JPH03229373A - Learning device for two-dimensional pattern - Google Patents

Learning device for two-dimensional pattern

Info

Publication number
JPH03229373A
JPH03229373A JP2025392A JP2539290A JPH03229373A JP H03229373 A JPH03229373 A JP H03229373A JP 2025392 A JP2025392 A JP 2025392A JP 2539290 A JP2539290 A JP 2539290A JP H03229373 A JPH03229373 A JP H03229373A
Authority
JP
Japan
Prior art keywords
dimensional pattern
line
pattern
learning device
feature extraction
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
JP2025392A
Other languages
Japanese (ja)
Inventor
Tadashi Kaneko
正 金子
Atsuharu Yamamoto
淳晴 山本
Hidemi Takahashi
秀実 高橋
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.)
Panasonic Holdings Corp
Original Assignee
Matsushita Electric Industrial Co 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 Matsushita Electric Industrial Co Ltd filed Critical Matsushita Electric Industrial Co Ltd
Priority to JP2025392A priority Critical patent/JPH03229373A/en
Publication of JPH03229373A publication Critical patent/JPH03229373A/en
Pending legal-status Critical Current

Links

Landscapes

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

Abstract

PURPOSE:To accurately learn the features of the line width, etc., by using a feature extracting operator window in accordance with the line direction/width of a two-dimensional pattern and applying the AND and OR operations to the binary value of images of each picture element. CONSTITUTION:An operator operation part 11 contains a feature extracting operator window store part 10 which includes a feature extracting operator window having the length larger than the number of width picture elements by >=2 picture elements in the direction orthogonal to the line of a two-dimensional pattern and at least the length equal to 2 picture elements in the same direction as the pattern line respectively. Then the part 11 works so as to store the binary image of the two-dimensional pattern in an entire area of the operator window. The data on the binary image are inputted to an AND arithmetic part 12 and an OR arithmetic part 13. Then the minimum line width and the maximum line width are obtained via an AND and an OR applied to the numerical value of the binary image respectively. As a result, the features of the line width, etc., are accurately learnt.

Description

【発明の詳細な説明】 産業上の利用分野 本発明は、プリント基板、集積回路のパターンやそのマ
スクの欠陥検査時の前処理行程に用いられ、検査対象の
2次元パターンの線の方向、線幅等の特徴を学習し、実
検査等の判断基準にする2次元パターンの学習装置に関
するものである。
DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention is used in a pre-processing process when inspecting patterns of printed circuit boards, integrated circuits, and their masks for defects. The present invention relates to a two-dimensional pattern learning device that learns characteristics such as width and uses the characteristics as criteria for actual inspections, etc.

従来の技術 一般に、プリント基板のパターン形成工程においては、
多数のフォトマスク処理やエツチング工程により表面に
必要なパターンが形成されるが、プリント基板に対する
実装密度が高くなるのにつれて、形成されるパターンの
線幅は細くなり、パターンも複雑化する傾向になる。こ
のように、パターンが微細で複雑化すると、断線、ピン
ホール等の欠陥も生じ易くなるから、この欠陥の検査を
正確かつ迅速に行なうことが、各種製品の製造に際して
、非常に重要になってきている。
Conventional technology Generally, in the pattern forming process of printed circuit boards,
The necessary patterns are formed on the surface through multiple photomask processing and etching processes, but as the mounting density on printed circuit boards increases, the line width of the formed patterns tends to become thinner and the patterns become more complex. . As the patterns become finer and more complex, defects such as wire breaks and pinholes are more likely to occur, so it is extremely important to accurately and quickly inspect these defects when manufacturing various products. ing.

