JPH0498471A - Parallel linear diagram extracting system - Google Patents

Parallel linear diagram extracting system

Info

Publication number
JPH0498471A
JPH0498471A JP2212632A JP21263290A JPH0498471A JP H0498471 A JPH0498471 A JP H0498471A JP 2212632 A JP2212632 A JP 2212632A JP 21263290 A JP21263290 A JP 21263290A JP H0498471 A JPH0498471 A JP H0498471A
Authority
JP
Japan
Prior art keywords
parallel
processing
line
extracted
azimuth
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.)
Granted
Application number
JP2212632A
Other languages
Japanese (ja)
Other versions
JP3080097B2 (en
Inventor
Taku Furukawa
卓 古川
Hideaki Arita
秀昶 有田
Yuzo Hirai
平井 有三
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.)
Nippon Steel Corp
Original Assignee
Nippon Steel Corp
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 Nippon Steel Corp filed Critical Nippon Steel Corp
Priority to JP02212632A priority Critical patent/JP3080097B2/en
Publication of JPH0498471A publication Critical patent/JPH0498471A/en
Application granted granted Critical
Publication of JP3080097B2 publication Critical patent/JP3080097B2/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Landscapes

  • Image Analysis (AREA)

Abstract

PURPOSE:To process at high speed by dedicated hardware, parallel processing and so on by extracting a parallel linear diagram by drawing picture data. a convolution integral, threshold value processing and a logical operation between the picture data. CONSTITUTION:The linear diagram of a specified azimuth is extracted by execut ing a convolution integral by a linear detecting matrix and threshold processing. Next, three checking points are taken in the azimuth right-angled to the azimuth of a load matrix on a picture to extract the linear diagram of the specified azimuth, on a same line and at equal distance, a logical operation is executed between the picture element values of the points and the parallel linear diagram of the specified azimuth in the drawing is extracted. The processing is executed on the plural azimuths and when the total sum or the logical sum of the result of is is taken, the parallel line of any azimuth can be extracted. Thus, high speed processing is attained by making into hardware of parallel calculation.

Description

【発明の詳細な説明】 〔産業上の利用分野〕 本発明は、計算機に読み込まれた地図、機械図面等の図
面画像データ中の平行線図形の抽出に関する。
DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to the extraction of parallel line figures from drawing image data such as maps and machine drawings read into a computer.

[従来の技術〕 平行線の抽出法は、地図図形処理の道路図形等の抽出に
おいて検討されている。平行線の追跡方法としては、平
行線の内側を2本のベクトルで並行に追跡しながら、ベ
クトルの内積をとり平行線を抽出するもの(電子通信学
会論文誌、シo1.J67DNo12. ppH429
−11426,1984,12) 、画素の輪郭線を画
素が常に右側に(るようにとると、平行線の候補となる
ベクトル(ペアベクトル)の向きが、逆向きになること
を利用して、平行線を求める方法などがあるが(AUT
OCARTOJAPAN論文集、昭62)、いずれにし
ても、一般に平行線の検出処理は難しく、処理は複雑で
高速処理は困難であった。
[Prior Art] Parallel line extraction methods have been studied in the extraction of road figures and the like in map figure processing. A method for tracing parallel lines is to trace the inside of the parallel line in parallel with two vectors and extract the parallel line by taking the inner product of the vectors (Transactions of the Institute of Electronics and Communication Engineers, No. 1. J67D No. 12. ppH429)
-11426, 1984, 12), by taking advantage of the fact that if the outline of a pixel is taken so that the pixel is always on the right side, the directions of vectors (paired vectors) that are candidates for parallel lines will be opposite, There are ways to find parallel lines (AUT
OCARTO JAPAN Collection of Papers, 1986) In any case, parallel line detection processing is generally difficult, the processing is complex, and high-speed processing is difficult.

〔発明が解決しようとする課題〕[Problem to be solved by the invention]

図面を認識・理解、特に地図図形認識の道路図形抽出と
いった処理では、様々な種類の線図形が含まれている図
面から、平行線を抽出する処理が必要となってくる。平
行線の抽出は正確さや、扱うデータ量の膨大さから、処
理の高速化が要求される。
In the process of recognizing and understanding drawings, especially the extraction of road figures for map figure recognition, it is necessary to extract parallel lines from drawings that include various types of line figures. Parallel line extraction requires high-speed processing due to accuracy and the huge amount of data to be handled.

