JP2006030093A - Method for detecting wavy defect - Google Patents

Method for detecting wavy defect Download PDF

Info

Publication number
JP2006030093A
JP2006030093A JP2004212293A JP2004212293A JP2006030093A JP 2006030093 A JP2006030093 A JP 2006030093A JP 2004212293 A JP2004212293 A JP 2004212293A JP 2004212293 A JP2004212293 A JP 2004212293A JP 2006030093 A JP2006030093 A JP 2006030093A
Authority
JP
Japan
Prior art keywords
line segment
pixel
candidate line
candidate
pixels
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
JP2004212293A
Other languages
Japanese (ja)
Inventor
Noritsugu Hamada
徳亜 濱田
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sumitomo Electric Industries Ltd
Original Assignee
Sumitomo Electric Industries Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sumitomo Electric Industries Ltd filed Critical Sumitomo Electric Industries Ltd
Priority to JP2004212293A priority Critical patent/JP2006030093A/en
Publication of JP2006030093A publication Critical patent/JP2006030093A/en
Pending legal-status Critical Current

Links

Images

Landscapes

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

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for detecting wavy defects, which can precisely detect the wavy defect, even when the surface of an object to be measured has a satin finish. <P>SOLUTION: Variations in luminance along line segments, each including a certain number of pixels are acquired from a target pixel in an image omnidirectionaly about the target pixel, and a candidate segment n1 in the direction in which variations in the luminance becomes minimum, is extracted. Then, variations in the luminance along line segments, each including the certain number of pixels are acquired from the top of the candidate segment n1 in a plurality of directions about the top pixel, and a candidate segment n2 in the direction in which the variation in luminance becomes minimum, is extracted. Furthermore, when the angle between the candidate segment n1 and the candidate segment n2 is not more than a prescribed value, variations in the luminance along line segments, each including the certain number of pixels are acquired from the top of the candidate segment n2 in a plurality of directions about the top pixel, and a candidate segment in the direction in which the variation in the luminance becomes minimum, is extracted, and in the same way, above steps are repeated to extract candidate segments n3 to nm. If the number m of the continuously extracted candidate segments n1 to nm is not less than a certain value, the plurality of candidate segments n1 to nm are determined as being wavy defects. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

本発明は、筋状欠陥の検出方法に関するものである。特に、梨地状表面を有する検査対象に生じた筋状の欠陥を精度良く検出することができる筋状欠陥の検出方法に関するものである。   The present invention relates to a method for detecting streak defects. In particular, the present invention relates to a method for detecting a streak defect that can accurately detect a streak defect generated in an inspection object having a satin surface.

セラミックスや銅タングステンといった粉末冶金材料は半導体レーザ・発光ダイオードの放熱基板材料といった電子部品や、軸受け・ローラといった耐摩耗部品など、様々な分野で使用されている。これらの部品は高温・高速摺動などの厳しい条件下で使用されることから、品質保証は非常に重要である。この品質保証項目の中でも、特にキズについては製品の熱伝導性の低下、破損の原因となるため、最も重要な保証項目の一つといえる。   Powder metallurgy materials such as ceramics and copper tungsten are used in various fields such as electronic parts such as semiconductor lasers and heat dissipation substrate materials for light emitting diodes, and wear-resistant parts such as bearings and rollers. Since these parts are used under severe conditions such as high temperature and high speed sliding, quality assurance is very important. Among these quality assurance items, scratches, in particular, are one of the most important assurance items because they cause a decrease in the thermal conductivity and damage of the product.

一方、物体表面に存在する筋状の欠陥(引っかき傷など)を照明ムラや物体表面の斑点から区別して光学的に自動検出する方法が特許文献1に開示されている。この文献に記載の発明では、筋状の欠陥は一般に細長く連続した欠陥であることを前提に、(1)斑点の明度よりも筋状欠陥の明度の方が明るい場合は、注目画素の周囲で1番目と2番目に明るい画素の方向を筋状の欠陥の方向と判断し、(2)斑点の明度よりも筋状欠陥の明度の方が暗い場合は、注目画素の周囲で1番目と2番目に暗い画素の方向を筋状の欠陥の方向と判断するロジックを利用している。   On the other hand, Patent Document 1 discloses a method of automatically detecting optically a streak-like defect (such as a scratch) existing on an object surface by distinguishing it from illumination unevenness or spots on the object surface. In the invention described in this document, assuming that the streak defect is generally a long and thin defect, (1) if the lightness of the streak defect is brighter than the lightness of the spot, it is around the pixel of interest. The direction of the first and second brightest pixels is determined as the direction of the streak defect. (2) If the lightness of the streak defect is darker than the brightness of the spot, the first and second pixels around the pixel of interest The logic that determines the direction of the second dark pixel as the direction of the streak defect is used.

特開平10-289319号公報Japanese Patent Laid-Open No. 10-289319

しかし、上記の検出方法でも、検査対象の表面が梨地状の場合に精度良く筋状欠陥を検出することができない場合があった。   However, even with the above-described detection method, there are cases in which a streak defect cannot be detected with high accuracy when the surface of the inspection object is a satin finish.

特許文献1に記載の技術では、注目画素の周囲で1番目と2番目に明るい(または暗い)画素の方向を筋状の欠陥の方向と判断するため、筋状欠陥の方が常に斑点よりも明るく(または暗く)なければ欠陥を認識することができない。   In the technique described in Patent Document 1, the direction of the first and second brightest (or darkest) pixels around the pixel of interest is determined as the direction of the streak defect. The defect cannot be recognized unless it is bright (or dark).

