JPH05273143A - Image quantization processor - Google Patents

Image quantization processor

Info

Publication number
JPH05273143A
JPH05273143A JP7089892A JP7089892A JPH05273143A JP H05273143 A JPH05273143 A JP H05273143A JP 7089892 A JP7089892 A JP 7089892A JP 7089892 A JP7089892 A JP 7089892A JP H05273143 A JPH05273143 A JP H05273143A
Authority
JP
Japan
Prior art keywords
image
light
subject
quantization
elements
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
JP7089892A
Other languages
Japanese (ja)
Inventor
Yuichi Matsui
祐一 松居
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.)
Kobe Steel Ltd
Original Assignee
Kobe Steel 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 Kobe Steel Ltd filed Critical Kobe Steel Ltd
Priority to JP7089892A priority Critical patent/JPH05273143A/en
Publication of JPH05273143A publication Critical patent/JPH05273143A/en
Pending legal-status Critical Current

Links

Landscapes

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

Abstract

PURPOSE:To detect the surface image of a specimen at a high speed by moving a quantization processing means linearly arranged with many quantizing elements relatively to the specimen in the direction perpendicular to the arrangement direction of the quantizing elements. CONSTITUTION:After the position detection by an image extracting device 8, the reflected light caused by a scar is transmitted from either one light modulating element 3 of multiple position detecting elements 4a. The distribution of the elements 3 in shape recognizing elements 5 in the light transmission state is symmetrical on the right and left against the face containing the elements 3 in the elements 4a in the light transmission state, i.e., the face parallel with the scanning direction, when observed from above an image processor. The tip of the scar image is necessarily located at the upper center of an extracting region, and the position of the scar image in the extracting region can be specified at this level. The shape of the scar can be classified at a high speed with a high recognition factor.

Description

【発明の詳細な説明】Detailed Description of the Invention

【0001】[0001]

【産業上の利用分野】本発明は,鉄,非鉄金属,合成樹
脂,ガラス,半導体などの被検体の表面を2次元的に走
査する走査器において,上記被検体の表面から測定され
る反射光に基づいて,上記光検体表面の光学的に検出さ
れる画像を量子化処理することのできる画像量子化処理
装置に関するものである。
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a scanner for two-dimensionally scanning the surface of an object such as iron, non-ferrous metal, synthetic resin, glass and semiconductor, and the reflected light measured from the surface of the object. The present invention relates to an image quantization processing device capable of performing a quantization process on the image optically detected on the surface of the optical specimen based on the above.

【0002】[0002]

【従来の技術】従来,この種の画像量子化処理装置を応
用した表面検査装置としては,例えば第34回システム
制御情報学会研究発表講演会論文集(ISCIE,第1
03頁乃至第104頁,’90−5)に開示された装置
が挙げられる。上記開示の表面検査装置1a を図5に示
す。同図に示す表面検査装置1a によれば,画像センサ
部が,走行する鋼板等の被検体表面に光源からのレーザ
光または白色光を照射し,上記被検体表面からの反射光
を主走査方向に画素毎に順次走査する受光部により受光
し,反射光の光強度をその強弱に応じた電気信号に変換
し出力する。そして,画像処理部は,図6に示すよう
に,上記画像センサ部からの画素毎の電気信号に基づい
て,例えばある閾値を用いて画素毎に0(白い部分の画
素)または1(ハッチングされた画素)の値に2値化さ
れた2次元の量子化画像を作成する。そして,この量子
化画像をもとに,上記被検体表面の疵や欠陥と目される
欠陥画像の切り出し処理を行い,上記切り出された欠陥
画像に関する,例えば長さh,幅w,面積等の特徴量を
抽出する。引き続いて,判定部は,画像処理部からの欠
陥画像の特徴量に基づいて,上記被検体表面の疵や欠陥
の種別やこれらの等級を判定する。上記判定部では,上
記したような判定を行うために,例えば逆伝播重み学習
型,いわゆるバックプロパゲーション型のニューラルネ
ットワークが用いられており,上記各種の特徴量に対応
して予め用意された学習用入力データにより学習済のニ
ューラルネットワークに,上記画像処理部により演算さ
れた欠陥画像に関する特徴量がデータ入力されて演算さ
れる。このような表面検査装置は,判定部における演算
を高速に行うことができる利点を有している。上記従来
例では,量子化処理の後に切り出し処理を行っている
が,その逆を行うことも当然可能である。本発明におい
てもいずれを先に行うこともできる。
2. Description of the Related Art Conventionally, as a surface inspection apparatus to which this kind of image quantization processing apparatus is applied, for example, the 34th System Control Information Society Research Presentation Lecture Proceedings (ISCIE, 1st
The apparatus disclosed in pages 03 to 104, '90 -5) can be mentioned. The surface inspection apparatus 1 a disclosed above is shown in FIG. According to the surface inspection apparatus 1a shown in the figure, the image sensor unit irradiates the surface of the object to be inspected, such as a running steel plate, with laser light or white light from a light source, and performs main scanning with the light reflected from the surface of the object to be inspected. Light is received by a light receiving unit that sequentially scans each pixel in the direction, and the light intensity of the reflected light is converted into an electric signal according to its intensity and output. Then, as shown in FIG. 6, the image processing unit uses 0 (white pixel) or 1 (hatched) for each pixel based on the electric signal from the image sensor unit for each pixel, for example, using a certain threshold value. Pixel) to create a two-dimensional quantized image. Then, based on this quantized image, a defect image which is regarded as a flaw or a defect on the surface of the subject is cut out, and the cut out defect image, for example, the length h, the width w, the area, etc. Extract feature quantities. Subsequently, the determination unit determines the types of flaws and defects on the surface of the subject and their grades based on the feature amount of the defect image from the image processing unit. In the determination unit, for example, a back-propagation weight learning type, a so-called back propagation type neural network is used to perform the determination as described above, and learning prepared in advance corresponding to the various feature amounts is performed. The feature amount related to the defect image calculated by the image processing unit is input as data to the neural network that has been learned from the input data for calculation. Such a surface inspection apparatus has an advantage that the calculation in the determination unit can be performed at high speed. In the above-mentioned conventional example, the cutout process is performed after the quantization process, but it is naturally possible to perform the reverse process. In the present invention, either of them can be performed first.