本発明では、基本的な処理を組み合わせて、正確に平行
線の抽出ができ、ハードウェア化、あるいは並列計算に
よって高速処理が可能となる平行線図形の抽出方式を提
供することを目的とする。
An object of the present invention is to provide a method for extracting parallel line figures that can accurately extract parallel lines by combining basic processing and that enables high-speed processing through hardware or parallel calculation.

[課題を解決するための手段〕 図面の画像データに対して荷重マトリックスとの畳み込
み積分を行い、決められた閾値未満の画素値をOとする
閾値処理と、画像データの画素対応での論理演算を行う
[Means for solving the problem] Perform convolution integration with a weight matrix on the image data of the drawing, perform threshold processing in which pixel values less than a predetermined threshold are set to O, and logical operations based on pixel correspondence of the image data. I do.

畳み込み積分は、直線状に中心部が正の値で周辺部が負
の値からなる荷重マトリックス(第2図)を用意し、こ
れらの荷重マトリックスと図面画像データとの積和演算
によって畳み込み積分を行い、図面中の特定方位の線図
形を抽出する。
The convolution integral is performed by preparing a linear weight matrix (Figure 2) consisting of positive values at the center and negative values at the periphery, and calculating the convolution integral by multiplying and adding these weight matrices and the drawing image data. and extract line figures in a specific direction in the drawing.

論理演算は、荷重マトリックスの方位と直角の方位に、
同一線上、等距離に3点の検査点(第3図)をとり、こ
れらの点の画素値の間で論理演算を行うことで、図面中
の平行線図形を抽出する。
The logical operation is, in the direction perpendicular to the direction of the load matrix,
Parallel line figures in the drawing are extracted by taking three inspection points (FIG. 3) on the same line and at equal distances and performing logical operations between the pixel values of these points.

〔作 用〕[For production]

本発明の平行線抽出方式は、図面画像データとの畳み込
み積分と閾値処理と画像データ間の論理演算によって平
行線図形を抽出するものである。
The parallel line extraction method of the present invention extracts parallel line figures through convolution with drawing image data, threshold processing, and logical operations between image data.

直線検出マトリックスによる畳み込み積分、閾値処理を
実行することによって特定方位の線図形を抽出する。次
に特定方位の線図形を抽出した画像について荷重マトリ
ックスの方位と直角の方位tこ、同一線上、等距離に3
点の検査点をとり、これらの点の画素値の間で論理演算
を行い、図面中の特定方位の平行線図形を抽出する。
A line figure in a specific direction is extracted by performing convolution integration and threshold processing using a straight line detection matrix. Next, for the image in which the line figure in a specific direction has been extracted, the direction t perpendicular to the direction of the load matrix is placed on the same line and at the same distance.
Point inspection points are taken, logical operations are performed between the pixel values of these points, and a parallel line figure in a specific direction in the drawing is extracted.

以上の処理を複数の方位について行い、その結果の総和
、あるいは論理和をとればあらゆる方位の平行線を抽出
できる。
By performing the above processing for multiple directions and calculating the sum or logical sum of the results, parallel lines in any direction can be extracted.

(実施例) 第1図は本発明の一実施例を説明するための図であり、
第2図は第1図の平行線抽出部■〜■の各部の処理手順
を説明するための図である。平行線分は以下に記す3つ
の部分(畳み込み積分(1)、閾値処理(2)、論理演
算(3))の処理で検出する。
(Example) FIG. 1 is a diagram for explaining an example of the present invention,
FIG. 2 is a diagram for explaining the processing procedure of each of the parallel line extraction units ① to ② in FIG. 1. Parallel line segments are detected by processing of three parts (convolution integral (1), threshold processing (2), and logical operation (3)) described below.

第6図は幅が6.12.24画素である同心円の図形(
a)から幅12画素の平行線の抽出処理を行った結果を
示したものである。まず、畳み込み積分(1)で特定方
位の線図形の抽出を行う。
Figure 6 shows a concentric circle whose width is 6, 12, and 24 pixels (
This figure shows the result of extracting parallel lines with a width of 12 pixels from a). First, a line figure in a specific direction is extracted using convolution (1).