一般に粉末冶金材料の表面には梨地状のランダムな斑点模様がある。この斑点模様は検査対象表面を画像とした場合、明暗模様として表されるため、筋状欠陥の認識の障害となる。特に、この斑点は筋状欠陥よりも必ず明るいか必ず暗い状態であるとは限らず、その場合には正確に欠陥を検知することができない。   In general, the surface of powder metallurgy material has a textured random spot pattern. Since this spotted pattern is expressed as a bright and dark pattern when the surface to be inspected is an image, it becomes an obstacle to the recognition of streak defects. In particular, this spot is not always brighter or darker than the streak defect, and in that case, the defect cannot be detected accurately.

従って、本発明の主目的は、梨地状等のように微細な凹凸や明暗がまだらに存在する場合でも精度良く筋状欠陥を検出することができる筋状欠陥の検出方法を提供することにある。   Accordingly, a main object of the present invention is to provide a method for detecting a streak defect that can accurately detect a streak defect even when fine irregularities or light and dark are present in a mottled state such as a satin shape. .

本発明は、検査対象の画像において、筋状欠陥上は筋状欠陥のない直線上に比べて輝度のばらつきが小さいことを利用して上記目的を達成する。   The present invention achieves the above object by utilizing the fact that in the image to be inspected, the variation in luminance is smaller on the streak defect than on the straight line without the streak defect.

本発明検出方法は、コンピュータを用いて検査対象の画像に画像処理を施すことで検査対象の欠陥を検出する筋状欠陥の検出方法であって、以下のステップを有することを特徴とする。   The detection method of the present invention is a streak defect detection method for detecting a defect to be inspected by performing image processing on an image to be inspected using a computer, and includes the following steps.

画像の注目画素から一定画素数の線分における輝度のばらつきを注目画素から全方向にわたって求めて、最も輝度のばらつきの小さい方向の候補線分n1を抽出するステップ。   A step of obtaining a luminance variation in a line segment having a certain number of pixels from the target pixel of the image in all directions from the target pixel and extracting a candidate line segment n1 in a direction having the smallest luminance variation.

候補線分n1の先端から一定画素数の線分における輝度のばらつきを先端画素から複数方向にわたって求めて、最も輝度のばらつきの小さい方向の候補線分n2を抽出するステップ。   Obtaining a variation in luminance in a line segment having a fixed number of pixels from the tip of the candidate line segment n1 in a plurality of directions from the leading pixel, and extracting a candidate line segment n2 in the direction having the smallest luminance variation.

候補線分n2が候補線分n1に対して所定の角度内の場合には、さらに候補線分n2の先端から一定画素数の線分における輝度のばらつきを先端画素から複数方向にわたって求めて、最も輝度のばらつきの小さい方向の候補線分n3…nmの抽出を繰り返し同様に行なうステップ。   When the candidate line segment n2 is within a predetermined angle with respect to the candidate line segment n1, the luminance variation in the line segment of a certain number of pixels from the tip of the candidate line segment n2 is further obtained from the tip pixel in a plurality of directions, and the most Step of repeatedly extracting candidate line segments n3... Nm in a direction where luminance variation is small.

連続して抽出された候補線分n1〜nmまでの候補線分数mが一定数以上であれば、その複数本の候補線分n1〜nmを筋状の欠陥と判定するステップ。   A step of determining the plurality of candidate line segments n1 to nm as a streak defect if the number m of candidate line segments to consecutive candidate line segments n1 to nm is a predetermined number or more.

一般に、筋状欠陥はある程度の長さからなる細長い線状に連続している。また、検査対象の表面にまだらの明暗がある場合であっても、通常、筋状欠陥自身はほぼ一様な明度である。このことを利用し、本発明では、注目画素から一定画素数分の線分における輝度のばらつきが最小となる方向の候補線分を求め、ある程度の数の候補線分が連続する場合に、この連続する候補線分群を筋状欠陥と認識することとしている。この欠陥検出方法によれば、検査対象表面に明度がまだらに存在する場合であっても、筋状欠陥とそれ以外の正常な検査対象表面とを正確に区別し、欠陥の認識率を向上することができる。   In general, streak defects are continuous in an elongated line having a certain length. In addition, even if the surface to be inspected has mottled light and dark, the streak defect itself usually has almost uniform lightness. Utilizing this fact, the present invention obtains candidate line segments in the direction in which the luminance variation in the line segment for a certain number of pixels from the target pixel is minimized, and when a certain number of candidate line segments continue, A continuous group of candidate line segments is recognized as a streak defect. According to this defect detection method, even when lightness is present on the surface to be inspected, it is possible to accurately distinguish streak defects from other normal inspection target surfaces and improve the defect recognition rate. be able to.

以下、本発明方法をより詳しく説明する。   Hereinafter, the method of the present invention will be described in more detail.

本発明方法では、まず検査対象の画像を取得する。この画像の取得には、CCDカメラなどの公知のカメラを利用することができる。   In the method of the present invention, first, an image to be inspected is acquired. A known camera such as a CCD camera can be used to acquire this image.

得られた画像から注目画素の抽出を行う。注目画素の抽出方法としては、検査対象の画像における個々の画素を注目画素とすることが挙げられる。その場合、各注目画素に対して候補線分の抽出や各候補線分同士の角度の分析などの処理を順次行ない、最終的に全画素について同様の処理を行う。その他、注目画素の数を一定数に制限してもよい。注目画素の数を限定することで、より短時間で筋状欠陥の検出処理を行なうことができる。注目画素の限定は、検査対象の画像を二値化処理し、その二値化画像において明部(または暗部)として認識された画素を注目画素とすることが考えられる。例えば、引っかきキズがキズ以外の箇所よりも明るい(暗い)傾向の場合、二値化画像で明部(暗部)として表示された画素を注目画素とすればよい。さらには、二値化処理を行っても注目画素の数が多い場合、二値化処理する際のしきい値を変えることで、必ず一定数(例えば100個)以下に注目画素の数を限定できるようにしてもよい。   The pixel of interest is extracted from the obtained image. As a pixel-of-interest extraction method, an individual pixel in an image to be inspected is used as a pixel of interest. In that case, processing such as extraction of candidate line segments and analysis of angles between the candidate line segments is sequentially performed on each target pixel, and finally the same processing is performed on all pixels. In addition, the number of target pixels may be limited to a certain number. By limiting the number of pixels of interest, the streak defect detection process can be performed in a shorter time. It is conceivable that the pixel of interest is limited by subjecting the image to be inspected to binarization processing, and using the pixel recognized as a bright part (or dark part) in the binarized image as the pixel of interest. For example, in the case where the scratches tend to be brighter (darker) than portions other than the scratches, a pixel displayed as a bright part (dark part) in the binarized image may be set as the target pixel. Furthermore, if the number of pixels of interest is large even after binarization processing, the number of pixels of interest must be limited to a certain number (for example, 100) or less by changing the threshold value for binarization processing. You may be able to do it.

