)填充与更新置信度：将搜索出的匹配快的值对优先权最大的块对应赋值，然后重新计算置信度。
进一步地，步骤(1)利用多次改变光路在暗场环境下进行图像的拍摄，获得反光弱的待检测源图像。
进一步地，步骤(3)中拉伸系数设为1.5。
进一步地，步骤(4)设置阈值T为150，将重构后的图像与源图像相减后的差值图像E(a，b)与阈值T进行比较，(a,b)指像素点的坐标，假定图像中最左上方的点为起始点，其中a表示像素点与起始点在行数上的距离，b表示像素点与起始点在列数上的距离，E(a，b)表示差值图像在第a行第b列的像素点的像素值；若E(a,b)≤T时，E(a,b)＝0；若E(a,b)>T时，E(a,b)＝255。
进一步地，步骤(6)设定9*9的掩模，以9*9大小的掩模为单位进行优先权的计算，选出优先权最大的块，并搜索得到最有的匹配快。
进一步地，步骤(7)根据不同的检测物体进行不同的设置，搜索区域中心点为优先权最大的块的中心点，在该搜索区域内进行搜索，找出最优匹配块。
本发明与现有技术相比，具有如下优点和有益效果：
(1)本发明通过一种针对类镜面物体的高光区域自适应匀光的方法对光照不均的图像的灰度变化分析，提高了光照不均区域检测的准确度，过程简单，可操作性强。
(2)本发明具有较强的鲁棒性，对于不同的强度的不均匀光照都可达到很好的效果。
(3)本发明还具有较强的适用性，不仅可以运用在类镜面物体缺陷的检测，还能推广到其他灰度对比强烈的物体的检测。
附图说明
图1是本发明图像标记示意图；
图2是本发明自适应匀光方法的总流程图；
图3是本发明待修复区域标记流程图；
图4是最佳匹配搜索流程图。
图5是实例中针对类镜面物体的高光区域自适应匀光的方法实施结果图。
具体实施方式
以下结合附图和实例对本发明的具体实施作进一步说明，但本发明的实施和保护不限于此。
本发明一种针对类镜面物体的高光区域自适应匀光的方法，包括以下步骤：
(1)获取待检测的源图像I：将类镜面物体放入一个暗室中，暗室内表面粗糙，光源产生的光线经由暗室内部粗糙的表面的第一次反射照在磨砂半球体表面，再经由磨砂半球体表面的二次透光照在待测物体上，由视觉系统采集而得到待测物体源图像I。经由两次光路变化来将能量集中的光分散来减少图像上高光的强度与区域。
(2)对待检测源图像进行小波分解：对待检测的源图像进行2层小波分解，小波函数可采用“db8”，提取出小波分解后第一层、第二层的近似矩阵CA1、CA2，和第一、二层小波分解的细节系数(包括水平分量、垂直分量、对角分量)，并取出第二层小波分解近似矩阵的最大值MAX＝1035.3227，和最小值MIN＝421.4969。
(3)对第二层小波分解近似矩阵进行自适应拉伸处理：将第二层小波分解近似矩阵CA2中的任一点x的灰度值依次与源图像的平均灰度average＝129.1114进行比较，如果小于average，则
Figure PCTCN2018113245-appb-000003
如果大于average，则
Figure PCTCN2018113245-appb-000004
其中A(x)是x 点经拉伸后的灰度值，G是拉伸系数，为0.1，X MAX＝255是最大值源图像的最大灰度值，X MIN＝102是最大值源图像的最小灰度值，X是近似矩阵CA2上x点的灰度值。
(4)进行小波重构后得到高光区域标记图像I1：对第二层小波分解的细节系数置零处理后进行小波重构，得到新的第一层小波分解的近似矩阵，对第一层小波分解的细节系数置零处理，再次进行小波重构得到新图像。新图像与源图像进行对应相减，设定阈值T＝143,提取出高光区域，并以绿色标记显示在源图像上，该新图像标记为I2。
(5)参数初始化：初始化等照线方向，置信度C＝O、数据项D＝-0.1000。
(6)计算优先权：根据I2，寻找绿色标记的高光区域，即为待填充区域，并提取出其边缘，其他区域标记为已知区域。设定以边缘点为中心点的9*9的掩模，计算掩模的优先权P (p)＝C (p)*D (p),C (p)是该掩模的置信度，D (p)是该掩模的数据项。
(7)最佳匹配块搜索：根据计算出的优先权取出优先权最大的块为填充目标块，在以该优先权最大的块的中心点为中心的600*800的区域块内进行搜索，找出与优先权最大的块的灰度差值最小的匹配块。
(8)填充与更新置信度：将搜索出的匹配快的值对优先权最大的块对应赋值，然后重新计算置信度。
图5中(a)是源图像，图中亮度较高区域即为高光区域，其与周边邻域在颜色和像素值上存在明显的差异；对源图像进行2层小波分解、 拉伸、和重建，检测提取出高光区域在源图像中的位置，如图5中(b)所示，绿色标记区域即为提取出的高光区域；再运行本文算法中的搜素和填充等步骤对提取出的高光区域实现自动填充，最终得到结果，如图5中(c)所示。从肉眼观察，图5中(c)中高光区域的亮度明显降低；从数值角度上而言，图5中(a)的所有像素点的灰度值中最大为180，最小为112，而进过本文算法处理过之后的图5中(c)的所有像素值点的灰度值最大为140，最小值为112，而图像的对比度(最大值与最小值的差值：最小值)也有60.7％降到了25％，效果显著。