線図形は第3図のような直線検出用のマトリックスで入
力画像データとの積和演算を実行する。マトリックスの
各要素は直線状に中心部が正で周辺部が負であり、それ
らの値はマトリックスの方位と垂直方向に2つのGau
ss関数の差(DOG間数)で算出する。
The line figure is a matrix for straight line detection as shown in FIG. 3, and a product-sum operation is performed with the input image data. Each element of the matrix is linearly positive at the center and negative at the periphery, and their values are divided by two Gauss in the direction perpendicular to the orientation of the matrix.
Calculate using the difference in ss functions (number of DOGs).

DOG(x) −(1/□ o +) ・exp(x2
/2 ct +”)<l/F7Tty z) ・exp
(−x2/2 tt z”)  (1)ここで、σ1.
σ2:  Gauss関数の標準偏差(σ2/σ、・1
,6)、X;中心線からの距離とする。畳み込み積分は
第(1)式でサイズMXNの荷重マトリックスを作成し
くN/2=3σ)、このマトリックスと画像データの間
で積和をとるものとする。
DOG(x) −(1/□ o +) ・exp(x2
/2 ct +”)<l/F7Tty z) ・exp
(-x2/2 tt z”) (1) Here, σ1.
σ2: Standard deviation of Gaussian function (σ2/σ, ・1
, 6), X: Distance from the center line. In the convolution integral, a weight matrix of size MXN is created using equation (1) (N/2=3σ), and the sum of products is calculated between this matrix and image data.

荷重マトリックスの方位が第3図に示す方位の場合、処
理対象画像をf(i、j)、荷重マトリックスをw(i
、j)、結果画像をg(i、j)とすると、閾値処理(
2)では決められた閾値以下の数値を0にする。これに
よりマトリックスの方位と同一の方位の線分を検出する
。第6図(b)は垂直方位の線図形の抽出を行った結果
である。
When the orientation of the load matrix is as shown in Fig. 3, the image to be processed is f(i, j) and the load matrix is w(i
, j), and the resulting image is g(i, j), threshold processing (
In 2), values below a predetermined threshold are set to 0. As a result, line segments having the same orientation as the matrix are detected. FIG. 6(b) shows the result of extracting line figures in the vertical direction.

次に、論理演算(3)では、畳み込み積分後の画像につ
いて平行か否かの判定を以下の方法で行う。まず、対象
画像の中に着目点をとり、上記の直線検出用マトリック
スと直角の方向に着目点の両側に特定の距離!に検査点
X1−1.Xt++をとる(第4図)。
Next, in logical operation (3), it is determined whether or not the images after convolution are parallel or not using the following method. First, take a point of interest in the target image, and set a specific distance on both sides of the point of interest in a direction perpendicular to the straight line detection matrix above! Inspection point X1-1. Take Xt++ (Figure 4).

上記の着目点が互いに平行な線分上に存在するのであれ
ば、左右の検査点X1−1+ Xi4.はどちらかに存
在すればよい。したがって、点Xの画素値をP (x)
とし、P (x)が二値(0,1)の場合、着目点が平
行な線分上に存在すれば、 P(xt) AND (P(xt−+) ORP(xt
−+))= 1    (3)となる。この式は画素値
P (x)が多値となる場合は、閾値をpとして、 MIN (P(xt)、 MAX(P(X=−+)+ 
P(xt−+)) )≧P平行線が互いに平行な2本の
線分と限定された場合は、左右の検査点X1−1+ X
i。1は同時に存在しないはずである。したがって、点
Xの画素値P(X)が二値であれば、 P(xX) AND (p(x6−+) XORP(X
!、1))= 1    (5)となり、P (x)が
多値となる場合は、MIN [P(Xi)l (MAX
 (P(Xi−1)I P(Xi41))MIN (P
(J−1)、 P(x;、+)) ) ]≧p  (6
)で表記できる。これらの3点の画素値から以上の式の
値を計算し、出力画素値とすれば、平行線が抽出される
。しかしながら第5図(a)のような図形では、式(5
)あるいは(6)によっても、両端の線が残る(第5図
(b))。この場合、以上の処理を2回掛けることで両
端の線は消去することができる。第6図(C)は抽出さ
れた垂直方位の線図形について論理演算を行い、幅12
画素の平行線を抽出した結果である。
If the above points of interest are on line segments parallel to each other, the left and right inspection points X1-1+Xi4. should exist in either one. Therefore, the pixel value of point X is P (x)
If P (x) is binary (0, 1), and the point of interest is on a parallel line segment, then P(xt) AND (P(xt-+) ORP(xt
−+))=1 (3). When the pixel value P (x) is multi-valued, this formula is expressed as MIN (P(xt), MAX(P(X=-+)+
P(xt-+)) ) ≧P If the parallel lines are limited to two line segments parallel to each other, the left and right inspection points X1-1+
i. 1 should not exist at the same time. Therefore, if the pixel value P(X) at point X is binary, P(xX) AND (p(x6-+) XORP(X
! , 1)) = 1 (5), and when P (x) is multivalued, MIN [P(Xi)l (MAX
(P(Xi-1)I P(Xi41))MIN (P
(J-1), P(x;, +)) ]≧p (6
) can be written. If the value of the above equation is calculated from the pixel values of these three points and used as the output pixel value, parallel lines will be extracted. However, in a figure like FIG. 5(a), the equation (5
) or (6), lines at both ends remain (Fig. 5(b)). In this case, the lines at both ends can be erased by performing the above process twice. FIG. 6(C) shows that logical operations are performed on the extracted line figures in the vertical direction, and the width is 12.
This is the result of extracting parallel lines of pixels.