注目画素が抽出できたら、その注目画素から一定画素数分の線分における輝度のばらつきを注目画素から全方向にわたって求め、ばらつきが最も小さい候補線分n1を抽出する。候補線分は筋状欠陥の候補となる画像上の仮想線であり、後述するように、この候補線分が一定本数以上連続する場合に筋状欠陥が存在すると認識する。   When the pixel of interest can be extracted, the luminance variation in a line segment of a certain number of pixels from the pixel of interest is obtained from the pixel of interest in all directions, and the candidate line segment n1 having the smallest variation is extracted. The candidate line segment is an imaginary line on the image that is a candidate for the streak defect, and as will be described later, it is recognized that a streak defect exists when the candidate line segment continues for a certain number or more.

この線分の長さを規定する一定画素数は一画素のサイズや筋状欠陥の一般的な長さ等を考慮して任意に選択すれば良い。例えば数画素〜数十画素とすることが挙げられる。この線分の輝度のばらつきを求めるには、注目画素を中心とする放射状の複数本の線分の各々について輝度を求めておくことが好ましい。そして、線分を構成する画素の輝度のばらつきが最も小さい線分を抽出すればよい。代表的には、まず注目画素から設定方向に伸びる線分の輝度のばらつきを求め、この設定方向から順次所定角度のピッチで一定画素数分の線分における輝度のばらつきを順次求め、注目画素の周囲360°にわたって同様の処理を行えばよい。この角度ピッチは任意に選択すれば良いが、大きく設定すると筋状欠陥を見落とす可能性が高くなり、小さすぎると処理時間が長くなるので、適宜な角度を選べばよい。例えば1°ピッチで線分の輝度を求めれば良い。   The fixed number of pixels defining the length of the line segment may be arbitrarily selected in consideration of the size of one pixel, the general length of the streak defect, and the like. For example, it may be set to several pixels to several tens of pixels. In order to obtain the luminance variation of the line segment, it is preferable to obtain the luminance for each of a plurality of radial line segments centered on the target pixel. Then, it is only necessary to extract a line segment having the smallest variation in luminance of pixels constituting the line segment. Typically, first, the luminance variation of the line segment extending from the target pixel in the setting direction is obtained, and the luminance variation in the line segment for a certain number of pixels is sequentially obtained from the setting direction at a predetermined angle pitch. The same processing may be performed over the surrounding 360 °. The angle pitch may be arbitrarily selected, but if it is set large, the possibility of overlooking the streak defect increases, and if it is too small, the processing time becomes long. Therefore, an appropriate angle may be selected. For example, the luminance of the line segment may be obtained at a 1 ° pitch.

候補線分n1が抽出できれば、候補線分n1の先端を新たな注目画素として、上記と同様に次の候補線分n2を抽出する。つまり、候補線分n1の先端から一定画素数分の線分における輝度のばらつきを先端画素から全方向にわたって求め、そのばらつきが最も小さい候補線分n2を抽出する。   If the candidate line segment n1 can be extracted, the next candidate line segment n2 is extracted in the same manner as described above using the tip of the candidate line segment n1 as a new target pixel. That is, the luminance variation in the line segment for a certain number of pixels from the tip of the candidate line segment n1 is obtained in all directions from the tip pixel, and the candidate line segment n2 having the smallest variation is extracted.

上述したように、最初に抽出した注目画素に対しては、その周囲360°にわたって各線分の輝度のばらつきを求めて最初の候補線分n1を決定する必要があるが、それ以降の候補線分を求める際の注目画素に対しては、所定角度のピッチで一定画素数分の線分における輝度のばらつきを求める範囲をより限定しても良い。例えば注目画素の周囲360°にわたって各線分の輝度のばらつきを求めるのではなく、注目画素を含み候補線分と直交する線分で規定される180°の範囲にわたって所定角度のピッチで一定画素数分の線分における輝度のばらつきを求めてもよい。通常、ある候補線分に連続する次の候補線分は、ある候補線分と直交する線分よりも前方(ある候補線分の先端側)に存在するからである。   As described above, for the first pixel of interest extracted, it is necessary to determine the first candidate line segment n1 by determining the luminance variation of each line segment over 360 ° around it. For the pixel of interest when obtaining the value, the range for obtaining the luminance variation in the line segment for a certain number of pixels at a pitch of a predetermined angle may be further limited. For example, instead of calculating the luminance variation of each line segment around 360 ° around the pixel of interest, it is a fixed number of pixels at a predetermined angle pitch over a 180 ° range defined by the line segment that includes the pixel of interest and is orthogonal to the candidate line segment. Variations in luminance in the line segments may be obtained. This is because the next candidate line segment that is continuous with a certain candidate line segment usually exists in front of a line segment that is orthogonal to the certain candidate line segment (at the tip side of a certain candidate line segment).