【0003】[0003]

【発明が解決しようとする課題】上述のように上記従来
の表面検査装置1a では,画像処理部において2次元の
量子化画像が作成される際に,全ての画素について個々
の画素毎に上記閾値処理を適用し,量子化(2値化)処
理を行う必要がある。上記したように,全ての画素につ
いて,量子化処理を行ったり,或いは欠陥画像の切り出
し処理を実行するのでは,上記画像処理部における演算
時間を多く必要とし,特に高速で走行する被検体の表面
を正確に検査しようとする場合には,適切な方法とは言
い難い。他方,上記量子化画像全体の大きさが一定の場
合に画像の解像度を向上させようとしたり,或いは画素
の大きさが一定の場合に量子化画像全体を広くしようと
いった画像処理機能の向上化を図ろうとする場合には,
演算対象となる画素数を多くしなければならない。この
様な場合,従来装置ではなおさら画像処理部における演
算に時間がかかり過ぎるといった問題があった。本発明
は,上記従来の問題点に鑑みてなされたものであって,
被検体の表面画像の検出を高速化することのできる画像
量子化処理装置の提供を目的とする。
In [0006] the conventional surface inspection apparatus as described above 1 a, when the image processing unit is a two-dimensional quantized image is created, the for all the pixels for each individual pixel It is necessary to apply threshold processing and perform quantization (binarization) processing. As described above, performing the quantization process or the defect image cutout process for all the pixels requires a large amount of calculation time in the image processing unit, and in particular, the surface of the subject running at a high speed. It is hard to say that it is an appropriate method when trying to accurately inspect. On the other hand, it is necessary to improve the image processing function such that the resolution of the image is improved when the size of the entire quantized image is constant, or the entire quantized image is widened when the size of the pixel is constant. When trying to plot,
The number of pixels to be calculated must be increased. In such a case, the conventional apparatus has a problem that the calculation in the image processing unit takes much longer. The present invention has been made in view of the above conventional problems,
An object of the present invention is to provide an image quantization processing device capable of speeding up detection of a surface image of a subject.

【0004】[0004]

【課題を解決するための手段】上記目的を達成するため
に,本発明が採用する主たる手段は,その要旨とすると
ころが,被検体表面を露光走査して得た被検体表面画像
を量子化処理する画像量子化処理装置において,上記被
検体からの光線の通過経路に設けられ,被検体からの光
線の光量に応じて量子化された信号を出力する微小面積
の量子化素子を直線状に多数配置してなる量子化処理手
段を上記量子化素子の配置方向に直角の方向に被検体に
対して相対移動させることを特徴とする画像量子化処理
装置である。尚,この発明における量子化素子及び量子
化処理手段は以下に述べる実施例に示された構成に限定
されず,上記の各機能を達成するあらゆる構成を含むも
のである。
In order to achieve the above object, the main means adopted by the present invention is the gist of the present invention, in which a subject surface image obtained by exposing and scanning the subject surface is quantized. In the image quantization processing device, the number of linear quantization elements provided in the passage of the light ray from the subject and outputting a quantized signal according to the light quantity of the light ray from the subject is linearly increased. In the image quantization processing apparatus, the arranged quantization processing means is moved relative to the subject in a direction perpendicular to the arrangement direction of the quantization elements. The quantizing element and the quantizing processing means in the present invention are not limited to the structures shown in the embodiments described below, but include any structure that achieves each of the above functions.

