CN102880868A - Engineering drawing vector conversion and primitive semantic extraction method - Google Patents
Engineering drawing vector conversion and primitive semantic extraction method Download PDFInfo
- Publication number
- CN102880868A CN102880868A CN2012102773683A CN201210277368A CN102880868A CN 102880868 A CN102880868 A CN 102880868A CN 2012102773683 A CN2012102773683 A CN 2012102773683A CN 201210277368 A CN201210277368 A CN 201210277368A CN 102880868 A CN102880868 A CN 102880868A
- Authority
- CN
- China
- Prior art keywords
- pel
- engineering drawing
- extracting method
- length
- point
- 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
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides an engineering drawing vector conversion and primitive semantic extraction method. The method comprises the following steps: S1, utilizing a scripting language to extract information data in a PDF (Portable Document Format) drawing document or generate corresponding BMP (Bitmap) image data, and further pre-processing the document in the image format; S2, utilizing the seed segment algorithm to vectorize straight-line primitives in the image data; S3, vectorizing circles/circular arcs in the image data; S4, identifying the primitive relationship and utilizing the text/graphics separation algorithm based on connected region to remove labeling frames and graphics information on the drawing images; and S5, outputting the result. Through pretreatment of the text/graphics separation algorithm and the improvement on the seed segment vector conversion method, the engineering drawing vector conversion and primitive semantic extraction method improves the vector conversion accuracy and processing speed, and better solves the difficult problems in extracting the pattern semantic information through the rule matching algorithm, thereby providing a better data foundation for intelligent machine drawing reading and three-dimensional reconstruction.
Description
Technical field
The present invention relates to engineering drawing pel semantic information extracts and the drawing vectorization technical field.
Background technology
Raster image vector quantized (Raster to Vector Conversion, referred to as RVC), utilize exactly the complex arts such as computer graphics, Digital Image Processing, artificial intelligence, the engineering drawing with the preservation of raster image form that is input to by scanner in the computing machine, be identified as vector graphics, the process that the form that can identify with CAD software is at last preserved.The raster graphic vector quantization not only in mechanical design field, more is with a wide range of applications in industries such as various engineering designs, geography information, has important theory and practice meaning.
In addition, except primitive information, a Practical Project drawing mainly is made of three parts such as graphic element, size marking, literal.Can be by following formal representation: P={E, D, T} wherein, the P--engineering drawing, the E--graphic element, the D--size marking, the various information structures that this three part of T--textual description is expressed the main contents of engineering drawing semantics.Explanatory note wherein partly comprises the information such as some materials, precision, processing request of product.General these explanations do not relate to the three-dimensional body information of part, so the pattern semanteme on the ordinary meaning refers to figure and size marking two aspects.Extract and the literal extraction by mark, the basis can be provided for the three-dimensional reconstruction of the drawing data behind the vector quantization.
Common vectorization method has: based on the algorithm of refinement, dynamically Framework Arithmetic, outline algorithm, mesh model algorithm and based on distance of swimming algorithm etc. and based on the kind subsegment algorithm of integral body identification etc., also have simultaneously artificial intelligence technology is fused in the middle of the vectorization method, such as nerual network technique and fuzzy recognition technology etc.
The method that the pel extraction of semantics relates to mainly contains rule match, picture and text separation, three contents of literal identification.Rule match is to utilize fixing rule judgment to go out the pels such as marking line, outer molded line, arrow, blanking line; The fundamental purpose that picture and text separate is from the engineering drawing that picture and text mix, and all character marks are separated, and belongs in essence a kind of image segmentation algorithm, usually utilizes contour feature to judge; Literal identification mainly utilizes the characteristic quantity algorithm to realize.
Have at present the vpstudio software of Germany, the multiple drawing vectorization softwares such as the R2v software of Able company, these softwares can better carry out the vector quantization of drawing to a certain extent, but they are also imperfect, the accuracy rate of vector quantization is not very high, has various weak points.On the pel extraction of semantics, effect is all poor, can not accurately identify the contents (all being that literal is identified as the multistage straight line basically) such as arrow, mark simultaneously.