次に、候補線分n2が候補線分n1に対して所定の角度内にあるかどうかを調べる。通常、筋状欠陥は細長く線状に伸びており、その途中にある程度の湾曲があったとしてもV型に折れ曲がるなどの極端な屈曲はない。そのため、候補線分n2が候補線分n1に対して所定の角度内にあるかを調べ、所定の角度内になければ候補線分n2と候補線分n1は筋状欠陥を構成するものではないと判断する。所定の角度は任意に選択すれば良いが、例えば±10°以内とすることが挙げられる。   Next, it is examined whether the candidate line segment n2 is within a predetermined angle with respect to the candidate line segment n1. Usually, a streak defect is elongated and linearly extended, and even if there is a certain amount of curvature in the middle, there is no extreme bending such as bending into a V shape. Therefore, it is checked whether the candidate line segment n2 is within a predetermined angle with respect to the candidate line segment n1, and if it is not within the predetermined angle, the candidate line segment n2 and the candidate line segment n1 do not constitute a streak defect. Judge. The predetermined angle may be arbitrarily selected, and may be within ± 10 °, for example.

一方、候補線分n2と候補線分n1が所定の角度内にあれば、両候補線分は筋状欠陥の可能性があると判断して、さらに次の候補線分n3の抽出を続ける。すなわち、候補線分n2の先端を新たな注目画素として、上記と同様に次の候補線分n3を抽出する。続いて、候補線分n3が候補線分n2に対して所定の角度内にあるかどうかを調べる。そして、このような処理を順次繰り返し行い、連続する候補線分n4…nmを順次求めていく。   On the other hand, if the candidate line segment n2 and the candidate line segment n1 are within a predetermined angle, it is determined that both candidate line segments may have a streak defect, and further extraction of the next candidate line segment n3 is continued. That is, the next candidate line segment n3 is extracted in the same manner as described above using the tip of the candidate line segment n2 as a new target pixel. Subsequently, it is examined whether the candidate line segment n3 is within a predetermined angle with respect to the candidate line segment n2. Then, such processes are sequentially repeated to sequentially obtain the candidate line segments n4... Nm.

この処理の繰り返しは新たな候補線分が抽出できなくなるまで行うことが好ましい。新たな候補線分が抽出できなくなる場合とは、注目画素から所定画素数分の線分が検査対象画像の輪郭に達した場合、あるいは注目画素が検査対象画像の輪郭に達した場合の他、候補線分n+1と候補線分nが所定の角度内になかった場合が挙げられる。   This process is preferably repeated until no new candidate line segment can be extracted. The case where a new candidate line segment cannot be extracted means that when a predetermined number of lines from the target pixel reaches the contour of the inspection target image, or when the target pixel reaches the contour of the inspection target image, There is a case where the candidate line segment n + 1 and the candidate line segment n are not within a predetermined angle.

そして、連続して抽出できた候補線分(候補線分n1〜候補線分nm)の数mが一定数以上であった場合に、その候補線分群を筋状欠陥であると認識する。この連続する数mは任意に設定すればよい。例えば4本以上などとすることが挙げられる。   When the number m of candidate line segments (candidate line segment n1 to candidate line segment nm) that can be extracted continuously is a certain number or more, the candidate line segment group is recognized as a streak defect. This continuous number m may be set arbitrarily. For example, it may be 4 or more.

本発明方法の検査対象は、特に限定されるものではない。粉末冶金材料に代表される梨地状の表面を有する材料も検査対象に含まれる。また、検査対象の画像も表面画像に限らず、X線透視画像など、背景に不規則な明暗が生じる画像であれば適用できる。   The inspection object of the method of the present invention is not particularly limited. Materials having a textured surface typified by powder metallurgy materials are also included in the inspection object. Further, the image to be inspected is not limited to the surface image, and can be applied as long as the image has irregular brightness on the background, such as an X-ray fluoroscopic image.

本発明の筋状欠陥の検出方法によれば、筋状欠陥上の輝度は筋状欠陥のない直線上の輝度に比べてばらつきが小さいことを利用することで、特に梨地状の表面を有する検査対象であっても正確に筋状欠陥を検出することができる。   According to the method for detecting a streak defect of the present invention, an inspection having a satin-like surface in particular is achieved by utilizing the fact that the brightness on the streak defect is less varied than the brightness on a straight line without a streak defect. Even if it is a target, a streak defect can be accurately detected.

また、注目画素を抽出する際に、検査対象の表面画像に対して二値化処理などの画像処理を施すことで注目画素数を少なくすれば、検査に要する時間を短縮することができる。   Further, when extracting the target pixel, if the number of target pixels is reduced by performing image processing such as binarization processing on the surface image to be inspected, the time required for the inspection can be shortened.

以下、本発明の実施の形態を説明する。   Embodiments of the present invention will be described below.

図1は本発明方法に利用する検査装置の機能ブロック図である。   FIG. 1 is a functional block diagram of an inspection apparatus used in the method of the present invention.

この検査装置は、図1に示すように、検査対象を撮影してその表面の画像を取得するカメラ100と、コンピュータ200Aでカメラからの画像を画像処理する画像処理装置200および画像を表示するディスプレイ300から構成される。   As shown in FIG. 1, this inspection apparatus includes a camera 100 that captures an image of an inspection object and acquires an image of the surface thereof, an image processing apparatus 200 that performs image processing on an image from the camera with a computer 200A, and a display that displays the image Consists of 300.

カメラ100はコンピュータ200Aからの指令に基づいて検査対象Wの表面画像を撮影する。ここでは、CCDカメラを用いて、検査対象Wである銅タングステンの表面画像を撮影する。銅タングステンは粉末冶金材料であり、不規則なまだらの明暗斑点を持ついわゆる梨地状の表面を有する。   The camera 100 captures a surface image of the inspection target W based on a command from the computer 200A. Here, a surface image of copper tungsten, which is the inspection object W, is taken using a CCD camera. Copper tungsten is a powder metallurgy material and has a so-called satin-like surface with irregular mottled light and dark spots.

