JP3036054B2 - Appearance inspection method - Google Patents
Appearance inspection methodInfo
- Publication number
- JP3036054B2 JP3036054B2 JP2300748A JP30074890A JP3036054B2 JP 3036054 B2 JP3036054 B2 JP 3036054B2 JP 2300748 A JP2300748 A JP 2300748A JP 30074890 A JP30074890 A JP 30074890A JP 3036054 B2 JP3036054 B2 JP 3036054B2
- Authority
- JP
- Japan
- Prior art keywords
- degree
- separation
- defective products
- determining
- inspection
- 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.)
- Expired - Fee Related
Links
Landscapes
- Length Measuring Devices By Optical Means (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Devices For Executing Special Programs (AREA)
- Image Analysis (AREA)
Description
【発明の詳細な説明】 産業上の利用分野 本発明は、検査物の画像情報より得られた特徴量を用
いた外観検査に関するものである。Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an appearance inspection using a feature obtained from image information of an inspection object.
従来の技術 ファジィ推論による外観検査は検査物の画像情報より
得られる各種特徴量を入力とし、検査結果を出力とす
る。この際、入力とする特徴量の種類やメンバシップ関
数は人間の勘や経験や試行錯誤によって決定される。2. Description of the Related Art Inspection by fuzzy inference involves inputting various features obtained from image information of an inspection object and outputting an inspection result. At this time, the type of the feature quantity to be input and the membership function are determined by human intuition, experience, and trial and error.
発明が解決しようとする課題 ファジィ推論による外観検査を行うにあたって、その
推論方法の決定方法の効率化が課題となっている。従来
の方法では、特徴量の選択あるいはメンバシップ関数の
決定に膨大な時間がかかる。Problems to be Solved by the Invention In performing an appearance inspection by fuzzy inference, it has been an issue to increase the efficiency of a method of determining an inference method. In the conventional method, it takes an enormous amount of time to select a feature amount or determine a membership function.
本発明は、ファジィ推論のための効率の良い特徴量の
選択あるいはメンバシップ関数の決定方法を提供するこ
とを目的とする。SUMMARY OF THE INVENTION It is an object of the present invention to provide a method for efficiently selecting a feature amount or determining a membership function for fuzzy inference.
課題を解決するための手段 本発明は、検査対象の画像情報より得られた特徴量の
良品と不良品の分離の度合いを表す分離度の決定を行
い、前記分離度を用いて検査する上で特徴量の選択を行
ない、前記選択された特徴量をファジィ推論を用いる
際、前記分離度の決定で用いた平均、標準偏差を用いて
メンバシップ関数を決定して検査結果を出力することを
特徴とする。Means for Solving the Problems The present invention determines the degree of separation indicating the degree of separation between non-defective and non-defective products of the feature amount obtained from the image information of the inspection target, and performs inspection using the degree of separation. Selecting a feature amount, and using fuzzy inference on the selected feature amount, determining a membership function using an average and a standard deviation used in the determination of the degree of separation, and outputting an inspection result. And
作用 この構成によると、人間の勘や経験や試行錯誤によら
ないファジィ推論のための効率の良い特徴量の選択とメ
ンバシップ関数の決定を行うことができる。Operation With this configuration, it is possible to select an efficient feature amount and determine a membership function for fuzzy inference without human intuition, experience, and trial and error.