第5図は、”サーキットテクノロジーVo1.3No、
1 (1988年)のページ2〜12に発表された、従
来の欠陥検査のアルゴリズムであり、検査装置はプリン
ト基板のパターン部1情報を読込めるオペレータ操作部
2を有し、オペレータ操作部2にはオペレータ収納部3
が設けられろ。オペレータ収納部は基本的に第6図(a
)のような十字オペレータA及び同図(b)のようなX
字オペレータBを有し、オペレータ操作部2で、表のテ
ーブルに基づいて、パターン1の状態により各オペレー
タAとBのいずれか一方又は両方が用いられる。つまり
、線幅検査で、例えば十字オペレータAが選択されると
、オペレータ操作部2でその十字オペレータAをパター
ン1に重合させてデータを抽出する。この場合、線幅小
欠陥があると、第7図(a)のように、十字オペレータ
Aにおいて上下の腕U、Dの画素Cの値が小さくなり、
ピンホールの欠陥では(b)のように、画素の値Cが4
つの腕U、R1D、Lに分散する。したがって、上述の
4つの腕U、R,D、Lと画素の値Cが判定部4に送ら
れ、欠陥の有無が判断されることになる。
Figure 5 shows “Circuit Technology Vol. 1.3 No.
1 (1988), pages 2 to 12, the inspection device has an operator operation section 2 that can read information from the pattern section 1 of a printed circuit board. is operator storage section 3
be established. The operator storage compartment is basically as shown in Figure 6 (a)
) and X as shown in (b) of the same figure.
In the operator operation unit 2, one or both of the operators A and B is used depending on the state of pattern 1 based on the table. That is, when, for example, a crosshair operator A is selected in the line width inspection, the operator operation section 2 superimposes the crosshair operator A on the pattern 1 and extracts data. In this case, if there is a small line width defect, the value of the pixel C of the upper and lower arms U and D of the cross operator A becomes small, as shown in FIG. 7(a).
In the case of a pinhole defect, as shown in (b), the pixel value C is 4.
Distributed into three arms U, R1D, and L. Therefore, the above-mentioned four arms U, R, D, L and the pixel value C are sent to the determination section 4, and the presence or absence of a defect is determined.

そして、微小欠陥を生じている場合は、欠陥コードによ
り致命欠陥判定部6に進み、判定テーブルと比較され、
欠陥状態がテーブルのものと一致した場合は致命欠陥と
判定するように構成されている。
If a minute defect has occurred, the defect code is used to proceed to the fatal defect determination section 6, where it is compared with a determination table.
If the defect status matches that in the table, the defect is determined to be a fatal defect.

以下余白 発明が解決しようとする課題 ところで、前述した従来例のものにあっては、検査前の
パターン1の正常な情報が無く、十字又はX字のオペレ
ータ八又はBと、その画素の値Cとのみで欠陥の有無が
判断される。このため、検査の精度を向上するには、オ
ペレータ抽出データを多くする必要があり、検査ロジy
りのハードウェアの複雑化を招き、高速検査の障害にな
る等の課題がある。
Problems to be Solved by the Invention of the Margin Below However, in the conventional example described above, there is no normal information of pattern 1 before inspection, and the operator 8 or B of the cross or X-shape and the value C of that pixel The presence or absence of a defect can be determined only by Therefore, in order to improve inspection accuracy, it is necessary to increase the amount of operator extracted data, and
This poses problems, such as increasing the complexity of the hardware and impeding high-speed inspection.

本発明は、前述したような従来の課題に鑑みてなされた
もので、その目的とするところは、検査対象のパターン
を前処理で学習して実検査時の判断基準に用い、検査行
程を簡略化できる、高速検査用2次元パターンの学習装
置を提供することにあるO 課題を解決するための手段 前記目的を達成するため、本発明の2次元パターンの学
習装置は、2次元パターンの線に対して線と直交する方
向には幅面素数より2画素以上多く、線と同一方向には
少なくとも2画素の長さの特徴抽出オペレータ窓を有す
る特徴抽出オペレータ窓収納部と、特徴抽出オペレータ
窓の全域に2次元パターンの2値画像をとらえるように
操作するオペレータ操作部と、2値画像の数値の線方向
に並ぶ各列の全部の論理和を求める論理和演算部と、論
理積を求める論理積演算部と、論理和と論理積の数値か
ら線幅の最大及び最小に基づくパターン特徴、座標が一
致するか否かを学習する特徴学習部と、学習データを記
憶する記憶部とを備えるものである。
The present invention was made in view of the conventional problems mentioned above, and its purpose is to simplify the inspection process by learning the pattern to be inspected in preprocessing and using it as a judgment criterion during actual inspection. An object of the present invention is to provide a two-dimensional pattern learning device for high-speed inspection, which allows the lines of a two-dimensional pattern to be On the other hand, there is a feature extraction operator window storage section having a feature extraction operator window with a length of at least two pixels longer than the width prime number in the direction perpendicular to the line and at least two pixels in the same direction as the line, and an operator operation unit that operates to capture a binary image of a two-dimensional pattern; a logical sum operation unit that calculates the logical sum of all the columns of numerical values in the binary image arranged in the line direction; and a logical product unit that calculates the logical product. It is equipped with an arithmetic unit, a feature learning unit that learns whether or not pattern features and coordinates match based on the maximum and minimum line widths from numerical sums and logical products, and a storage unit that stores learning data. be.