【0005】[0005]

【作用】被検体からの所定光量以上の光が量子化素子に
当たると,その素子から量子化された信号が出力され
る。上記量子化素子は平面的に多数配置されているので
上記量子化された信号群は被検体表面の量子化された画
像を表すことになる。この装置では,通常行われるよう
な各素子への駆動電圧の走査を行う必要が全くなく,量
子化素子に被検体表面からの光が入射した瞬間に,量子
化処理を行うことができ,極めて高速動作の画像処理装
置が実現できる。
When the quantizing element is exposed to a predetermined amount of light or more from the subject, the element outputs a quantized signal. Since the plurality of quantizing elements are arranged in a plane, the quantized signal group represents a quantized image of the surface of the subject. In this device, it is not necessary to scan the drive voltage to each element as is usually done, and the quantization processing can be performed at the moment when the light from the surface of the subject is incident on the quantization element, A high-speed image processing device can be realized.

【0006】[0006]

【実施例】以下,添付図面を参照して,本発明を具体化
した実施例につき説明し,本発明の理解に供する。な
お,以下の実施例は,本発明を具体化した一例であっ
て,本発明の技術的範囲を限定する性格のものではな
い。また以下の実施例は2次元表面の画像検出に関する
ものであるが,本発明は一次元的な検出についても当然
適用可能である。ここに図1は本発明の適用例である欠
陥検査装置の構成要素としての画像切り出し装置8を示
す配線ブロック図,図2は本発明の一実施例に係る量子
化処理装置13を示す配線ブロック図,図3は上記画像
処理化処理装置を用いて被検体表面の表面欠陥画像など
を切り出して量子化処理する状態を示す概念図,図4は
表面欠陥画像と画像処理装置の関係を示す平面図であ
る。
Embodiments of the present invention will be described below with reference to the accompanying drawings to provide an understanding of the present invention. In addition, the following embodiments are examples of embodying the present invention and are not intended to limit the technical scope of the present invention. Further, although the following embodiments relate to image detection of a two-dimensional surface, the present invention is naturally applicable to one-dimensional detection. 1 is a wiring block diagram showing an image clipping device 8 as a component of a defect inspection apparatus as an application example of the present invention, and FIG. 2 is a wiring block showing a quantization processing device 13 according to an embodiment of the present invention. 3 and 4 are conceptual views showing a state in which a surface defect image or the like on the surface of a subject is cut out and quantized by using the image processing device, and FIG. 4 is a plane showing a relationship between the surface defect image and the image processing device. It is a figure.