カメラ100で撮影された画像は画像処理装置200に取り込まれる。画像処理装置200は、この表面画像を記憶する画像メモリ210、画像メモリ210から読み出した画像における検査対象の輪郭を抽出する輪郭抽出部220、画像から注目画素を抽出する注目画素抽出部230、注目画素から所定画素分の線分の輝度を抽出する線分輝度抽出部240を有する。さらに、これら各線分のうち最も輝度のばらつきの小さい線分を抽出する候補線分抽出部250、候補線分同士の角度を比較する線分角度比較部260、これら候補線分の連続数から筋状欠陥の有無を判定する筋状欠陥判定部270ならびに欠陥画像の表示指令部280を有する。そして、欠陥の検出された検査対象Wの画像は、筋状欠陥を明示してディスプレイ300上に表示される。   An image photographed by the camera 100 is taken into the image processing apparatus 200. The image processing apparatus 200 includes an image memory 210 that stores the surface image, a contour extraction unit 220 that extracts a contour of an inspection target in the image read from the image memory 210, a target pixel extraction unit 230 that extracts a target pixel from the image, It has a line segment luminance extraction unit 240 that extracts the luminance of a predetermined line segment from the pixel. Further, among these line segments, a candidate line segment extraction unit 250 that extracts a line segment having the smallest luminance variation, a line segment angle comparison unit 260 that compares the angles of the candidate line segments, and a straight line based on the continuous number of these candidate line segments. A streak defect determination unit 270 that determines the presence or absence of a line defect and a display instruction unit 280 for a defect image. Then, the image of the inspection target W in which the defect is detected is displayed on the display 300 with the streak defect clearly shown.

このような装置により筋状の欠陥を検出する手順を図2のフローチャートに基づいて説明する。その際、100〜300までの符号は図1を参照する。   A procedure for detecting a streak defect using such an apparatus will be described with reference to the flowchart of FIG. At that time, reference numerals 100 to 300 refer to FIG.

まず、コンピュータ200Aからの指令によりカメラ100は検査対象Wの表面画像を取得する(ステップS1)。取得した画像は、画像処理装置の画像メモリ210に記憶される。   First, in response to a command from the computer 200A, the camera 100 acquires a surface image of the inspection object W (step S1). The acquired image is stored in the image memory 210 of the image processing apparatus.

次に、この画像の中から試料の輪郭を輪郭抽出部220で抽出し、その輪郭座標を求めておく(ステップS2)。この輪郭座標は、後にキズの候補線分を求める処理の終了判断に利用する。   Next, the contour of the sample is extracted from the image by the contour extraction unit 220, and the contour coordinates are obtained (step S2). The contour coordinates are used later to determine the end of the process for obtaining the scratch candidate line segment.

抽出された輪郭座標から一定画素分内側の枠内を注目画素の抽出範囲として選択する(ステップS3)。この選択は注目画素抽出部230により行なう。候補線分は注目画素から一定画素分の線分を単位として求めるため、この抽出範囲内で注目画素を選択すれば、いずれの注目画素からも確実に一定画素分の線分の輝度を求めることができる。本例では1画素4.2μmのサイズで、20画素分の線分を単位として注目画素の抽出を行う。   A frame within a certain number of pixels from the extracted contour coordinates is selected as a target pixel extraction range (step S3). This selection is performed by the pixel-of-interest extraction unit 230. Since the candidate line segment is obtained from the target pixel in units of a line segment for a certain pixel, if the target pixel is selected within this extraction range, the luminance of the line segment for the certain pixel is reliably determined from any target pixel. Can do. In this example, the pixel of interest is extracted in units of 20 pixel line segments with a size of one pixel of 4.2 μm.

抽出範囲を選択したら、同じく注目画素抽出部230により、この抽出範囲の画像を二値化処理し、二値化画像から注目画素を抽出する(ステップS4)。注目画素はキズの候補線分を求める始点となる箇所である。本例では、筋状欠陥が暗部として認識される場合であり、二値化画像における黒点を注目画素とする。抽出範囲の全ての画素を注目画素としても良いが、全ての画素から候補線分を求める処理を行うと欠陥検出処理に時間を要するため、注目画素の数を限定することで処理時間を短縮化することができる。例えば、二値化処理の閾値を調整することで、注目画素数を100個に限定する。そして、100個の注目画素に順位付けし、注目画素T1〜注目画素T100の各々について順位に基づいて候補線分の抽出処理を行っていく。   When the extraction range is selected, the pixel-of-interest extraction unit 230 similarly binarizes the image in the extraction range, and extracts the pixel of interest from the binarized image (step S4). The pixel of interest is a starting point for obtaining a scratch candidate line segment. In this example, a streak defect is recognized as a dark part, and a black dot in a binarized image is set as a pixel of interest. All pixels in the extraction range may be used as the target pixel, but if the process for obtaining the candidate line segment from all the pixels is performed, the defect detection process takes time, so the processing time is shortened by limiting the number of target pixels. can do. For example, the number of pixels of interest is limited to 100 by adjusting the threshold value of the binarization process. Then, the 100 target pixels are ranked, and candidate line segment extraction processing is performed for each of the target pixel T1 to the target pixel T100 based on the order.