作用 このような構成によると、2次元パターンの縦横の線の
幅の2値画像が特徴抽出オペレータの総画素でとらえら
れ、この2値画像の数値を論理和することで最大線幅が
、論理積することで最小線幅が求まり、これらのデータ
により線幅の特徴が学習される。また、線幅が太き過ぎ
て座標が一致しないことも学習され、これらの学習デー
タが記憶される。したがって、パターンの欠陥検査にお
いて学習データな判断基準に用いて検査データと比べる
ことで、迅速に欠陥の有無が判断されるようになる。
Function: According to this configuration, a binary image of the vertical and horizontal line widths of a two-dimensional pattern is captured by the total pixels of the feature extraction operator, and by logically adding the values of this binary image, the maximum line width can be determined by By multiplying the values, the minimum line width is determined, and the characteristics of the line width are learned using this data. It is also learned that the line width is too thick and the coordinates do not match, and this learning data is stored. Therefore, by using learned data as a judgment criterion in pattern defect inspection and comparing it with inspection data, the presence or absence of a defect can be quickly determined.

実施例 以下、第1図ないし第6図を用い本発明の実施例を詳細
に説明する。
EXAMPLES Hereinafter, examples of the present invention will be described in detail using FIGS. 1 to 6.

第1図は本発明において2次元パターン学習で検査の前
処理を行う場合のブロック図であり、本発明の検査装置
は特徴抽出オペレータ窓収納部1゜並びに特徴抽出オペ
レータ窓りを有している。この特徴抽出オペレータ窓り
においては、学習対象の2次元パターンの線等との関係
から画素数が予め設定されている。
FIG. 1 is a block diagram when preprocessing for inspection is performed by two-dimensional pattern learning in the present invention, and the inspection apparatus of the present invention has a feature extraction operator window housing section 1° and a feature extraction operator window. . In this feature extraction operator window, the number of pixels is set in advance based on the relationship with lines, etc. of the two-dimensional pattern to be learned.

つまり、第2図(a)のようにパターンの横線Pで線幅
Wに対し8画素以上の場合にあっては、特徴抽出オペレ
ータ窓りには、横線Pと直交する縦方向に、必要な8個
に対して2個以上加算した(例えば10個)画素dが設
けられる。また、横方向にあっては、2個以上(例えば
8個)画素dが設けられ、したがって、全体画素dは1
0x8の2値画像に区画した構成となる。
In other words, if the horizontal line P of the pattern is 8 pixels or more relative to the line width W as shown in Figure 2(a), the feature extraction operator window will display the required number of pixels in the vertical direction orthogonal to the horizontal line P. Two or more (for example, 10) pixels d are provided compared to the 8 pixels. In addition, in the horizontal direction, two or more (for example, eight) pixels d are provided, so the total pixel d is 1
The configuration is divided into 0x8 binary images.

一方、オペレータ操作部11においては、パターン部1
のパターン状態により、横線Pの場合、前述した上記特
徴抽出オペレータ窓りを選択し、パターン上で操作して
各画素dに画像Xをとらえる。
On the other hand, in the operator operation section 11, the pattern section 1
According to the pattern state, in the case of horizontal line P, the above-described feature extraction operator window is selected and operated on the pattern to capture image X at each pixel d.