Summary of the invention
The object of the present invention is to provide a kind of engineering drawing identification of different-format and engineering drawing vectorization and pel semantic extracting method of output of solving.
For solving the problems of the technologies described above, engineering drawing vectorization of the present invention and pel semantic extracting method comprise the steps:
S1 utilizes the information data in the script extraction PDF drawing file or generates corresponding BMP format-pattern data, and the file of picture format is carried out further pre-service;
S2 utilizes kind of a subsegment algorithm that the straight line pel in the view data is carried out vector quantization;
S3 carries out vector quantization to the circle/circular arc in the view data;
S4, the pel relation recognition is utilized based on callout box, Word message on the Algorithm for Text/Photo Separation removal line drawing of connected region;
S5, Output rusults.
Described kind of subsegment algorithm comprises the steps: S2.1, search kind of subsegment; S2.2 plants the subsegment growth; S2.3 has identified wiping of straight line.
Described search kind subsegment comprises the steps:
S2.1.1 makes a square window centered by black pixel C, from the coboundary, the continuously black pixel fragment of getting in the direction of the clock on the window edge is candidate segment;
S2.1.2 is the candidate segment of L with untreated, length, and P seeks kind of a subsegment as initial point with its mid point;
S2.1.3 from P, presses the direction that CP determines, surveys the black pixel fragment length on the orthogonal directions take step-size in search r as step-length, and such as black pixel fragment length≤L, be the rule section, otherwise be irregular section, and take the current rule section mid point that records as Q;
S2.1.4 checks that the black pixel between P, the Q is communicated with the district;
S2.1.5 finishes to seek.
Among the described S2.1.2, if without untreated candidate segment, such as search radius R＜maximum search scope scope, R=R+r then, and skip to S2.1.1.Among the described S2.1.2, if without untreated candidate segment, such as search radius R 〉=maximum search scope scope, then finish to seek.
Among the described S2.1.3, the regular hop count n that records continuously; The irregular hop count m that records continuously, establishing the current rule section mid point that records is Q, if n * r 〉=K * L then enters S2.1.4.Among the described S2.1.3, the regular hop count n that records continuously; The irregular hop count m that records continuously, establishing a current rule section mid point that records is Q, if m * r＞scope or | CP|＞scope, then finish to seek and return S2.1.2.
Among the described S2.1.4, qualified if the black pixel between P, the Q is communicated with the district, then finish to seek, return successfully sign.Among the described S2.1.4, defective if the black pixel between P, the Q is communicated with the district, then return S2.1.3.
Described kind of subsegment growth comprises the steps:
S2.2.1 generates the axial orthogonal directions of seed segment length path with the Bresenham algorithm;
S2.2.2 generates the track path that begins from the central point of kind of subsegment by the long axis direction of kind of subsegment with the pointwise of Bresenham algorithm, point of every generation checks whether the pixel of this point correspondence is stain.
In described S2.2.2, as check that pixel corresponding to this point is stain, then add up the length of continuous stain, if length greater than noise threshold, then will the length of continuous white point be charged to path segment table before detecting stain. check that the quadrature at current path point place is communicated with length.In described S2.2.2, as check that pixel corresponding to this point is not stain, the length that then adds up continuous white point will be if length greater than noise threshold, then will the length of continuous stain be charged to path segment table before detecting white point.In described S2.2.2, if the length of continuous stain then finishes the tracking on the current direction greater than the threshold value of dotted line spacing before detecting white point.
Describedly identified wiping of straight line and comprised the steps:
S2.3.1 along the connection length on the path testing path orthogonal directions of straight line, has been divided into the intersection point section with the path and without the intersection point section, has wiped fully by live width without the intersection point section;
S2.3.2, branch's situation and the local configuration thereof of the center line both sides of detection intersection point place straight line.
In described S2.3.2, branch into as described both sides branch, then the trend with the both sides profile calculates the approximate border of wiping.In described S2.3.2, branch into as described one-sided branch, then a branchiess side is erased to the center line of straight line, opposite side calculates with the trend of profile.