【0007】本実施例に係る画像量子化処理装置13と
共に欠陥検出装置に用いられる画像切り出し装置8は,
図1に示すように,受光によって電気的抵抗が変化する
受光素子2a と印加電圧によって光透過率が変化する光
変調素子3とを直列に接続してなる位置検知素子4
a と,印加電圧によって光透過率が変化する光変調素子
3を複数個直列に接続してなる形状認識素子5とからな
り,該位置検知素子4a と該形状認識素子5とを直列に
接続して構成された複数個の画像切り出し素子6を並列
に接続し,該位置検知素子4a と該形状認識素子5との
各接続点において隣あう該接続点を抵抗7を有した配線
で接続した回路から構成されている。また上記画像切り
出し装置と組み合わせて用いられる量子化処理装置13
は,受光によって電気抵抗が変化する受光素子2b とあ
る一定の閾値以上の印加電圧によって発光する発光素子
9とを直列に接続してなる量子化素子10を並列に接続
して構成された量子化回路11と,受光によって電気抵
抗が変化する受光素子2b からなる位置検知素子4b
並列に接続して構成された位置検知回路12を,直列に
接続して構成されている。図3に示す如く,被検体14
の直上に上記画像切り出し装置8を設置し,その上に上
記量子化処理装置13を設置して上記量子化素子の配列
方向に直角に移動させると,露光走査器Lからの光の被
検体14表面での正反射光は,上記画像切り出し装置8
と上記量子化処理装置13に入射せず,被検体14表面
に異常部が存在した場合に異常部における乱反射光のみ
が,上記画像切り出し装置8と上記量子化処理装置13
に入射する。このような入射光が得られるように,画像
切り出し装置8,量子化処理装置13を有する画像処理
装置Aの配設位置が決められている。
The image cutout device 8 used in the defect detection device together with the image quantization processing device 13 according to the present embodiment is
As shown in FIG. 1, a position detecting element 4 is formed by connecting in series a light receiving element 2 a whose electric resistance changes by receiving light and a light modulating element 3 whose light transmittance changes by an applied voltage.
a and a shape recognition element 5 in which a plurality of light modulation elements 3 whose light transmittance changes according to an applied voltage are connected in series, and the position detection element 4 a and the shape recognition element 5 are connected in series. A plurality of image cutout elements 6 configured in parallel are connected in parallel, and the adjacent connection points of the position detection element 4 a and the shape recognition element 5 are connected by a wiring having a resistor 7. It is composed of a circuit. Further, a quantization processing device 13 used in combination with the above-described image cutting device
Is a quantum device constructed by connecting in parallel a quantizing device 10 formed by connecting in series a light receiving device 2 b whose electric resistance changes by receiving light and a light emitting device 9 emitting light by an applied voltage above a certain threshold value. And a position detection circuit 12 formed by connecting in parallel a position detection element 4 b composed of a light receiving element 2 b whose electric resistance changes by receiving light. As shown in FIG.
When the image cut-out device 8 is installed directly above, and the quantization processing device 13 is installed on the image cut-out device 8 and is moved at a right angle to the array direction of the quantization elements, the subject 14 of the light from the exposure scanner L is detected. The regular reflection light on the surface is the image cutting device 8 described above.
When the abnormal portion exists on the surface of the subject 14 without entering the quantization processing device 13, only diffused light in the abnormal portion is generated by the image cutting device 8 and the quantization processing device 13.
Incident on. The arrangement position of the image processing device A having the image clipping device 8 and the quantization processing device 13 is determined so that such incident light can be obtained.

【0008】この実施例では,画像切り出し装置8内の
位置検知素子4a は100個設け,形状認識素子5内の
光変調素子3は4個ずつ100列形成された。また,量
子化処理装置13内の位置検知素子4b は100個設け
られ,量子化素子10は4個ずつ100列形成された。
上記位置検知回路12が本発明の量子化素子に,また量
子化処理装置13が量子化処理手段にそれぞれ該当す
る。そして,量子化処理装置13内の各位置検知素子4
b が,画像切り出し装置8内の各位置検知素子4a 内の
光変調素子3の真上になるように配置し,量子化処理装
置13内の各量子化素子10内の受光素子2b が,画像
切り出し装置8内の各形状認識素子5内の各光変調素子
3の真上になるように配置した。そして図1のa端子と
b端子との間にa端子側が正となるように約5Vの電圧
を印加し,図2のc端子に+5Vの電圧を印加した状態
で,図3に示した様に,定常光を照射しながら被検体1
4である鉄板を,画像処理装置Aの下部に通過させた。
その際に,鉄板表面が平滑であれば,正反射光は表面検
査装置には照射されない。
[0008] In this embodiment, the position sensing element 4 a of the image clipping device 8 is provided 100, the light modulation device 3 of shape recognition element 5 is formed by four 100 columns. Further, 100 position detecting elements 4b were provided in the quantization processing device 13, and 100 rows of 4 quantizing elements 10 were formed.
The position detection circuit 12 corresponds to the quantization element of the present invention, and the quantization processing device 13 corresponds to the quantization processing means. Then, each position detection element 4 in the quantization processing device 13
b is arranged so as to be directly above the light modulation element 3 in each position detection element 4 a in the image clipping device 8, and the light receiving element 2 b in each quantization element 10 in the quantization processing device 13 is , Are arranged so as to be directly above the respective light modulation elements 3 in the respective shape recognition elements 5 in the image clipping device 8. Then, a voltage of about 5V is applied between the a terminal and the b terminal of FIG. 1 so that the a terminal side is positive, and a voltage of + 5V is applied to the c terminal of FIG. 2, as shown in FIG. Subject 1 while irradiating constant light
The iron plate of No. 4 was passed under the image processing apparatus A.
At that time, if the surface of the iron plate is smooth, the specular reflection light is not applied to the surface inspection device.