したがって、この場合のデータは縦方向の各列で[0、
o、 xlx、 x、 x、 x、 x、 x、 x、
 o、、 o]になり、これが8列だけ得られる。ここ
で、画像Xは、有無に応じて1と0の任意の値で示され
る。
Therefore, the data in this case is [0,
o, xlx, x, x, x, x, x, x,
o,, o], which yields only 8 columns. Here, the image X is indicated by an arbitrary value of 1 or 0 depending on the presence or absence of the image.

この画像データは論理積演算部12と論理和演算部13
に入力され、論理積演算部12は横の線方向に並ぶ各列
の全画像Xの論理積を求める(第2図(a)の場合には
、[0,0,0,0,0,1、・・・]となる)。同様
に、論理和演算部13は全画像Xの論理和を求め(第2
図(a)の場合には、[0,0,1,1,1,1、・・
・]となる)、これらの演演算部が特徴学習部14に入
力される。以上のように、特徴学習部14では、”1“
の数が最大となる論理和データにより最大線幅が求めら
れ、逆に最小となる論理積データにより最小線幅が求め
られて、線幅の特徴が学習され、この学習結果が記憶部
16に記憶される。つまり、実検査時に、記憶部16の
学習データは判断部4に出力され、検査の際のオペレー
タによる特徴認識と一致するか否かの判断基準となる。
This image data is processed by the AND operation section 12 and the OR operation section 13.
is input, and the logical product calculation unit 12 calculates the logical product of all images X in each column arranged in the horizontal line direction (in the case of FIG. 2(a), 1,...]). Similarly, the logical sum calculation unit 13 calculates the logical sum of all images X (the second
In the case of figure (a), [0, 0, 1, 1, 1, 1,...
), these operation units are input to the feature learning unit 14. As described above, in the feature learning unit 14, "1"
The maximum line width is determined by the logical sum data that has the maximum number of , and the minimum line width is determined by the logical product data that has the minimum number of , and the characteristics of the line width are learned. This learning result is stored in the storage unit 16. be remembered. That is, during the actual inspection, the learning data in the storage section 16 is output to the judgment section 4, and serves as a criterion for determining whether or not it matches the feature recognition by the operator during the inspection.

また、論理積データがすべて”1“の場合は、特徴学習
部14は線幅が太過ぎる学習パターンの座標と一致しな
いことが学習される。そして、このオペレータ窓りの座
標の不一致は記憶部16に登録され、実検査時の参考に
なるように構成されている。
Further, if all the logical product data are "1", the feature learning unit 14 learns that the line width does not match the coordinates of the learning pattern that is too thick. This discrepancy in the coordinates of the operator window is registered in the storage unit 16 and is configured to serve as a reference during actual inspection.

このように構成された2次元パターンの特徴学習部14
においては、パターンの部1の欠陥検査の前処理時に、
先ず、パターンの線の縦横、線幅に応じた特徴抽出オペ
レータ窓りが選択される。
The two-dimensional pattern feature learning unit 14 configured as described above
In this case, during pre-processing for defect inspection of pattern part 1,
First, a feature extraction operator window is selected depending on the length and width of the lines of the pattern and the line width.

そして、この特徴抽出オペレータ窓りの各画素dの画像
Xのデータが読込まれ、座標が一致する場合は、両端の
ビットが”0゛のときに、両端の”0”に挾まれた“1
“の連なる数を論理積及び論理和演算して、線幅の特徴
が学習される。
Then, the data of the image
Characteristics of line width are learned by performing AND and OR operations on successive numbers.