Described S3 utilizes circle and the circular-arc information in the least square method approximate algorithm extraction straight-line segment.Described S4 utilizes rule matching algorithm to extract the semantic informations such as arrow in straight-line segment, the arc section, marking line, boost line, blanking line, outer molded line; Utilize the OCR technology to carry out literal identification to wiping the image of identifying pel, utilize rule matching algorithm that literal is related with primitive information.Described S5 utilizes the XML form, and pel and the semantic information thereof of vector quantization export the file preservation to all.
Engineering drawing vectorization of the present invention and pel semantic extracting method pass through the pre-service to Algorithm for Text/Photo Separation, and to kind of the improvement of subsegment vectorization method, have improved accuracy rate and the processing speed of vector quantization; By rule matching algorithm, solved preferably the extraction difficult problem of pattern semantic information simultaneously, for machine intelligence interpreting blueprints and three-dimensional reconstruction provide preferably data basis.
Description of drawings
Fig. 1 is engineering drawing vectorization of the present invention and pel semantic extracting method process flow diagram;
Fig. 2 A is engineering drawing vectorization of the present invention and pel semantic extracting method Line vectorization search kind of subsegment process flow diagram;
Fig. 2 B is engineering drawing vectorization of the present invention and pel semantic extracting method Line vectorization kind subsegment growth process flow diagram;
Fig. 2 C is the process flow diagram of wiping that engineering drawing vectorization of the present invention and pel semantic extracting method Line vectorization have been identified straight line.
Embodiment
Below in conjunction with accompanying drawing engineering drawing vectorization of the present invention and pel semantic extracting method are described in further detail.
As shown in Figure 1 and Figure 2, engineering drawing vectorization of the present invention and pel semantic extracting method utilize the Ghostscript script to extract the information data in the PDF drawing file or generate corresponding BMP format-pattern data; To the further pre-service of the file of picture format (denoising and binaryzation); Utilize improved kind of all straight-line segment primitive information in the subsegment algorithm extraction image.
Seek the algorithm of planting subsegment since a black pixel c by definition as follows:
R is step-size in search, and scope is the maximum search scope, all determines according to drawing scanning resolution.Initially establish R=r.
Step1.1 makes a length of side as the square window of 2 * R centered by black pixel c, from the coboundary, the length of getting in the direction of the clock on the window edge is candidate segment near the continuously black pixel fragment of live width scope.
Step1.2 if untreated candidate segment (length is L) is arranged, seeks kind of a subsegment with mid point p as initial point.If without untreated candidate segment, such as R＜scope, R=R+r then, and skip to Step1.1; If R 〉=scope finishes to seek, return failure flags.
Step1.3 from p, presses the direction (upper and lower, left or right) that cp determines, surveys the black pixel fragment length on the orthogonal directions take r as step-length, and length approaches or is called the rule section less than the black pixel fragment of L, otherwise is irregular section.If the regular hop count of n for recording continuously; The irregular hop count of m for recording continuously.If the current rule section mid point that records is q, if n * r 〉=K * L then enters Step1.4; Otherwise, if m * r〉scope or | cp|〉scope, then finish to seek, return Step2.After several irregular section, when entering again the rule section, p is navigated to q, continue to seek, make search can skip noise among a small circle, damaged or intersection point.
Step1.4 checks p according to kind of a subsegment definition, and whether the black pixel between the q is communicated with the district qualified.If qualified, then finish to seek, return successfully sign; Otherwise, return Step1.3 and continue to seek.
This algorithm can begin by any one black pixel from image, seeks the kind subsegment of a nearest any direction straight line of (if search area memory exists).The long axis direction of the kind subsegment that finds can be used as the inceptive direction of this straight line, and minor axis length can be used as the initial value of live width.
Improvement to this algorithm is mainly reflected on the choosing of initial step length.Step-length is excessive, can omit tiny straight line, and step-length is too small, can affect program efficiency.By to selected window from square to rectangular change (namely the square of 2 * R becomes the rectangle of W * H), can be from less step-length, quick location kind of a subsegment.
Plant the growth algorithm of subsegment.