実施例 以下、本発明の一実施例を説明する。第１図は、ある
１つの特徴量についての良品と不良品の分布を表してお
り、図中１は良品の分布、２は不良品の分布、３は良品
の平均、４は不良品の平均、５は良品の標準偏差、６は
不良品の標準偏差を示している。第２図は第１図で示し
た特徴量の良品と不良品のメンバシップ関数の決定方法
を表しており、図中７は良品のメンバシップ関数、８は
不良品のメンバショプ関数を示している。第３図，第４
図は検査対象となる印刷文字の例であり、第３図は良品
を示す図、第４図は矢印の示す位置ににじみを有する不
良品を示す図である。Example Hereinafter, an example of the present invention will be described. FIG. 1 shows the distribution of non-defective products and non-defective products with respect to a certain characteristic amount, wherein 1 is the distribution of non-defective products, 2 is the distribution of defective products, 3 is the average of non-defective products, and 4 is the average of defective products. 5 indicates the standard deviation of a good product, and 6 indicates the standard deviation of a defective product. FIG. 2 shows a method of determining a membership function between a good product and a bad product of the feature amount shown in FIG. . FIG. 3, FIG.
FIG. 3 shows an example of a printed character to be inspected. FIG. 3 is a diagram showing a non-defective product, and FIG. 4 is a diagram showing a defective product having a blur at a position indicated by an arrow.
まず、検査の基準を決定するために呈示された検査物
の画像情報より得られた種々の特徴により、検査に用い
る上でより効率のよいものを選択するための基準となる
パラメータとして、式（１）に示す分離度を用いる。First, according to various characteristics obtained from the image information of the inspection object presented for determining the inspection criterion, the following equation (5) is used as a parameter serving as a criterion for selecting a more efficient one for use in the inspection. The degree of separation shown in 1) is used.
但し、avg1,avg2は各々ある１種類の特徴量の良品及
び不良品の各々における平均値3,4、σ1,σ２は同様に
特徴量の良品及び不良品の各々における標準偏差5,6で
ある。分離度は各種特徴量ごとに計算され、その表す意
味を第１図に示す。この図からもわかるように分離度は
良品と不良品の特徴量の分布の分離の度合いを表すもの
であり、分離度が大きい程その特徴量における分離の度
合いは大きい。なお、第１図に示した特徴量の分離度は
1.6である。 Here, avg1 and avg2 are the average values 3,4 for each of the non-defective products and the non-defective products of one type of characteristic amount, and σ1, σ2 are the standard deviations 5,6 for the non-defective products and the defective product of the characteristic amounts, respectively. . The degree of separation is calculated for each feature amount, and its meaning is shown in FIG. As can be seen from this figure, the degree of separation indicates the degree of separation of the distribution of the feature values between good and defective products. The higher the degree of separation, the greater the degree of separation in the feature value. Note that the degree of separation of the feature amounts shown in FIG.
1.6.
ファジィ推論に用いる多数の良品および不良品サンプ
ルの画像より計算された各特徴量について分離度を計算
した結果より、分離度の大きいものから数十種類の特徴
量を選択し、ファジィ推論の入力とする。From the results of calculating the degree of separation for each feature calculated from the images of many good and bad samples used for fuzzy inference, select dozens of features from those with the highest degree of separation and input fuzzy inference. I do.
第４図のようなにじみの不良を検出するようにファジ
イ推論を行わせる。入力部に与える特徴量としては、分
離度より選ばれた文字部分の面積，周囲長，骨格長，周
囲角度分布，骨格角度分布、射影長，濃度ヒストグラム
分布等を用いる。Fuzzy inference is performed so as to detect a blurring defect as shown in FIG. As the feature amount given to the input unit, the area, perimeter, skeleton length, peripheral angle distribution, skeleton angle distribution, projection length, density histogram distribution, etc. of the character portion selected from the degree of separation are used.
次に、分離度を用いて選択された特徴量をファジィ推
論する際、第２図のように良品のメンバシップ関数７、
不良品のメンバシップ関数８を決定するための基準とな
るパラメータとして、良品の平均3,不良品の平均4,良品
の標準偏差5,不良品の標準偏差６を用いる。ここで、第
１図で示した特徴量がファジィ推論の入力として選択さ
れた場合を例にとる。Next, when fuzzy inference is performed on the feature quantity selected using the degree of separation, as shown in FIG.