いい換えると、実検査時に、例えば十字オペレータAを
用いてパターン部1の線幅に応じたデータが得られ、欠
陥の有無が判断される場合に、上述の学習による線幅の
特徴が判断基準に用いられ、両者が一致するか否かが比
較され、一致する場合は正常と判断し、不一致の場合は
欠陥と判断されることになる。また、実際の検査行程に
おいて、断線、ショート、線太り、線細り等の不良欠陥
が含まれている場合は、記憶部16の情報は正しいもの
と判断される。
In other words, during actual inspection, when data corresponding to the line width of pattern portion 1 is obtained using, for example, cross operator A, and the presence or absence of a defect is determined, the line width characteristics obtained by the above-mentioned learning are used as the judgment criteria. It is used to compare the two to see if they match, and if they match, it is determined to be normal, and if they do not match, it is determined to be defective. Further, in the actual inspection process, if defects such as disconnections, short circuits, thick lines, thin lines, etc. are included, the information in the storage unit 16 is determined to be correct.

なお、特徴抽出オペレータ窓りの画像Xは0と定めるこ
ともできるが、この場合は、負論理で演算すればよいの
は、改めて説明するまでもない。
It should be noted that the image X in the feature extraction operator window can be set to 0, but in this case, it is needless to explain that it is sufficient to perform calculations using negative logic.

第2図(b)は縦線用のオペレータ窓であって、縦線P
′に対し横方向に長い特徴抽出オペレータ窓D′を用い
ているが、このようなオペレータ窓によっても、第2図
(a)について前述したのと同様に、学習できるのは明
らかである。
FIG. 2(b) is an operator window for vertical lines.
Although a feature extraction operator window D' that is long in the horizontal direction with respect to ' is used, it is clear that learning can be performed using such an operator window in the same manner as described above with respect to FIG. 2(a).

第3図は線幅は同一であるが方向の異なるパターンが混
在している場合の一例であるけれども、この場合は、折
れ線Pに対し第2図(a)のような特徴抽出オペレータ
窓りを用い、第3図(a)のように横方向からと、第3
図(b)のように斜め方向から読取ればよい。
Figure 3 is an example of a case where patterns with the same line width but different directions coexist. As shown in Figure 3(a), from the lateral direction and from the third
It is sufficient to read it from an oblique direction as shown in FIG.

また、第4図はパターンの中央に断線欠陥がある場合を
示す。この場合は、第4図(a)のように縦線Pの途中
に欠陥部Sがあると、第4図(b)のように欠陥部Sの
方向は学習できない。このため、このような欠陥の場合
は、画像の膨張又は収縮を行うと、第4図(c)のよう
に修正した上で、学習させることができる。
Further, FIG. 4 shows a case where there is a disconnection defect at the center of the pattern. In this case, if there is a defective part S in the middle of the vertical line P as shown in FIG. 4(a), the direction of the defective part S cannot be learned as shown in FIG. 4(b). Therefore, in the case of such a defect, if the image is expanded or contracted, it can be corrected as shown in FIG. 4(c) and then learned.

発明の効果 以上に述べてきたように、本発明によれば、2次元パタ
ーンの線の方向、線幅に応じた特徴抽出オペレータ窓を
用い、各画素の画像の2値数値を論理積及び論理和演算
ずろことで、線幅等の特徴を正確に学習できる。また、
この特徴により2次元パターンの大き過ぎ等の座標も学
習できることになる。
Effects of the Invention As described above, according to the present invention, a feature extraction operator window corresponding to the line direction and line width of a two-dimensional pattern is used to perform logical product and logic on the binary values of the image of each pixel. By performing summation operations, features such as line width can be learned accurately. Also,
This feature makes it possible to learn the coordinates of two-dimensional patterns that are too large.

し、たがって、本発明によるこの2次元パターンの学習
データをパターン欠陥検査の判断基準に用いると、検査
データと学習データを比較するだけで欠陥の有無が判断
され、高速検査が可能になる。
Therefore, if the learning data of this two-dimensional pattern according to the present invention is used as a criterion for pattern defect inspection, the presence or absence of a defect can be determined simply by comparing the inspection data and the learning data, making high-speed inspection possible.

また、本発明を用いる検査工程は非常に簡略化したもの
となる。
Furthermore, the inspection process using the present invention is greatly simplified.