The growth of planting subsegment is realized by the directed tracing technology based on the Bresenham scan conversion algorithm.The Bresenham algorithm is for the sequence of the point on the straight line path that generates fast any direction.Track path extends towards two opposite directions along the direction of major axis from the major axis of kind of subsegment.As long as satisfy following two conditions, plant subsegment and just can grow:
Condition 1, the length of continuous white point section is less than the threshold value of dotted line spacing on the track path;
Condition 2, on the track path continuously on the stain section every bit place with the path orthogonal directions be communicated with that length approaches or greater than kind of a subsegment live width.
It is damaged that condition 1 can guarantee to skip little bitmap, and have the ability of following the tracks of dotted line.Condition 2 guarantees that the bitmap that extends is communicated with the feature that comprises straight line to be known in the district, and can by crossing, adhesion is regional.
Path segment table is the chained list of black and white section continuous on the record path, acts on the tracing process of both direction, is initialized as sky; L and R are the counters of the black pixel that is communicated with path point of track path both sides, act on the tracing process of a direction.
Step2.1 generates the axial orthogonal directions of seed segment length path Vo[i with the Bresenham algorithm] (i=0 ... n).The mid point in path is at O (planting the central point of subsegment) and n=3 * W (W is the width of kind of subsegment).Initialization L and R are 0.
Step2.2 generates the track path that begins from O by the long axis direction of kind of subsegment with the pointwise of Bresenham algorithm.Point of every generation checks whether pixel corresponding to this point is stain.If so, then enter Step2.3, otherwise, Step2.4 entered.
Step2.3, the length of the continuous stain of accumulative total is if length greater than noise threshold, is then charged to path segment table with the length of upper one white section (as existing).Inspection is communicated with length at the quadrature at current path point P place.At this moment, the orthogonal directions path Vp at P point place needn't regenerate, and can obtain from the Vo translation transformation Vp[i]=Vo[i]+offset (i=0 ... n, offset=P-O).From mid point Vp[n/2] begin to check that to both sides black section on the Vp is communicated with length, if the total length approximate W of both sides, then the length with both sides is added to respectively L and R.When the difference of L and R during greater than the smaller value among L and the R, show that there are deviation in the direction of kind of subsegment and actual rectilinear direction, need to after adjusting, again follow the tracks of.Plant the dynamic adjustment of subsegment and guarantee that kind of subsegment can extend to maximum length by corresponding bitmap.
Step2.4, the length of the continuous white point of accumulative total is if length greater than noise threshold, is then charged to path segment table with the length of upper one black section (as existing).If length greater than the threshold value of dotted line spacing, then finishes the tracking on the current direction.
Whole tracing process does not relate to any concrete direction take the direction of kind of subsegment as parameter, therefore the straight line of any direction is all had identical processing power.The tracing process of other direction similarly, and is just different in the mode of the section of charging in path segment table.A direction is added the table tail to, and other direction adds gauge outfit to, and the 1st section of both direction will be merged into a black section.When follow the tracks of to the both direction of kind of subsegment finish after, recorded black, the white segment length of the order that whole extension path passes through in the path segment table.If only there is a black section, then this section is a solid line; Otherwise, whether exist with dotted modes similarly regularly between black by analyzing, white section, just can realize the one-off recognition to dotted line.After straight line was intactly identified, live width can be obtained by the mean value calculation of the quadrature segment length that is similar to W on the whole path accurately.
It is that two end points directions are done respectively growth that the improvement of this algorithm is mainly simple interest, simultaneously the quadrature in the more STEP3 consuming time being communicated with length check is optimized, employing checks the straight line correctness with the Bresenham algorithm of live width, the kind subsegment that has the deviation of directivity is not done the direction adjustment, directly think kind of a subsegment grown junction bundle, thereby raise the efficiency.
Wipe the bitmap of identifying straight line.
For an isolated straight-line segment, the operation of wiping bitmap by live width is easy to realize.But when straight line intersects with other figures or literal, intersection point
Diversity complicates the issue.In fact, owing to not yet identifying with its object that intersects, it is unpractiaca accurately calculating the part that the intersection point place should keep.Thereby GLV has adopted an approximate data that detects based on the profile of antinode branch bitmap.