【0009】鉄板表面にキズがあれば,鉄板の移動方向
に対するキズの先端部分で光が乱反射され,反射光の一
部が画像切り出し装置8下部から画像切り出し装置8の
位置検知素子4a に入射される。その結果,反射光の入
射した位置検知素子4a 内の受光素子2a の電気抵抗が
減少し,この受光素子2a と直列に接続された光変調素
子3に印加される電圧が増大し,この光変調素子3は光
透過状態になる。ゆえに,反射光は,この光変調素子3
を透過して,画像切り出し装置8の上部に設置された量
子化処理装置13内の1個の位置検知素子4b に入射
し,量子化処理装置13を作動させる。すなわち,画像
切り出し装置8が作動するときには,量子化処理装置1
3もほぼ同時に作動する。画像切り出し装置8内の1個
の位置検知素子4a 内の受光素子2a に反射光が入射す
ると,位置検知素子4a に印加される電圧が減少し,逆
に,形状認識素子5に印加される電圧は増大する。さら
に、反射光の照射された位置検知素子4aから,遠い位
置にある形状認識素子5ほど配線抵抗が大きくなり,そ
の結果,それぞれの形状認識素子5に印加される電圧
は,遠い位置にある形状認識素子5ほど小さくなる。す
なわち,反射光の照射された位置検知素子4a から,遠
い位置にある形状認識素子5内の光変調素子3には,閾
値以上の電圧が印加されず,光吸収状態になる。
[0009] If a flaw on the surface of the iron plate, the irregular reflection light at the tip portion of the flaw with respect to the moving direction of the steel plate, a portion of the reflected light is incident from the image cutout unit 8 lower the position sensing element 4 a of the image cutout unit 8 To be done. As a result, the electrical resistance of the light receiving element 2 a of the incident of the reflected light position detecting element in 4 a is reduced, the voltage applied to the light receiving element 2 a and the optical modulation element 3 connected in series is increased, The light modulation element 3 is in a light transmitting state. Therefore, the reflected light is reflected by the light modulator 3
After passing through, the light enters the one position detection element 4 b in the quantization processing device 13 installed on the upper part of the image clipping device 8 and operates the quantization processing device 13. That is, when the image cropping device 8 operates, the quantization processing device 1
3 also operates almost at the same time. When the reflected light is incident on the light receiving element 2 a in one position detecting element 4 a in the image clipping device 8, the voltage applied to the position detecting element 4 a is decreased, and conversely, it is applied to the shape recognition element 5. The applied voltage increases. Moreover, from the irradiated position sensing element 4 a of the reflected light, the wiring resistance as shape recognition element 5 is increased in the distant position, as a result, the voltage applied to each of the shape recognition element 5, located far The shape recognition element 5 becomes smaller. That is, from the irradiated position sensing element 4 a of the reflected light, the shape recognition element light modulation element 3 in 5 located far, not higher than the threshold voltage is applied, the light-absorbing state.