本発明で用いる特徴抽出オペレータ窓は、縦、横の2種
類と、画素の数の多少により2次元パターンの全てに適
用でき、画像の膨張、収縮により欠陥部を有する場合に
も適用できるから、検査装置として実用性の高いものと
なる・
The feature extraction operator window used in the present invention can be applied to all two-dimensional patterns depending on the two types, vertical and horizontal, and the number of pixels, and can also be applied to cases where the image has defects due to expansion or contraction. It becomes highly practical as an inspection device.

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

第1図は本発明の一実施例における2次元パターンの学
習装置のブロック結線図、第2図は特徴抽出オペレータ
窓を示すもので、(a)は横線用、(b)は縦線を示す
概念図、第3図(a)、(b)は方向の異なるパターン
が混在する場合の読取り状態を示す図、第4図(a)な
いしくc)はパターン中央に欠陥部を有する場合の修正
状態を示す図、第6図は従来のパターン欠陥検査工程を
示すブロック図、第6図は欠陥検査のオペレータを示す
もので、(a)は十字オペレータを、(b)はX字オペ
レータを示す図、第7図は欠陥検査状態を示すもので、
(a)は線幅小の欠陥を、(b)はピンホールの欠陥を
示す図である。 1・・・パターン部、1o・・・特徴抽出オペレータ窓
収納部、11・・・オペレータ操作部、12・・・論理
積演算部、13・・・論理和演算部、14・・・特徴学
習部、15・・・記憶部、P・・・線、W・・・線幅、
D・・・特徴抽出オペレータ窓。
Fig. 1 is a block diagram of a two-dimensional pattern learning device according to an embodiment of the present invention, and Fig. 2 shows a feature extraction operator window, where (a) shows horizontal lines and (b) shows vertical lines. Conceptual diagram, Figures 3 (a) and (b) are diagrams showing the reading state when patterns with different directions coexist, and Figures 4 (a) to c) are corrections when there is a defective part in the center of the pattern. Figure 6 is a block diagram showing the conventional pattern defect inspection process. Figure 6 shows the defect inspection operators. (a) shows the cross operator, and (b) shows the X operator. Figure 7 shows the defect inspection status.
(a) is a diagram showing a defect with a small line width, and (b) is a diagram showing a pinhole defect. DESCRIPTION OF SYMBOLS 1... Pattern part, 1o... Feature extraction operator window storage part, 11... Operator operation part, 12... Logical product operation part, 13... Logical sum operation part, 14... Feature learning Part, 15...Storage part, P...Line, W...Line width,
D...Feature extraction operator window.

Claims (5)