まず、候補線分の輝度を線分輝度抽出部240により求める。具体的には、図3に示すように、注目画素T1について、その画素から20画素分の線分(20画素線分L:L1、L2…)の輝度を求める。この図では、マトリックスを構成する正方形の一つが1画素に対応し、説明の便宜上、20画素線分Lの長さは20画素よりも少ない長さとなっている。この20画素線分Lの輝度分布は、1°ピッチで注目画素T1の周囲360°の範囲にわたって順次求め、合計360本の20画素線分L1…L360の輝度分布を抽出する(ステップS5)。   First, the line segment luminance extraction unit 240 obtains the luminance of the candidate line segment. Specifically, as shown in FIG. 3, the luminance of a line segment for 20 pixels from the pixel of interest T1 (20 pixel line segments L: L1, L2,...) Is obtained. In this figure, one of the squares constituting the matrix corresponds to one pixel, and for the convenience of explanation, the length of the 20 pixel line segment L is shorter than 20 pixels. The luminance distribution of the 20 pixel line segments L is sequentially obtained over a range of 360 ° around the target pixel T1 at a pitch of 1 °, and the luminance distributions of a total of 360 20 pixel line segments L1... L360 are extracted (step S5).

続いて、候補線分抽出部250により、360本の20画素線分の中から、最も輝度のばらつきの小さい線分をキズの候補線分n1として抽出する(ステップS6)。   Subsequently, the candidate line segment extraction unit 250 extracts the line segment having the smallest luminance variation from 360 20 pixel line segments as a scratch candidate line segment n1 (step S6).

候補線分n1を抽出したら、この候補線分の先端(注目画素と反対側端部)の画素を抽出する(ステップS7)。   When the candidate line segment n1 is extracted, a pixel at the tip (end opposite to the target pixel) of this candidate line segment is extracted (step S7).

この先端画素から20画素の範囲に試料の輪郭画素が重なるかどうかを調べる(ステップS8)。そのとき、先端画素が輪郭画素に重なれば、先端画素が試料の輪郭から候補線分の距離以下に接近していると判断し、後述するように、その時点までに抽出された候補線分の数を調べる(ステップS12へ)。   It is checked whether or not the contour pixel of the sample overlaps the range of 20 pixels from the tip pixel (step S8). At that time, if the tip pixel overlaps the contour pixel, it is determined that the tip pixel is closer than the distance of the candidate line segment from the contour of the sample, and the candidate line segment extracted up to that point as described later. Is checked (to step S12).

一方、先端画素が輪郭画素に重ならなければ、線分輝度抽出部240により、次の候補線分を求める処理を続ける。つまり、図4に示すように、先端画素E1から20画素分線分Lの輝度を求める。この20画素線分Lの輝度分布は、候補線分n1を基準として1°ピッチで注目画素の周囲359°の範囲にわたって順次求め、合計359本の20画素線分Lの輝度分布を抽出する(ステップS9)。候補線分n1を基準として先端画素E1の周囲359°の範囲について輝度分布を抽出するのは、候補線分n1と重複する線分を次の候補線分n2として認識しないためである。そして、359本の20画素線分の中から、最も輝度のばらつきの小さい線分をキズの候補線分n2として抽出する(ステップS10)。   On the other hand, if the leading edge pixel does not overlap the contour pixel, the line segment luminance extraction unit 240 continues the process of obtaining the next candidate line segment. That is, as shown in FIG. 4, the luminance of the 20-pixel segment L from the leading edge pixel E1 is obtained. The luminance distribution of the 20 pixel line segment L is sequentially obtained over a range of 359 ° around the pixel of interest at a pitch of 1 ° with reference to the candidate line segment n1, and a total of 359 luminance distributions of the 20 pixel line segment L are extracted ( Step S9). The reason why the luminance distribution is extracted in the range of 359 ° around the leading edge pixel E1 with the candidate line segment n1 as a reference is that the line segment overlapping the candidate line segment n1 is not recognized as the next candidate line segment n2. Then, a line segment with the smallest luminance variation is extracted from the 359 20 pixel line segments as a scratch candidate line segment n2 (step S10).

次に、線分角度比較部260で、候補線分n2と候補線分n1の角度θ(図4参照)を検出し、所定範囲内であるかどうかを調べる(ステップS11)。本例では、ある候補線分xとその前の候補線分x-1との角度θが10°以下かどうかを判断する。これは、筋状の欠陥は、急激に屈曲することなくある程度直線状に連続する特性を利用している。つまり、候補線分xと候補線分x-1の角度が10°以内にあれば、後の候補線分xは筋状欠陥の候補であると認識する。その場合、後の候補線分の先端画素E2を抽出し、ステップS7〜ステップS11までを繰り返して連続する候補線分を順次求めていく。つまり、候補線分n3の先端画素から20画素線分の輝度分布を求めて次の候補線分n4を抽出し、順次これを繰り返して候補線分n3〜nmを抽出する。これ対して、候補線分xと候補線分x-1の角度が10°以内になければ、その時点までに抽出された候補線分の数mを調べる。   Next, the line segment angle comparison unit 260 detects the angle θ (see FIG. 4) between the candidate line segment n2 and the candidate line segment n1, and checks whether it is within a predetermined range (step S11). In this example, it is determined whether or not the angle θ between a certain candidate line segment x and the previous candidate line segment x−1 is 10 ° or less. This utilizes the characteristic that a line-like defect continues linearly to some extent without sudden bending. That is, if the angle between the candidate line segment x and the candidate line segment x-1 is within 10 °, the subsequent candidate line segment x is recognized as a candidate for a streak defect. In that case, the leading end pixel E2 of the subsequent candidate line segment is extracted, and successive candidate line segments are sequentially obtained by repeating steps S7 to S11. That is, the luminance distribution of 20 pixel lines is obtained from the leading pixel of the candidate line segment n3, the next candidate line segment n4 is extracted, and this is repeated sequentially to extract candidate line segments n3 to nm. On the other hand, if the angle between the candidate line segment x and the candidate line segment x-1 is not within 10 °, the number m of candidate line segments extracted up to that point is examined.