Claims (6)

  1. 一种针对类镜面物体的高光区域自适应匀光的方法，其特征在于包括以下步骤：
    (1)获取待检测的源图像I：将类镜面物体放入一个暗室中，暗室内表面粗糙，光源产生的光线经由暗室内部粗糙的表面的第一次反射照在磨砂半球体表面，再经由磨砂半球体表面的二次透光照在待测物体上，由视觉系统采集而得到待测物体源图像I；经由两次光路变化来将能量集中的光分散来减少图像上高光的强度与区域；
    (2)对待检测源图像进行小波分解：对待检测的源图像进行2层小波分解，提取出经过2层小波分解后的第一层、第二层的近似矩阵CA1、CA2，和第一、二层小波分解的细节系数，同时取出第二层小波分解近似矩阵的最大值MAX，和最小值MIN；所述细节系数包含水平分量、垂直分量、对角分量；
    (3)对第二层小波分解近似矩阵进行自适应拉伸处理：取第二层小波分解近似矩阵CA2中的任意一个像素点x的灰度值依次与源图像的平均灰度值average进行比较，如果小于average，则
    Figure PCTCN2018113245-appb-100001
    如果大于average，则
    Figure PCTCN2018113245-appb-100002
    其中A(x)是x点经拉伸后的灰度值，G是拉伸系数，X MAX是源图像的最大灰度值，X MIN是源图像的最小灰度值，X是近似矩阵CA2上x点的灰度值；
    (4)进行小波重构后得到高光区域标记图像I1：对第二层小波分 解的细节系数置零处理后进行小波重构，得到新的第一层小波分解的近似矩阵，对第一层小波分解的细节系数置零处理，再次进行小波重构得到新图像；新图像与源图像进行对应相减，设定阈值T,提取出高光区域，并以绿色标记显示在源图像上，该新图像标记为I2；
    (5)参数初始化：初始化等照线方向，置信度、数据项；
    (6)计算优先权：根据I2，寻找绿色标记的高光区域，即为待填充区域，并提取出其边缘，其他区域标记为已知区域；设定以边缘点为中心点的9*9的掩模，计算掩模的优先权P (p)＝C (p)*D (p),C (p)是该掩模的置信度，D (p)是该掩模的数据项；
    (7)最佳匹配块搜索：根据计算出的优先权取出优先权最大的块为填充目标块，在以该优先权最大的块的中心点为中心的600*800的区域块内进行搜索，找出与优先权最大的块的灰度差值最小的匹配块；
    (8)填充与更新置信度：将搜索出的匹配快的值对优先权最大的块对应赋值，然后重新计算置信度。
  2. 根据权利要求1所述的一种针对类镜面物体的高光区域自适应匀光的方法，其特征在于步骤(1)利用多次改变光路在暗场环境下进行图像的拍摄，获得反光弱的待检测源图像。
  3. 根据权利要求1所述的一种针对类镜面物体的高光区域自适应匀光的方法，其特征在于步骤(3)中拉伸系数设为1.5。
  4. 根据权利要求1所述的一种针对类镜面物体的高光区域自适应匀光的方法，其特征在于步骤(4)设置阈值T为150，将重构后的图像与源图像相减后的差值图像E(a，b)与阈值T进行比较，(a,b) 指像素点的坐标，假定图像中最左上方的点为起始点，其中a表示像素点与起始点在行数上的距离，b表示像素点与起始点在列数上的距离，E(a，b)表示差值图像在第a行第b列的像素点的像素值；若E(a,b)≤T时，E(a,b)＝0；若E(a,b)>T时，E(a,b)＝255。
  5. 根据权利要求1所述的一种针对类镜面物体的高光区域自适应匀光的方法，其特征在于步骤(6)设定9*9的掩模，以9*9大小的掩模为单位进行优先权的计算，选出优先权最大的块，并搜索得到最有的匹配快。
  6. 根据权利要求1所述的一种针对类镜面物体的高光区域自适应匀光的方法，其特征在于步骤(7)根据不同的检测物体进行不同的设置，搜索区域中心点为优先权最大的块的中心点，在该搜索区域内进行搜索，找出最优匹配块。
PCT/CN2018/113245 2017-12-15 2018-10-31 一种针对类镜面物体的高光区域自适应匀光的方法 WO2019114453A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201711360310.4 2017-12-15
CN201711360310.4A CN108109138B (zh) 2017-12-15 2017-12-15 一种针对类镜面物体的高光区域自适应匀光的方法