【0010】本実施例の場合,距離の単位として,図1
における隣あう形状認識素子5間の間隔を採用すると,
反射光の照射された位置検知素子4a から,距離2単位
よりも大きい範囲にある形状認識素子5内の光変調素子
3は,すべて光吸収状態であった。その結果,キズによ
る反射光が,画像切り出し装置8内の1個の位置検知素
子4a に入射したことによって,距離2単位以下にある
形状認識素子5のみを通して,キズの全体または一部か
らの反射光を透過させることができた。このことによ
り,キズ画像の場所を認識しキズ画像の一部または全体
を切り出すことができた。さらに、画像切り出し装置8
を透過してきた反射光は,量子化処理装置13に入射さ
れる。すなわち,画像切り出し装置8内で光透過状態に
ある光変調素子3の真上にある量子化処理装置13内の
受光素子2b にのみ,反射光が入射する。その結果,距
離の単位として図2における量子化処理装置13内の隣
あう位置検知素子4b 間の間隔を採用すると,量子化処
理装置13内の反射光の入射した1個の位置検知素子4
b (受光素子2b )の電気抵抗と,その位置検知素子4
b からの距離(2単位)以内にある位置検知素子4b
直列に接続された量子化素子10内の受光素子2b の電
気抵抗は,いずれも減少する。本実施例においては,量
子化処理装置13内の位置検知素子4a内の1個と,量
子化回路11内の20個からなる合計21個の受光素子
b の電気抵抗が減少した。その結果,電気抵抗が減少
した受光素子2b に直列に接続された発光素子9に印加
される電圧が増大する。そして,印加電圧が閾値以上に
なった発光素子9のみが発光する。すなわち、閾値を基
準にした2値化(量子化)処理が行われる。以上の実施
例においては,CRTやCCD,液晶などで行われてい
るような,各画素に印加される電圧を走査する必要が全
くなく、光の並列性をそのまま利用しているために,キ
ズの場所の認識,切り出し,量子化処理をほぼ同時に行
うことができ,高速処理ができる。このようにして得ら
れたキズ画像の一部または全体に対して,ニューラルネ
ットワークを用いた認識処理を施した。
In the case of this embodiment, the unit of distance is as shown in FIG.
If the space between the adjacent shape recognition elements 5 in is adopted,
From the irradiated position sensing element 4 a of the reflected light, the light modulation device 3 of a distance in a range greater than 2 units shape recognition element 5 were all light-absorbing state. As a result, light reflected by the flaw, by entering into one position detecting element 4 a of the image clipping device 8, the distance only through the shape recognizing device 5 in the following two units, from all or part of the flaw The reflected light could be transmitted. As a result, it was possible to recognize the location of the scratch image and cut out a part or the whole of the scratch image. Furthermore, the image cropping device 8
The reflected light that has passed through is incident on the quantization processing device 13. That is, the reflected light is incident only on the light receiving element 2 b in the quantization processing device 13 which is directly above the light modulation element 3 in the light transmitting state in the image clipping device 8. As a result, if the distance between the adjacent position detection elements 4 b in the quantization processing device 13 in FIG. 2 is adopted as the unit of distance, one position detection element 4 in which the reflected light is incident in the quantization processing device 13 is adopted.
b (light receiving element 2 b ) electric resistance and its position detecting element 4
The electric resistance of the light receiving element 2 b in the quantizing element 10 connected in series with the position detecting element 4 b within a distance (2 units) from b is reduced. In the present embodiment, and one in the position detecting elements 4a of the quantization processor 13, the electrical resistance of the total of 21 light receiving elements 2 b of twenty quantization circuit 11 is reduced. As a result, the voltage applied to the light emitting element 9 connected in series with the light receiving element 2 b whose electric resistance has decreased increases. Then, only the light emitting element 9 whose applied voltage becomes equal to or higher than the threshold value emits light. That is, the binarization (quantization) process based on the threshold value is performed. In the above embodiments, there is no need to scan the voltage applied to each pixel as is done with CRTs, CCDs, liquid crystals, etc., and since the parallelism of light is used as it is, there is a flaw. The location recognition, segmentation, and quantization can be performed almost simultaneously, and high-speed processing can be performed. A recognition process using a neural network was applied to a part or the whole of the scratch image thus obtained.

【0011】本実施例によると,画像切り出し装置8に
よる位置検知後には,複数個からなる位置検知素子4a
の内のいずれか1個の光変調素子3から,キズに起因し
た反射光が透過することになる。また,光透過状態にな
った形状認識素子5内の光変調素子3の分布を,画像処
理装置Aの上部から見た場合,光透過状態になった位置
検知素子4a 内の光変調素子3を含む面(走査方向に平
行の面)に対して,左右対象となっている。すなわち、
切り出した領域の中で,キズ画像の先端は,必ず切り出
し領域の上部中央に位置することになり,切り出し領域
内でのキズ画像の位置を,このレベルで特定できる。そ
して,画像切り出し装置8によって切り出された反射光
による画像に対して,量子化処理装置13によってほぼ
同時に発光素子9による量子化画像が得られるが,この
量子化画像についても,キズ画像の先端に対応した光信
号が,必ず切り出し領域の上部中央の発光素子9から発
光する。このことは,切り出された画像の中で対象画像
の位置が変わると一般に認識率が低下するニューラルネ
ットワークに対して,有効に寄与し,認識率を向上させ
ることができる。以上の構成によって,キズの形状によ
る分類を,高速にかつ高い認識率で行うことができた。
According to this embodiment, after the position is detected by the image clipping device 8, a plurality of position detecting elements 4 a are formed.
From any one of the light modulation elements 3, the reflected light caused by the flaw is transmitted. Further, when the distribution of the light modulation elements 3 in the shape recognition element 5 in the light transmission state is viewed from above the image processing apparatus A, the light modulation elements 3 in the position detection element 4 a in the light transmission state are shown. Left and right are symmetrical with respect to the surface including (the surface parallel to the scanning direction). That is,
In the clipped area, the tip of the scratched image is always located at the upper center of the clipped area, and the position of the scratched image in the clipped area can be specified at this level. Then, with respect to the image by the reflected light cut out by the image cutout device 8, a quantized image by the light emitting element 9 can be obtained almost simultaneously by the quantization processing device 13, and this quantized image also appears at the tip of the scratch image. The corresponding optical signal always emits from the light emitting element 9 in the upper center of the cutout area. This effectively contributes to the neural network in which the recognition rate generally decreases when the position of the target image changes in the clipped image, and the recognition rate can be improved. With the above configuration, it was possible to perform classification by the shape of scratches at high speed and with high recognition rate.