以上のようにして1つの方位の平行線図形を抽出するこ
とができるが、あらゆる方向の平行線図形を抽出するに
は、以上の処理を複数方位について平行線抽出部■〜■
で行い、それぞれの結果の総和あるいは論理和を画像間
演算部■でとる必要がある(第1図)。
Parallel line figures in one direction can be extracted as described above, but in order to extract parallel line figures in all directions, the above process can be performed in the parallel line extraction section for multiple directions.
It is necessary to calculate the sum or logical sum of the respective results in the inter-image calculation section (Fig. 1).

この実施例では、直線検出用のマトリックスはπ/8毎
の角度で8方位のマトリックスを用意した。以上の処理
を8方位について行い、これらの結果画像の総和をとっ
た(第6図(d))。
In this embodiment, a matrix for detecting a straight line is prepared in eight directions at angles of π/8. The above processing was performed for eight directions, and the resulting images were summed (FIG. 6(d)).

〔発明の効果〕〔Effect of the invention〕

本発明によれば、地図、機械図面などの図面上の図形か
ら平行線を抽出する際、いかなる形状の図形についても
、積和演算、閾値処理、論理演算といった基本的な画像
処理の手法を組み合わせて処理することができるので、
専用ハードウェアや並列処理等で高速に処理することが
可能となる。
According to the present invention, when extracting parallel lines from figures on drawings such as maps and mechanical drawings, basic image processing techniques such as product-sum operations, threshold processing, and logical operations are combined to extract parallel lines from figures of any shape. Because it can be processed by
It becomes possible to perform high-speed processing using dedicated hardware, parallel processing, etc.

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

第1図は本発明の詳細な説明した系統図、第2図は平行
線抽出処理の手順の説明図、第3図は直線検出用マトリ
ックス、第4図は論理演算の説明図、第5図は複数の平
行線での処理を説明した図、第6図(a)〜(d)は異
なる幅の平行線図形が描かれた画像について平行線抽出
処理を実施した結果を示した画像である。 なお、図面に用いた符号において、 1・・・・・・・・・・・・・・・畳み込み積分2・・
・・・・・・・・・・・・・閾値処理3・・・・・・・
・・・・・・・・論理演算である。
Fig. 1 is a detailed system diagram of the present invention, Fig. 2 is an explanatory diagram of the procedure of parallel line extraction processing, Fig. 3 is a matrix for straight line detection, Fig. 4 is an explanatory diagram of logical operations, and Fig. 5 is a diagram explaining processing with multiple parallel lines, and Figures 6 (a) to (d) are images showing the results of parallel line extraction processing on images in which parallel line figures of different widths are drawn. . In addition, in the symbols used in the drawings, 1......Convolution integral 2...
・・・・・・・・・・・・Threshold processing 3・・・・・・・
......It is a logical operation.