Step3.1 along the connection length on the path testing path orthogonal directions of straight line, has been divided into the intersection point section with the path and without the intersection point section.Wipe fully by live width without the intersection point section; There is the processing of intersection point section to change Step3.2 over to.
Step3.2, branch's situation and the local configuration thereof of the center line both sides of detection intersection point place straight line.
Step3.3, if both sides branch, then the trend with the both sides profile calculates the approximate border of wiping; If one-sided branch then is erased to a branchiess side center line of straight line, opposite side calculates with the trend of profile.
Improvement to this algorithm is larger, directly utilizes the variation of memory bitmap data message, and the pixel value+1(on the straight-line segment of having identified is established 0 for white, and 1 for black).After straight-line detection was finished, the work of wiping of straight line is assignment in internal memory just like this; Simultaneously because what plant that the Bresenham algorithm with live width in the subsegment growth course seeks only is non-white pixel, the intersection point of line segment can not bring impact to seeking straight line.And utilize pixel value on the last memory bitmap, can find out fast all intersection points (pixel value is greater than 2), for follow-up data arrangement processing etc. the situation of line segment (line width variation overlap) provides foundation.
Utilize circle and circular-arc information in the least square method approximate algorithm extraction straight-line segment; Utilize rule matching algorithm to extract the semantic informations such as arrow in straight-line segment, the arc section, marking line, boost line, blanking line, outer molded line; Utilize the OCR technology to carry out literal identification to wiping the image of identifying pel, utilize rule matching algorithm that literal is related with primitive information; Utilize the XML form with all the pel of vector quantization and semantic information thereof export file to and preserve, as the basic data of next step three-dimensional reconstruction.
Below the preferred embodiment of the invention is specified, but the invention is not limited to embodiment, those of ordinary skill in the art also can make all modification that is equal to or replacement under the prerequisite of the invention spirit, the modification that these are equal to or replacement all are included in the application's claim limited range.
Claims (19)
1. engineering drawing vectorization and pel semantic extracting method is characterized in that, comprise the steps:
S1 utilizes the information data in the script extraction PDF drawing file or generates corresponding BMP format-pattern data, and the file of picture format is carried out further pre-service;
S2 utilizes kind of a subsegment algorithm that the straight line pel in the view data is carried out vector quantization;
S3 carries out vector quantization to the circle/circular arc in the view data;
S4, the pel relation recognition is utilized based on callout box, Word message on the Algorithm for Text/Photo Separation removal line drawing of connected region;
S5, Output rusults.
2. engineering drawing vectorization according to claim 1 and pel semantic extracting method is characterized in that, described kind of subsegment algorithm comprises the steps:
S2.1, search kind of subsegment;
S2.2 plants the subsegment growth;
S2.3 has identified wiping of straight line.
3. engineering drawing vectorization according to claim 2 and pel semantic extracting method is characterized in that, described search kind subsegment comprises the steps:
S2.1.1 makes a square window centered by black pixel C, from the coboundary, the continuously black pixel fragment of getting in the direction of the clock on the window edge is candidate segment;
S2.1.2 is the candidate segment of L with untreated, length, and P seeks kind of a subsegment as initial point with its mid point; If without untreated candidate segment, such as R＜scope, R=R+r then, and skip to S2.1.1; If R 〉=scope finishes to seek, return failure;
S2.1.3 from P, presses the direction that CP determines, surveys the black pixel fragment length on the orthogonal directions take step-size in search r as step-length, and such as black pixel fragment length≤L, be the rule section, otherwise be irregular section, and take the current rule section mid point that records as Q; Such as n * r 〉=K * L, enter S2.1.4; Otherwise as m * r＞scope or | CP|＞scope, finish to seek, return S2.1.2; Other situations replace the P point with the Q point, continue S2.1.3;
S2.1.4 checks that the black pixel between P, the Q is communicated with the district;
S2.1.5 finishes to seek.
4. engineering drawing vectorization according to claim 3 and pel semantic extracting method is characterized in that, among the described S2.1.2, if without untreated candidate segment, and such as search radius R＜maximum search scope scope, R=R+r then, and skip to S2.1.1.