Publications (1)

Publication Number Publication Date
WO2019114453A1 true WO2019114453A1 (zh) 2019-06-20

Family

ID=62217524

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/113245 WO2019114453A1 (zh) 2017-12-15 2018-10-31 一种针对类镜面物体的高光区域自适应匀光的方法

Country Status (2)

Country Link
CN (1) CN108109138B (zh)
WO (1) WO2019114453A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108109138B (zh) * 2017-12-15 2021-08-06 华南理工大学 一种针对类镜面物体的高光区域自适应匀光的方法
CN109829480A (zh) * 2019-01-04 2019-05-31 广西大学 物体表面高光特征检测与材质分类的方法及系统
CN110322409B (zh) * 2019-06-14 2021-08-31 浙江大学 一种基于标记图的改进型小波变换图像融合方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105354802A (zh) * 2015-10-23 2016-02-24 哈尔滨工业大学 基于双向梯度预测的高光谱图像连续谱段恢复方法
WO2016176209A1 (en) * 2015-04-28 2016-11-03 Dolby Laboratories Licensing Corporation Image highlight detection and rendering
CN106803240A (zh) * 2016-12-30 2017-06-06 大连海事大学 一种工业图像均光处理方法
CN108109138A (zh) * 2017-12-15 2018-06-01 华南理工大学 一种针对类镜面物体的高光区域自适应匀光的方法