【特許請求の範囲】[Claims] (1)2次元パターンの線に対して線と直交する方向に
は幅画素数より2画素以上多く、線と同一方向には少な
くとも2画素の長さの特徴抽出オペレータ窓を有する特
徴抽出オペレータ窓収納部と、特徴抽出オペレータ窓の
全域に2次元パターンの2値画像をとらえるように操作
するオペレータ操作部と、2値画像の数値の線方向に並
ぶ各列の全部の論理和を求める論理和演算部と、論理積
を求める論理積演算部と、論理和と論理積の数値から線
幅の最大及び最小に基づくパターン特徴、座標が一致す
るか否かを学習する特徴学習部と、学習データを記憶す
る記憶部とを備える2次元パターンの学習装置。
(1) A feature extraction operator window that has a feature extraction operator window with a length of at least 2 pixels in the direction perpendicular to the line of the two-dimensional pattern, which is more than the number of width pixels in the direction perpendicular to the line, and at least 2 pixels in the same direction as the line. A storage section, an operator operation section that operates to capture a binary image of a two-dimensional pattern in the entire area of the feature extraction operator window, and a logical sum that calculates the logical sum of all the columns of numerical values of the binary image arranged in the line direction. an arithmetic unit, an AND operation unit that calculates a logical product, a feature learning unit that learns whether pattern features and coordinates match based on the maximum and minimum line width from the values of the logical sum and the logical product, and learning data. A two-dimensional pattern learning device, comprising: a storage unit for storing a two-dimensional pattern;
(2)記憶部の学習データはパターン欠陥検査の判断基
準に用いる請求項1記載の2次元パターンの学習装置。
(2) The two-dimensional pattern learning device according to claim 1, wherein the learning data in the storage section is used as a criterion for pattern defect inspection.
(3)2値画像の数値は両端ビットが”1”の場合に”
O”とし、両端ビットが”O”の場合に”1”とする請
求項1記載の2次元パターンの学習装置。
(3) The numerical value of a binary image is "1" when both end bits are "1".
The two-dimensional pattern learning device according to claim 1, wherein the two-dimensional pattern learning device is set to “1” when both end bits are “O”.
(4)2次元パターン内に欠陥部を有する場合に、欠陥
部の位置を含む領域を2次元画像として膨張又は収縮に
より修正する機能を有する請求項1記載の2次元パター
ンの学習装置。
(4) The two-dimensional pattern learning device according to claim 1, which has a function of, when a two-dimensional pattern has a defective portion, correcting the area including the position of the defective portion by expanding or contracting it as a two-dimensional image.
(5)2次元パターンの線の方向が異なるものが混在す
る場合に、各線の方向と特徴抽出オペレータ窓の方向が
一致するように傾けて操作する請求項1記載の2次元パ
ターンの学習装置。
(5) The two-dimensional pattern learning device according to claim 1, wherein when two-dimensional patterns with different line directions coexist, the device is operated by tilting so that the direction of each line coincides with the direction of the feature extraction operator window.
JP2025392A 1990-02-05 1990-02-05 Learning device for two-dimensional pattern Pending JPH03229373A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2025392A JPH03229373A (en) 1990-02-05 1990-02-05 Learning device for two-dimensional pattern

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2025392A JPH03229373A (en) 1990-02-05 1990-02-05 Learning device for two-dimensional pattern

Publications (1)

Publication Number Publication Date
JPH03229373A true JPH03229373A (en) 1991-10-11

Family

ID=12164616

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2025392A Pending JPH03229373A (en) 1990-02-05 1990-02-05 Learning device for two-dimensional pattern

Country Status (1)

Country Link
JP (1) JPH03229373A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140034201A1 (en) * 2011-04-15 2014-02-06 Bridgestone Corporation Pneumatic tire for two-wheeled motor vehicle

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140034201A1 (en) * 2011-04-15 2014-02-06 Bridgestone Corporation Pneumatic tire for two-wheeled motor vehicle

Similar Documents

Publication Publication Date Title
WO2018068415A1 (en) Detection method and system for wrong part
JPH04115144A (en) Image processor and automatic optical inspection apparatus using same
JPS5951135B2 (en) object inspection equipment
JPH0723847B2 (en) Printed circuit board pattern inspection method
JPH03229373A (en) Learning device for two-dimensional pattern
JP3344338B2 (en) Image defect detection apparatus and defect detection method
JP2002008029A (en) Image inspection device
JP4546607B2 (en) Non-defective pattern registration method and pattern inspection method
JPS61151410A (en) Pattern defect detection and its apparatus
JP2794924B2 (en) Wiring pattern inspection equipment
JPS6385432A (en) Inspecting method for linear defect
JPS6361373A (en) Method and device for inspecting shape
TWI786894B (en) Detection method
JP2003098117A (en) Apparatus for detecting defect of image and its defect detection method
JP3106370B2 (en) Defect detection and type recognition of printed circuit boards using graph information
JP2819905B2 (en) Wiring pattern inspection equipment
JPH0763816A (en) Printed board inspection method and apparatus, centralized inspection system of prinrted board
Foster et al. Automated visual inspection of bare printed circuit boards using parallel processor hardware
JPH05108798A (en) Picture processing method and picture processor
JPH01296142A (en) Two-dimensional image comparison inspecting device
JPS62259184A (en) Detecting device for pattern crack
JPH09147056A (en) Method and device for checking appearance of mark
JPS62184579A (en) Pattern inspection device
JPH0812697B2 (en) Pattern defect inspection system
JPH03252546A (en) Wiring pattern inspection device