5. engineering drawing vectorization according to claim 3 and pel semantic extracting method is characterized in that, among the described S2.1.2, if without untreated candidate segment, such as search radius R 〉=maximum search scope scope, then finish to seek.
6. engineering drawing vectorization according to claim 3 and pel semantic extracting method is characterized in that, among the described S2.1.3, and the regular hop count n that records continuously; The irregular hop count m that records continuously, establishing the current rule section mid point that records is Q, if n * r 〉=K * L then enters S2.1.4.
7. engineering drawing vectorization according to claim 3 and pel semantic extracting method is characterized in that, among the described S2.1.3, and the regular hop count n that records continuously; The irregular hop count m that records continuously, establishing a current rule section mid point that records is Q, if m * r＞scope or | CP|＞scope, then finish to seek and return S2.1.2.
8. engineering drawing vectorization according to claim 3 and pel semantic extracting method is characterized in that, among the described S2.1.4, if the black pixel between P, the Q is communicated with district's qualified (satisfying kind of a subsegment definition standard), then finish to seek, and return successfully sign.
9. engineering drawing vectorization according to claim 3 and pel semantic extracting method is characterized in that, and be defective if the black pixel between P, the Q is communicated with the district among the described S2.1.4, then returns S2.1.3.
10. engineering drawing vectorization according to claim 2 and pel semantic extracting method is characterized in that, described kind of subsegment growth comprises the steps:
S2.2.1 generates the axial orthogonal directions of seed segment length path with the Bresenham algorithm;
S2.2.2 generates the track path that begins from the central point of kind of subsegment by the long axis direction of kind of subsegment with the pointwise of Bresenham algorithm, point of every generation checks whether the pixel of this point correspondence is stain.
11. engineering drawing vectorization according to claim 10 and pel semantic extracting method, it is characterized in that, in described S2.2.2, as check that pixel corresponding to this point is stain, then add up the length of continuous stain, if length greater than noise threshold, then will the length of continuous white point be charged to path segment table before detecting stain. check that the quadrature at current path point place is communicated with length.
12. engineering drawing vectorization according to claim 10 and pel semantic extracting method, it is characterized in that, in described S2.2.2, as check that pixel corresponding to this point is not stain, then add up the length of continuous white point, if length greater than noise threshold, then will the length of continuous stain be charged to path segment table before detecting white point.
13. engineering drawing vectorization according to claim 12 and pel semantic extracting method is characterized in that, in described S2.2.2, if the length of continuous stain then finishes the tracking on the current direction greater than the threshold value of dotted line spacing before detecting white point.
14. engineering drawing vectorization according to claim 2 and pel semantic extracting method is characterized in that, have describedly identified wiping of straight line and comprise the steps:
S2.3.1 along the connection length on the path testing path orthogonal directions of straight line, has been divided into the intersection point section with the path and without the intersection point section, has wiped fully by live width without the intersection point section;
S2.3.2, branch's situation and the local configuration thereof of the center line both sides of detection intersection point place straight line.
15. engineering drawing vectorization according to claim 14 and pel semantic extracting method is characterized in that, in described S2.3.2, branch into as described both sides branch, then the trend with the both sides profile calculates the approximate border of wiping.
16. engineering drawing vectorization according to claim 14 and pel semantic extracting method, it is characterized in that, in described S2.3.2, branch into as described one-sided branch, then a branchiess side is erased to the center line of straight line, opposite side calculates with the trend of profile.
17. engineering drawing vectorization according to claim 1 and pel semantic extracting method is characterized in that, described S3 utilizes circle and the circular-arc information in the least square method approximate algorithm extraction straight-line segment.
18. engineering drawing vectorization according to claim 1 and pel semantic extracting method is characterized in that, described S4 utilizes rule matching algorithm to extract the semantic informations such as arrow in straight-line segment, the arc section, marking line, boost line, blanking line, outer molded line; Utilize the OCR technology to carry out literal identification to wiping the image of identifying pel, utilize rule matching algorithm that literal is related with primitive information.