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100535944C (zh) * 2007-06-29 2009-09-02 北京大学 一种用于视觉凸壳绘制的高光区域选取方法及其装置
CN102393913B (zh) * 2011-10-31 2017-04-05 北京航空航天大学 一种基于光谱指纹特征的弱小目标精确跟踪方法
CN102819735B (zh) * 2012-08-17 2015-07-15 深圳辉锐天眼科技有限公司 基于视频帧图像的火焰检测方法
CN103198483B (zh) * 2013-04-07 2015-09-30 西安电子科技大学 基于边缘和光谱反射率曲线的多时相遥感图像配准方法
CN103164704B (zh) * 2013-04-12 2016-05-11 山东师范大学 一种基于混合高斯模型的虹膜图像分割算法
CN104200673B (zh) * 2014-09-01 2016-04-06 西南交通大学 一种基于道路图像的路面湿滑检测方法
CN104299230B (zh) * 2014-09-25 2017-02-15 北京理工大学 一种利用红黑小波变换的图像显著性检测方法
CN104732235B (zh) * 2015-03-19 2017-10-31 杭州电子科技大学 一种消除夜间道路反光干扰的车辆检测方法
CN104751181B (zh) * 2015-04-02 2018-01-12 山东大学 一种基于相对丰度的高光谱图像解混方法
CN105184806A (zh) * 2015-10-15 2015-12-23 南京工程学院 一种度量图像模糊程度的方法
CN105957042B (zh) * 2016-06-07 2019-02-15 北京理工大学 内镜图像的高光区域消除方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016176209A1 (en) * 2015-04-28 2016-11-03 Dolby Laboratories Licensing Corporation Image highlight detection and rendering
CN105354802A (zh) * 2015-10-23 2016-02-24 哈尔滨工业大学 基于双向梯度预测的高光谱图像连续谱段恢复方法
CN106803240A (zh) * 2016-12-30 2017-06-06 大连海事大学 一种工业图像均光处理方法
CN108109138A (zh) * 2017-12-15 2018-06-01 华南理工大学 一种针对类镜面物体的高光区域自适应匀光的方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LIU, CHUN: "The Research on fast image inpainting algorithms", ELECTR. TECHN. & INF. SCIENCE, CHINA MASTER'S THESES, no. 7, 15 July 2013 (2013-07-15), pages 30 - 36 *
ZHANG, ZHEN ET AL.: "Improved MASK Dodging method based on Wavelet", JOURNAL OF REMOTE SENSING, vol. 13, no. 6, 15 November 2009 (2009-11-15), pages 1079, XP055618722 *

Also Published As

Publication number Publication date
CN108109138A (zh) 2018-06-01
CN108109138B (zh) 2021-08-06

Similar Documents

Publication Publication Date Title
CN105334219B (zh) 一种残差分析动态阈值分割的瓶口缺陷检测方法
TWI787296B (zh) 光學檢測方法、光學檢測裝置及光學檢測系統
CN104458755B (zh) 一种基于机器视觉的多类型材质表面缺陷检测方法
WO2019114453A1 (zh) 一种针对类镜面物体的高光区域自适应匀光的方法
CN108416766B (zh) 双侧入光式导光板缺陷视觉检测方法
TW201702586A (zh) 光學薄膜缺陷辨識方法及其系統
CN110610496B (zh) 一种对光照变化鲁棒的荧光胶缺陷分割方法
CN108846397B (zh) 一种基于图像处理的电缆半导电层自动检测方法
CN109001212A (zh) 一种基于机器视觉的不锈钢汤勺缺陷检测方法
JP2004294202A (ja) 画面の欠陥検出方法及び装置
WO2009000689A1 (en) Method for optical inspection of a matt surface and apparatus for applying this method
US20200096454A1 (en) Defect detection system for aircraft component and defect detection method for aircraft component
JPWO2019151393A1 (ja) 食品検査システム、食品検査プログラム、食品検査方法および食品生産方法
US11680909B2 (en) Automated inspection of foreign materials, cracks and other surface anomalies
CN109064439B (zh) 基于分区的单侧入光式导光板暗影缺陷提取方法
CN114119591A (zh) 一种显示屏画面质量检测方法
CN115272256A (zh) 亚像素级传感光纤路径高斯提取方法及系统
CN115100206A (zh) 用于具有周期图案纺织物的印花缺陷识别方法
CN114018946B (zh) 一种基于OpenCV的高反光瓶盖缺陷检测方法
Xia et al. An automatic machine vision-based algorithm for inspection of hardwood flooring defects during manufacturing
KR101862310B1 (ko) 얼룩결함 검사장치 및 얼룩결함 검사방법
CN108805854B (zh) 一种复杂环境下药片快速计数与完整性检测方法
CN116934752B (zh) 一种基于人工智能的玻璃检测方法及系统
CN114092682A (zh) 基于机器学习的小金具类缺陷检测算法
CN104573635B (zh) 一种基于三维重建的微小高度识别方法

Legal Events

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

Ref document number: 18888441

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 30.09.2020)

122 Ep: pct application non-entry in european phase

Ref document number: 18888441

Country of ref document: EP

Kind code of ref document: A1