Claims (3)

【特許請求の範囲】[Claims] (1)図面の画像データに対して荷重マトリックスとの
畳み込み積分を行い、決められた閾値未満の画素値を0
とする閾値処理を行い、画像データの画素対応での論理
演算を行うことを特徴とする平行線図形抽出方式。
(1) Convolution integration with the weight matrix is performed on the image data of the drawing, and pixel values below a predetermined threshold are zeroed out.
A parallel line figure extraction method characterized by performing threshold processing and performing logical operations based on pixel correspondence of image data.
(2)前記畳み込み積分は、直線状に中心部が正の値で
周辺部が負の値からなる荷重マトリックスを用意し、こ
れらの荷重マトリックスと図面画像データとの積和演算
によって畳み込み積分を行い、図面中の特定方位の線図
形を抽出することを特徴とする請求項1記載の平行線図
形抽出方式。
(2) The convolution integral is performed by preparing a linear load matrix with positive values at the center and negative values at the periphery, and performing the convolution integral by multiplying and adding these load matrices and the drawing image data. 2. The parallel line figure extraction method according to claim 1, wherein a line figure in a specific direction in a drawing is extracted.
(3)前記論理演算は、荷重マトリックスの方位と直角
の方位に、同一線上、等距離に3点の検査点をとり、こ
れらの点の画素値の間で論理演算を行うことで、図面中
の平行線図形を抽出することを特徴とする請求項1また
は2記載の平行線図形抽出方式。
(3) The above logical operation is performed by taking three inspection points on the same line and at equal distances in a direction perpendicular to the direction of the load matrix, and performing a logical operation between the pixel values of these points. 3. The parallel line figure extraction method according to claim 1, wherein a parallel line figure is extracted.
JP02212632A 1990-08-10 1990-08-10 Parallel line figure extraction method Expired - Fee Related JP3080097B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP02212632A JP3080097B2 (en) 1990-08-10 1990-08-10 Parallel line figure extraction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP02212632A JP3080097B2 (en) 1990-08-10 1990-08-10 Parallel line figure extraction method

Publications (2)

Publication Number Publication Date
JPH0498471A true JPH0498471A (en) 1992-03-31
JP3080097B2 JP3080097B2 (en) 2000-08-21

Family

ID=16625877

Family Applications (1)

Application Number Title Priority Date Filing Date
JP02212632A Expired - Fee Related JP3080097B2 (en) 1990-08-10 1990-08-10 Parallel line figure extraction method

Country Status (1)

Country Link
JP (1) JP3080097B2 (en)

Also Published As

Publication number Publication date
JP3080097B2 (en) 2000-08-21

Similar Documents

Publication Publication Date Title
CN102209974B (en) Feature value extracting device, object identification device, and feature value extracting method
Wang et al. Recognition and location of the internal corners of planar checkerboard calibration pattern image
CN101639947A (en) Image-based plant three-dimensional shape measurement and reconstruction method and system
CN102722731A (en) Efficient image matching method based on improved scale invariant feature transform (SIFT) algorithm
US8121415B2 (en) Combining feature boundaries
CN105894521A (en) Sub-pixel edge detection method based on Gaussian fitting
CN112233076B (en) Structural vibration displacement measurement method and device based on red round target image processing
CN106529548A (en) Sub-pixel level multi-scale Harris corner detection algorithm
CN115546223A (en) Method and system for detecting loss of fastening bolt of equipment under train
CN115457277A (en) Intelligent pavement disease identification and detection method and system
CN113989653B (en) Method for extracting river plane geometric index and topological structure
Khosravi et al. Vehicle speed and dimensions estimation using on-road cameras by identifying popular vehicles
CN111062415A (en) Target object image extraction method and system based on contrast difference and storage medium
JPH0498471A (en) Parallel linear diagram extracting system
CN103761725B (en) A kind of video plane detection method based on innovatory algorithm
CN113095359B (en) Method and system for detecting radiographic image marking information
CN110705479A (en) Model training method, target recognition method, device, equipment and medium
CN112967303B (en) Image edge detection and edge precision verification method thereof
CN103759647B (en) Module sectional construction face member geometric properties measuring method
JPH0498470A (en) Classifying system for line width of linear diagram
CN110349174B (en) Sliding rail multi-parameter measurement method and measurement device
US20220019797A1 (en) Methods and systems for agricultural block mapping
Somawirata et al. Road Detection Based on Region Grid Analysis Using Structural Similarity
Wang et al. 3D shape measurement based on computer vision
CN116740153A (en) Quick registration method for agile satellite remote sensing image based on surface texture reference plane

Legal Events

Date Code Title Description
LAPS Cancellation because of no payment of annual fees