このように、候補線分の先端画素から20画素の範囲に試料の輪郭が重なった場合または候補線分xと候補線分x-1の角度が10°以内になかった場合は、それぞれの時点までに抽出された候補線分mの数が規定値以上であるかどうかを筋状欠陥判定部270で比較する(ステップS12)。ここでは、m<4の場合に筋状の欠陥がないと判断し(ステップS13)、m≧4の場合に筋状の欠陥があると判断する(ステップS14)。   In this way, if the sample outline overlaps the range of 20 pixels from the tip pixel of the candidate line segment, or if the angle between the candidate line segment x and the candidate line segment x-1 is not within 10 °, each time point The streak defect determining unit 270 compares whether or not the number of candidate line segments m extracted so far is greater than or equal to a specified value (step S12). Here, it is determined that there is no streak defect when m <4 (step S13), and it is determined that there is a streak defect when m ≧ 4 (step S14).

つまり、候補線分の連続数が3本以下の場合は、検査対象の表面に筋状のキズはないが梨地状表面の明暗のため偶然に輝度のばらつきが最小である線分がほぼ直線状に連続した場合であると認識し、候補線分の連続数が4本以上の場合は、これだけの数の候補線分が直線状に連続している以上、筋状のキズがあると認識する。   In other words, if the number of consecutive candidate line segments is 3 or less, the surface of the object to be inspected has no streak-like scratches, but the line segment with the smallest variation in brightness by chance due to the light and dark surface of the textured surface is almost linear If the number of consecutive candidate line segments is 4 or more, recognize that there are streak scratches as long as this number of candidate line segments continue in a straight line. .

このステップS13でキズがないと判断されれば、未処理の注目画素があるかどうかを調べ(ステップS15)、未処理の注目画素があれば、その注目画素に対してステップS5〜ステップS12までの処理を行う。つまり、注目画素T1について筋状のキズがないと判定されれば、次に注目画素T2について同様の欠陥検出処理を行い、順次注目画素T3…と同様の処理を行う。   If it is determined in step S13 that there is no scratch, it is checked whether there is an unprocessed pixel of interest (step S15). If there is an unprocessed pixel of interest, steps S5 to S12 are performed for the pixel of interest. Perform the process. That is, if it is determined that there is no streak-like defect for the target pixel T1, the same defect detection process is performed for the target pixel T2, and the same process as the target pixel T3 is sequentially performed.

一方、ステップS14でキズがあると判断されれば、そのキズを検査対象の表面画像と共にディスプレイに表示するように欠陥画像表示指令部280から指令を出力する(ステップS16)。この指令により、図5に示すように、検査対象Wの画像上に、筋状の欠陥Dが黒いラインとして表示される。   On the other hand, if it is determined in step S14 that there is a scratch, a command is output from the defect image display command unit 280 so that the scratch is displayed on the display together with the surface image to be inspected (step S16). By this command, as shown in FIG. 5, the streak defect D is displayed as a black line on the image of the inspection object W.

そして、全ての注目画素からの検査でキズがないと判断されるか、いずれかの注目画素からの検査でキズがあると判定されれば、その検査対象についての欠陥検査を終了する。   If it is determined that there is no flaw in the inspection from all the target pixels, or if it is determined that there is a flaw in the inspection from any of the target pixels, the defect inspection for the inspection target is ended.

本発明方法では、複数の注目画素を始点として、その各々に対して順次キズの有無を検査するが、いずれかの注目画素を始点とした検査でキズがあると判定されれば、その検査対象はキズが見つかった時点で不合格品となるため、その時点で未処理の注目画素が残っていても検査処理を終了して構わない。もちろん、キズを検出しても検査を続行し、複数のキズを検出することも可能である。   In the method of the present invention, a plurality of target pixels are used as starting points, and the presence or absence of scratches is sequentially inspected for each of the target pixels. Since a defective product is found when a scratch is found, the inspection process may be terminated even if an unprocessed target pixel remains at that time. Of course, even if a scratch is detected, the inspection can be continued and a plurality of scratches can be detected.

このように、本発明方法によれば、梨地状の表面を有する検査対象の筋状欠陥を高精度に検出することができる。   Thus, according to the method of the present invention, it is possible to detect a streak defect to be inspected having a satin-like surface with high accuracy.

本発明は、セラミックスや金属などの粉末冶金材料など、表面が梨地状になっている材料における筋状欠陥の検出に利用することができる。特に、ヒートシンク材料に用いられる銅タングステンの表面欠陥の検出に好適に利用できる。   INDUSTRIAL APPLICABILITY The present invention can be used for detecting streak defects in a material having a textured surface, such as a powder metallurgy material such as ceramics or metal. In particular, it can be suitably used for detecting surface defects of copper tungsten used as a heat sink material.

本発明方法に利用する検査装置の機能ブロック図である。It is a functional block diagram of the test | inspection apparatus utilized for this invention method. 本発明方法の処理手順を示すフローチャートである。It is a flowchart which shows the process sequence of this invention method. 注目画素から候補線分を見出す手順の説明図である。It is explanatory drawing of the procedure which finds a candidate line segment from an attention pixel. ある候補線分から次の候補線分を見出していく手順の説明図である。It is explanatory drawing of the procedure which finds the next candidate line segment from a certain candidate line segment. 筋状の欠陥があった検査対照の画像を示す模式図である。It is a schematic diagram which shows the image of the test | inspection control | contrast with a streaky defect.