As parameters serving as references for determining the defective product membership function 8, an average of good products 3, an average of defective products 4, a standard deviation of good products 5, and a standard deviation 6 of defective products are used. Here, a case where the feature quantity shown in FIG. 1 is selected as an input of fuzzy inference will be described as an example.
良品メンバシップ関数７の決定は、良品の平均３を中
心に良品の標準偏差５の幅だけ両側の適合度を1.0とし
て、次に良品の標準偏差５の幅で両側の適応度が0.0に
なるように直線で結ぶ。不良品のメンバシップ関数８も
同様にして決定する。The non-defective membership function 7 is determined by setting the fitness on both sides to the standard deviation of the non-defective product by the width of the standard deviation 5 around the mean 3 of the non-defective products, and then the fitness on both sides becomes 0.0 by the width of the standard deviation 5 of the non-defective product. Connect with a straight line. The defective product membership function 8 is similarly determined.
なお、本実施例において、ファジィ推論に入力する特
徴量を分離度の大きいものから数十種類選択したがその
数について限定しない。また、平均を中心に標準偏差の
幅だけ適合度を1.0とし、次に標準偏差の幅で適合度が
0.0になるように直線で結んだが適合度1.0の幅、適合度
1.0から0.0の直線の傾き、位置については限定しない。
また、適合度1.0から0.0の間は直線と限定しない。In the present embodiment, dozens of feature amounts to be input to fuzzy inference are selected from those having a large degree of separation, but the number is not limited. In addition, the fitness is set to 1.0 with the width of the standard deviation around the mean, and then the fitness is calculated with the width of the standard deviation.
Connected with a straight line so that it becomes 0.0, but the width of conformity 1.0, conformity
The slope and position of the straight line from 1.0 to 0.0 are not limited.
The degree of conformity between 1.0 and 0.0 is not limited to a straight line.
発明の効果 以上のように本発明によれば、ファジィ推論による外
観検査において、従来のような入力とする特徴量の種類
やメンバシップ関数を人間の勘や経験や試行錯誤によっ
て決定する方法に比べ、分離度を用い効率の良い特徴量
の選択とメンバシップ関数の決定を行なうことができ
る。Effect of the Invention As described above, according to the present invention, in the appearance inspection by fuzzy inference, compared with the conventional method of determining the type of the feature amount to be input and the membership function by human intuition, experience, and trial and error. In addition, it is possible to efficiently select a feature amount and determine a membership function by using the degree of separation.
第１図は分離度の示す意味を表す概念図、第２図はメン
バシップ関数の決定方法を表す図、第３図，第４図は検
査対象となる印刷文字の例であり、第３図は良品を示す
図、第４図は矢印の示す位置ににじみを有する不良品を
示す図である。 １……良品の分布、２……不良品の分布、３……良品の
平均、４……不良品の平均、５……良品の平均±標準偏
差、６……不良品の平均±標準偏差、７……良品のメン
バシップ関数、８……不良品のメンバシップ関数。FIG. 1 is a conceptual diagram showing the meaning of the degree of separation, FIG. 2 is a diagram showing a method of determining a membership function, and FIGS. 3 and 4 are examples of print characters to be inspected. FIG. 4 is a diagram showing a non-defective product, and FIG. 4 is a diagram showing a defective product having a blur at a position indicated by an arrow. 1 ... distribution of non-defective products, 2 ... distribution of defective products, 3 ... average of non-defective products, 4 ... average of defective products, 5 ... average ± standard deviation of non-defective products, 6 ... average ± standard deviation of defective products , 7... Good membership function, 8.