19. engineering drawing vectorization according to claim 1 and pel semantic extracting method is characterized in that, described S5 utilizes the XML form, and pel and the semantic information thereof of vector quantization export the file preservation to all.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012102773683A CN102880868A (en) | 2012-08-06 | 2012-08-06 | Engineering drawing vector conversion and primitive semantic extraction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012102773683A CN102880868A (en) | 2012-08-06 | 2012-08-06 | Engineering drawing vector conversion and primitive semantic extraction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102880868A true CN102880868A (en) | 2013-01-16 |
Family
ID=47482185
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012102773683A Pending CN102880868A (en) | 2012-08-06 | 2012-08-06 | Engineering drawing vector conversion and primitive semantic extraction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102880868A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103246751A (en) * | 2013-05-28 | 2013-08-14 | 国家电网公司 | Secondary drawing information identification and reading system |
| CN104252715A (en) * | 2014-09-05 | 2014-12-31 | 北京大学 | Single line image-based three-dimensional reconstruction method |
| CN105550363A (en) * | 2016-01-06 | 2016-05-04 | 北京联合大学 | Image semi-structural representing method based on XGMl |
| CN106339518A (en) * | 2015-07-10 | 2017-01-18 | 陈明德 | Computation system and method for engineering cost |
| CN108304848A (en) * | 2018-01-10 | 2018-07-20 | 链家网（北京）科技有限公司 | Extraction method, system, electronic equipment and the storage medium of house type feature |
| CN110458857A (en) * | 2019-08-12 | 2019-11-15 | 厦门美图之家科技有限公司 | Central symmetry pel detection method, device, electronic equipment and readable storage medium storing program for executing |
| CN111027429A (en) * | 2019-11-29 | 2020-04-17 | 陈韬文 | Data preprocessing method and system for intelligent identification of electrical drawings |
| CN112329411A (en) * | 2020-11-10 | 2021-02-05 | 湖北福泰建筑装饰工程有限公司 | Engineering budget accounting management and control management software |
| CN112632844A (en) * | 2019-09-24 | 2021-04-09 | 国际商业机器公司 | Extracting and analyzing information from engineering drawings |
| CN116740750A (en) * | 2023-06-30 | 2023-09-12 | 江苏方天电力技术有限公司 | Reconstruction method and device for identifying size information based on engineering drawing |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5761328A (en) * | 1995-05-22 | 1998-06-02 | Solberg Creations, Inc. | Computer automated system and method for converting source-documents bearing alphanumeric text relating to survey measurements |
| CN1333511A (en) * | 2000-07-14 | 2002-01-30 | 纬衡知识产权有限公司 | Raster image vector quantized computer automatic processing method |
-
2012
- 2012-08-06 CN CN2012102773683A patent/CN102880868A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5761328A (en) * | 1995-05-22 | 1998-06-02 | Solberg Creations, Inc. | Computer automated system and method for converting source-documents bearing alphanumeric text relating to survey measurements |
| CN1333511A (en) * | 2000-07-14 | 2002-01-30 | 纬衡知识产权有限公司 | Raster image vector quantized computer automatic processing method |
Non-Patent Citations (2)
| Title |
|---|
| 刘云鹏: "工程图纸自动识别技术的研究", 《万方数据库》, 31 December 2005 (2005-12-31) * |
| 宋继强等: "基于种子段的方向无关的直线矢量化方法", 《软件学报》, vol. 11, no. 9, 30 September 2000 (2000-09-30), pages 58 - 63 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103246751A (en) * | 2013-05-28 | 2013-08-14 | 国家电网公司 | Secondary drawing information identification and reading system |
| CN103246751B (en) * | 2013-05-28 | 2016-04-20 | 国家电网公司 | The identification of secondary drawing information and reading system |
| CN104252715A (en) * | 2014-09-05 | 2014-12-31 | 北京大学 | Single line image-based three-dimensional reconstruction method |
| CN104252715B (en) * | 2014-09-05 | 2017-05-03 | 北京大学 | Single line image-based three-dimensional reconstruction method |
| CN106339518A (en) * | 2015-07-10 | 2017-01-18 | 陈明德 | Computation system and method for engineering cost |
| CN105550363A (en) * | 2016-01-06 | 2016-05-04 | 北京联合大学 | Image semi-structural representing method based on XGMl |
| CN105550363B (en) * | 2016-01-06 | 2019-03-22 | 北京联合大学 | The semi-structured representation method of image based on XGML |