【0012】本実施例では,表面走査装置に印加する電
圧が5Vの場合について述べたが,印加電圧が1V,1
0V,30V,100Vの場合においても,各素子の電
気抵抗を変えることにより,同様の手法でキズの認識を
することができた。また,キズに限らず,被検体表面の
異物質や汚れなども認識することができた。以上述べた
実施例では所定の画像を切り出して2値化する目的で,
量子化装置と画像切り出し装置が組み合わされている
が,別段画像切り出しを行わない場合には,画像切り出
し装置と組み合わせる必要はまったくない。また,まっ
たく別個の画像切り出し装置と組み合わせることも可能
である。
In this embodiment, the case where the voltage applied to the surface scanning device is 5V has been described, but the applied voltage is 1V, 1V.
Even in the case of 0V, 30V, and 100V, it was possible to recognize the flaw by the same method by changing the electric resistance of each element. Moreover, not only scratches but also foreign substances and dirt on the surface of the subject could be recognized. In the embodiment described above, in order to cut out a predetermined image and binarize it,
Although the quantizer and the image cutout device are combined, it is not necessary to combine with the image cutout device if the image cutout is not performed separately. It is also possible to combine it with a completely separate image cropping device.

【0013】[0013]

【発明の効果】本発明は,上記したように構成されてい
る。それによって,被検体表面の画像の量子化処理を光
の並列性を利用した並列処理を行うことができ,画像の
各画素ごとに時分割処理を行う必要がなく、高速な画像
量子化処理が実現できる。そして,量子化処理された表
面異常の画像に対して,ニューラルネットワークによる
認識を行うことにより,表面異常部の形状による分類
を,高速かつ高い認識率で行うことができる。
The present invention is constructed as described above. As a result, it is possible to perform the quantization process of the image of the surface of the subject by using the parallelism of light, and it is not necessary to perform the time division process for each pixel of the image, and the high-speed image quantization process can be performed. realizable. Then, the quantized image of the surface abnormality is recognized by the neural network, so that the shape of the surface abnormality portion can be classified at high speed and with a high recognition rate.

【0014】[0014]

【図面の簡単な説明】[Brief description of drawings]

【図1】 本発明の一適用例である表面欠陥検出装置の
一構成要素である画像切り出し装置8を示す配線ブロッ
ク図。
FIG. 1 is a wiring block diagram showing an image clipping device 8 which is a component of a surface defect detection device as an application example of the present invention.

【図2】 本発明の一実施例に係る画像量子化処理装置
13を示す配線ブロック図。
FIG. 2 is a wiring block diagram showing an image quantization processing device 13 according to an embodiment of the present invention.

【図3】 上記画像切り出し装置を用いて被検体表面の
表面欠陥画像などを切り出して量子化処理する状態を示
す概念図。
FIG. 3 is a conceptual diagram showing a state in which a surface defect image or the like on a surface of a subject is cut out and quantized by using the image cutting device.

【図4】 表面欠陥画像と欠陥検出装置の関係を示す平
面図。
FIG. 4 is a plan view showing the relationship between a surface defect image and a defect detection device.

【図5】 従来の表面検査装置の概念図。FIG. 5 is a conceptual diagram of a conventional surface inspection device.

【図6】 従来の量子化画像から特徴部を抽出する手順
を示す機能ブロック図。
FIG. 6 is a functional block diagram showing a procedure for extracting a characteristic portion from a conventional quantized image.

【符号の説明】[Explanation of symbols]

a ,2b …受光素子 3…光変調素子 4a ,4b …位置検知素子 5…形状認識素子 6…画像切り出し素子 7…抵抗 8…画像切り出し装置 9…発光素子 10…量子化素子 11…量子化回路 12…位置検知回路 13…量子化処理装置 14…被検体2 a , 2 b ... Light receiving element 3 ... Light modulation element 4 a , 4 b ... Position detecting element 5 ... Shape recognition element 6 ... Image cutting element 7 ... Resistor 8 ... Image cutting device 9 ... Light emitting element 10 ... Quantization element 11 ... Quantization circuit 12 ... Position detection circuit 13 ... Quantization processing device 14 ... Subject

Claims (2)