符号の説明Explanation of symbols

100 カメラ
200 画像処理装置 200A コンピュータ 210 画像メモリ
220 輪郭抽出部 230 注目画素抽出部 240線分輝度抽出部
250 候補線分抽出部 260 線分角度比較部 270 筋状欠陥判定部
280 欠陥画像表示指令部
300 ディスプレイ
T 注目画素 L 20画素線分 n 候補線分 D 欠陥
100 cameras
200 Image processor 200A Computer 210 Image memory
220 Outline extraction unit 230 Pixel of interest extraction unit 240 line segment luminance extraction unit
250 Candidate line segment extraction unit 260 Line segment angle comparison unit 270 Streak defect determination unit
280 Defect image display command section
300 displays
T Target pixel L 20 pixel line segment n Candidate line segment D Defect

Claims (2)

コンピュータを用いて検査対象の画像に画像処理を施すことで検査対象の欠陥を検出する筋状欠陥の検出方法であって、
前記画像の注目画素から一定画素数の線分における輝度のばらつきを注目画素から全方向にわたって求めて、最も輝度のばらつきの小さい方向の候補線分n1を抽出するステップと、
候補線分n1の先端から一定画素数の線分における輝度のばらつきを先端画素から複数方向にわたって求めて、最も輝度のばらつきの小さい方向の候補線分n2を抽出するステップと、
候補線分n2が候補線分n1に対して所定の角度内の場合には、さらに候補線分n2の先端から一定画素数の線分における輝度のばらつきを先端画素から複数方向にわたって求めて、最も輝度のばらつきの小さい方向の候補線分n3…nmの抽出を繰り返し同様に行なうステップと、
連続して抽出された候補線分n1〜nmまでの候補線分数mが一定数以上であれば、それら複数本の候補線分n1〜nmを筋状の欠陥と判定するステップとを有することを特徴とする筋状欠陥の検出方法。
A streak defect detection method for detecting a defect to be inspected by performing image processing on an image to be inspected using a computer,
Obtaining a luminance variation in a line segment of a certain number of pixels from the target pixel of the image in all directions from the target pixel, and extracting a candidate line segment n1 in a direction having the smallest luminance variation;
Obtaining a variation in luminance in a line segment of a certain number of pixels from the tip of the candidate line segment n1 over a plurality of directions from the tip pixel, and extracting a candidate line segment n2 in the direction with the smallest luminance variation;
When the candidate line segment n2 is within a predetermined angle with respect to the candidate line segment n1, the luminance variation in the line segment of a certain number of pixels from the tip of the candidate line segment n2 is further obtained from the tip pixel in a plurality of directions, and the most A step of repeatedly extracting candidate line segments n3... Nm in a direction where the luminance variation is small;
If the number m of candidate line segments to consecutively extracted candidate line segments n1 to nm is equal to or greater than a predetermined number, the step of determining the plurality of candidate line segments n1 to nm as streak defects is included. A method for detecting a characteristic streak defect.
予め表面画像に二値化処理を施して、筋状の欠陥の候補点を抽出するステップを有し、
この候補点を注目画素とすることを特徴とする請求項1に記載の筋状欠陥の検出方法。
A step of binarizing the surface image in advance to extract candidate points for streak defects;
2. The streak defect detection method according to claim 1, wherein the candidate point is set as a target pixel.
JP2004212293A 2004-07-20 2004-07-20 Method for detecting wavy defect Pending JP2006030093A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2004212293A JP2006030093A (en) 2004-07-20 2004-07-20 Method for detecting wavy defect

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2004212293A JP2006030093A (en) 2004-07-20 2004-07-20 Method for detecting wavy defect

Publications (1)

Publication Number Publication Date
JP2006030093A true JP2006030093A (en) 2006-02-02

Family

ID=35896620

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2004212293A Pending JP2006030093A (en) 2004-07-20 2004-07-20 Method for detecting wavy defect

Country Status (1)

Country Link
JP (1) JP2006030093A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2019066263A (en) * 2017-09-29 2019-04-25 清水建設株式会社 Crack detector, crack detection method, and computer program
JP2020201616A (en) * 2019-06-07 2020-12-17 東京計器株式会社 Image processing device, image processing program, and image processing method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2019066263A (en) * 2017-09-29 2019-04-25 清水建設株式会社 Crack detector, crack detection method, and computer program
JP2020201616A (en) * 2019-06-07 2020-12-17 東京計器株式会社 Image processing device, image processing program, and image processing method

Similar Documents

Publication Publication Date Title
JP5672480B2 (en) Apparatus and method for determining shape of terminal end of bead
JP4150390B2 (en) Appearance inspection method and appearance inspection apparatus
US20080175466A1 (en) Inspection apparatus and inspection method
JP6922539B2 (en) Surface defect determination method and surface defect inspection device
US6304323B1 (en) Method for detecting defect in bottle
US20220011241A1 (en) Surface-defect detecting method, surface-defect detecting apparatus, steel-material manufacturing method, steel-material quality management method, steel-material manufacturing facility, surface-defect determination model generating method, and surface-defect determination model
WO2017071406A1 (en) Method and system for detecting pin of gold needle element
WO2022153852A1 (en) Image analysis device, image analysis method, and program
JP4322230B2 (en) Surface defect inspection apparatus and surface defect inspection method
CN110426395B (en) Method and device for detecting surface of solar EL battery silicon wafer
KR101862310B1 (en) Apparatus and Method for Detecting Mura Defects
JP2009074952A (en) Visual inspection method
JP2007248051A (en) Method for inspecting defect of object surface
JP2009109243A (en) Device for inspecting resin sealing material
JP3988440B2 (en) Appearance inspection method and appearance inspection apparatus
JP2002243655A (en) Method and equipment for visual inspection of electronic component
JP2006030093A (en) Method for detecting wavy defect
JP4015436B2 (en) Gold plating defect inspection system
JP3159063B2 (en) Surface defect inspection equipment
JP2006145415A (en) Surface inspection method and device
JP4349960B2 (en) Surface defect inspection equipment
KR101053779B1 (en) Metal mask inspection method of display means
JP2003247954A (en) Defect detection method for round body circumference
JP2004170374A (en) Apparatus and method for detecting surface defect
JP2005031952A (en) Image processing inspection method and image processing inspection device