───────────────────────────────────────────────────── フロントページの続き (56)参考文献 特開 昭63−55677（ＪＰ，Ａ) 特開 平２−140886（ＪＰ，Ａ) (58)調査した分野(Int.Cl.7，ＤＢ名) G06T 7/00 G01B 11/24 G01N 21/88 G06F 9/44 554 ────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-63-55677 (JP, A) JP-A-2-140886 (JP, A) (58) Fields investigated (Int. Cl. 7 , DB name) G06T 7/00 G01B 11/24 G01N 21/88 G06F 9/44 554
Claims (1)
良品と不良品の分離の度合いを表す分離度の決定を行
い、前記分離度を用いて検査する上で特徴量の選択を行
ない、前記選択された特徴量をファジィ推論を用いる
際、前記分離度の決定で用いた平均、標準偏差を用いて
メンバシップ関数を決定して検査結果を出力する外観検
査方法。1. A method for determining a degree of separation between a non-defective product and a non-defective product based on feature information obtained from image information to be inspected, and selecting a feature value for the inspection using the separability. A visual inspection method for determining a membership function using an average and a standard deviation used in determining the degree of separation and outputting an inspection result when using the selected feature amount by fuzzy inference.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2300748A JP3036054B2 (en) | 1990-11-05 | 1990-11-05 | Appearance inspection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2300748A JP3036054B2 (en) | 1990-11-05 | 1990-11-05 | Appearance inspection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JPH04171587A JPH04171587A (en) | 1992-06-18 |
| JP3036054B2 true JP3036054B2 (en) | 2000-04-24 |
Family
ID=17888627
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP2300748A Expired - Fee Related JP3036054B2 (en) | 1990-11-05 | 1990-11-05 | Appearance inspection method |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JP3036054B2 (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH06195467A (en) * | 1992-12-25 | 1994-07-15 | Iwaki Electron Corp Ltd | Automatic mf function generating system |
| AU6802694A (en) * | 1993-05-28 | 1994-12-20 | Axiom Bildverarbeitungssysteme Gmbh | An automatic inspection apparatus |
| JP4528014B2 (en) * | 2004-04-05 | 2010-08-18 | 株式会社日立ハイテクノロジーズ | Sample inspection method |
-
1990
- 1990-11-05 JP JP2300748A patent/JP3036054B2/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| JPH04171587A (en) | 1992-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US5592572A (en) | Automated portrait/landscape mode detection on a binary image | |
| DE10317917B4 (en) | System and method for bounding and classifying regions within a graphic image | |
| US7609847B2 (en) | Methods and systems for determining object layouts | |
| DE1524323A1 (en) | Graphic analysis | |
| JPH08335271A (en) | Picture processing method of picture diagnosis support device | |
| US20060282777A1 (en) | Batch processing of images | |
| CN113012157A (en) | Visual detection method and system for equipment defects | |
| JP3036054B2 (en) | Appearance inspection method | |
| JP3009205B2 (en) | Inspection method and apparatus | |
| JP3708042B2 (en) | Image processing method and program | |
| US5649028A (en) | Connect-the-dots drawing production device | |
| JPH04346187A (en) | Quality decision method for subject to be detected | |
| JPH05282456A (en) | Pattern recognition device | |
| JPH10285399A (en) | Binarization method for image | |
| JPH09305757A (en) | Device and method for discriminating quantity of noise in image | |
| CN115564702A (en) | Model training method, system, device, storage medium and defect detection method | |
| JPH1049669A (en) | Method for processing picture | |
| JP2671818B2 (en) | Image processing device | |
| JP3462960B2 (en) | Image processing method | |
| JP4741289B2 (en) | Image processing apparatus and image processing method | |
| JP3361756B2 (en) | Extended graphic display attribute equal division method and apparatus, and recording medium storing extended graphic display attribute equal division program | |
| JPH103539A (en) | Method for processing picture | |
| JPH0773327A (en) | Broken line/chain line recognizing method for automatic drawing recognizing device | |
| JP2000155806A (en) | Character recognition method and its device and dictionary preparation method and its device and character quality judgment method and recording medium | |
| JP3193135B2 (en) | Dot pattern dot position detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| LAPS | Cancellation because of no payment of annual fees |