【特許請求の範囲】[Claims] 【請求項1】 被検体表面を露光走査して得た被検体表
面画像を量子化処理する画像量子化処理装置において,
上記被検体からの光線の通過経路に設けられ,被検体か
らの光線の光量に応じて量子化された信号を出力する微
小面積の量子化素子を直線状に多数配置してなる量子化
処理手段を上記量子化素子の配置方向に直角の方向に被
検体に対して相対移動させることを特徴とする画像量子
化処理装置。
1. An image quantization processing device for performing a quantization process on a subject surface image obtained by exposing and scanning the subject surface,
Quantization processing means which is provided in the passage path of the light beam from the subject and has a large number of linearly arranged quantizing elements having a small area for outputting a quantized signal according to the light quantity of the light beam from the subject. An image quantization processing device, characterized in that is moved relative to the subject in a direction perpendicular to the arrangement direction of the quantization element.
【請求項2】 被検体の表面を2次元的に露光走査する
露光走査器によって,上記被検体表面から測定される反
射光の測定量に基づいて,上記被検体表面の光学的に検
出される状態からの反射光でなる画像を量子化処理する
画像処理装置において,受光によって電気抵抗が変化す
る受光素子とある一定の閾値以上の印加電圧によって発
光する発光素子とを直列に接続してなる量子化素子を並
列に接続して構成された量子化装置と,受光によって電
気抵抗が変化する受光素子からなる位置検知素子を並列
に接続して構成された位置検知装置とを,直列に接続
し,該位置検知装置側が正となり該量子化装置側が負と
なるように,同一電源からの電圧を印加する回路からな
ることを特徴とする画像量子化処理装置。
2. An exposure scanner that two-dimensionally exposes and scans the surface of the subject to optically detect the surface of the subject based on the amount of reflected light measured from the surface of the subject. In an image processing device for performing a quantization process on an image formed by reflected light from a state, a quantum device formed by connecting in series a light receiving element whose electric resistance changes by light reception and a light emitting element which emits light by an applied voltage above a certain threshold value. And a position detecting device configured by connecting in parallel a position detecting element composed of a light receiving element whose electric resistance changes by light reception, and a quantizing device configured by connecting parallelizing elements in parallel, An image quantization processing device comprising a circuit for applying a voltage from the same power source so that the position detection device side becomes positive and the quantization device side becomes negative.
JP7089892A 1992-03-27 1992-03-27 Image quantization processor Pending JPH05273143A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP7089892A JPH05273143A (en) 1992-03-27 1992-03-27 Image quantization processor

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP7089892A JPH05273143A (en) 1992-03-27 1992-03-27 Image quantization processor

Publications (1)

Publication Number Publication Date
JPH05273143A true JPH05273143A (en) 1993-10-22

Family

ID=13444813

Family Applications (1)

Application Number Title Priority Date Filing Date
JP7089892A Pending JPH05273143A (en) 1992-03-27 1992-03-27 Image quantization processor

Country Status (1)

Country Link
JP (1) JPH05273143A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103512903A (en) * 2013-09-16 2014-01-15 青海中控太阳能发电有限公司 Method and system for automatically measuring surface clearness of heliostat

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103512903A (en) * 2013-09-16 2014-01-15 青海中控太阳能发电有限公司 Method and system for automatically measuring surface clearness of heliostat

Similar Documents

Publication Publication Date Title
KR0159925B1 (en) Process for determining the optical quality of flat glasss or flat glass products
JP3178644B2 (en) Defect detection method for transparent plate
US5076692A (en) Particle detection on a patterned or bare wafer surface
JP4351522B2 (en) Pattern defect inspection apparatus and pattern defect inspection method
US6175645B1 (en) Optical inspection method and apparatus
JP2795595B2 (en) Defect detection method for transparent plate
KR100669845B1 (en) Scanning method and system for inspecting anomalies on surface
EP0459489B1 (en) Method of reading optical image of inspected surface and image reading system employable therein
EP0179309A2 (en) Automatic defect detection system
WO2006108137A2 (en) Glass inspection systems and methods for using same
GB2051349A (en) Automatic defecting inspection apparatus
EP0016551A1 (en) Apparatus for inspecting translucent articles for faults
US8135207B2 (en) Optical inspection tools featuring parallel post-inspection analysis
US8547547B2 (en) Optical surface defect inspection apparatus and optical surface defect inspection method
JPH05273143A (en) Image quantization processor
KR20040017838A (en) Autothresholding of noisy images
CN111638226B (en) Detection method, image processor and detection system
JPH05273144A (en) Image processor
US6201623B1 (en) Surface topography enhancement
JPH0511574B2 (en)
JPH05276453A (en) Image segmenting device
Batchelor et al. Defect detection on the internal surfaces of hydraulics cylinders for motor vehicles
JPH1063913A (en) Quality decision device
US11415528B2 (en) Method and apparatus for automated in-line inspection of optically transparent materials
JPH05332947A (en) Surface inspection device and method