| CN108304848A (en) * | 2018-01-10 | 2018-07-20 | 链家网（北京）科技有限公司 | Extraction method, system, electronic equipment and the storage medium of house type feature |
| CN110458857A (en) * | 2019-08-12 | 2019-11-15 | 厦门美图之家科技有限公司 | Central symmetry pel detection method, device, electronic equipment and readable storage medium storing program for executing |
| CN112632844A (en) * | 2019-09-24 | 2021-04-09 | 国际商业机器公司 | Extracting and analyzing information from engineering drawings |
| US11270105B2 (en) | 2019-09-24 | 2022-03-08 | International Business Machines Corporation | Extracting and analyzing information from engineering drawings |
| CN112632844B (en) * | 2019-09-24 | 2024-04-12 | 国际商业机器公司 | Method and system for extracting and analyzing information from engineering drawing |
| CN111027429A (en) * | 2019-11-29 | 2020-04-17 | 陈韬文 | Data preprocessing method and system for intelligent identification of electrical drawings |
| CN111027429B (en) * | 2019-11-29 | 2024-01-12 | 广东工业大学 | Data preprocessing method and system for intelligent recognition of electrical drawing |
| CN112329411A (en) * | 2020-11-10 | 2021-02-05 | 湖北福泰建筑装饰工程有限公司 | Engineering budget accounting management and control management software |
| CN116740750A (en) * | 2023-06-30 | 2023-09-12 | 江苏方天电力技术有限公司 | Reconstruction method and device for identifying size information based on engineering drawing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102880868A (en) | Engineering drawing vector conversion and primitive semantic extraction method | |
| Sun et al. | Disp r-cnn: Stereo 3d object detection via shape prior guided instance disparity estimation | |
| EP3506162B1 (en) | Method and apparatus for determining matching relationship between point cloud data | |
| CN109948510B (en) | Document image instance segmentation method and device | |
| EP3955213A1 (en) | Three-dimensional object modeling method, image processing method, and image processing device | |
| Wu et al. | Rapid localization and extraction of street light poles in mobile LiDAR point clouds: A supervoxel-based approach | |
| Khotanzad et al. | Contour line and geographic feature extraction from USGS color topographical paper maps | |
| CN105528614B (en) | A kind of recognition methods of the cartoon image space of a whole page and automatic recognition system | |
| Yang et al. | Detecting rotated objects as gaussian distributions and its 3-d generalization | |
| CN100464347C (en) | Vector graphics identifying method for engineering CAD drawing | |
| Biasutti et al. | Lu-net: An efficient network for 3d lidar point cloud semantic segmentation based on end-to-end-learned 3d features and u-net | |
| CN110838105B (en) | Business process model image recognition and reconstruction method | |
| CN113628291B (en) | Multi-shape target grid data vectorization method based on boundary extraction and combination | |
| CN110263794B (en) | Training method of target recognition model based on data enhancement | |
| CN112990183B (en) | Method, system and device for extracting homonymous strokes of offline handwritten Chinese characters | |
| CN111738055A (en) | Multi-class text detection system and bill form detection method based on same | |
| CN102750531A (en) | Method for detecting handwriting mark symbols for bill document positioning grids | |
| KR20200075940A (en) | Real-time data set enlarging system, method of enlarging data set in real-time, and computer-readable medium having a program recorded therein for executing the same | |
| CN114241469A (en) | Information identification method and device for electricity meter rotation process | |
| CN113538585B (en) | High-precision multi-target intelligent identification, positioning and tracking method and system based on unmanned aerial vehicle | |
| US20020006224A1 (en) | Computer automated process for vectorization of raster images | |
| Bao et al. | Vectorizing line drawings with near-constant line width | |
| CN112364863A (en) | Character positioning method and system for license document | |
| CN116541912A (en) | Method and device for generating wiring diagram based on CAD image recognition | |
| Martens et al. | Cross domain matching for semantic point cloud segmentation based on image segmentation and geometric reasoning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C05 | Deemed withdrawal (patent law before 1993) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130116 |