CN111368757A

CN111368757A - Machine learning-oriented column large sample building drawing layer classification method and system

Info

Publication number: CN111368757A
Application number: CN202010156016.7A
Authority: CN
Inventors: 刘仕杰
Original assignee: Glodon Co Ltd
Current assignee: Glodon Co Ltd
Priority date: 2020-03-09
Filing date: 2020-03-09
Publication date: 2020-07-03

Abstract

The invention discloses a machine learning-oriented column proof building drawing layer classification method, which is used for a server and comprises the following steps: and importing the CAD construction drawing, preprocessing, dividing the CAD construction drawing into sub-drawings according to the graphic elements, establishing a corresponding attribute information mapping relation, analyzing the column large-sample characteristic layer, identifying and classifying, and finishing the classification of the column large-sample CAD construction drawing characteristic layer according to the table layer. The method carries out business modeling on the building component expressed by the CAD building drawing, analyzes and refines the business characteristics expressed by the design drawing, and further provides the concept of the feature layer of the building component to guide the automatic classification of the layer; an automatic layer classification algorithm from the business features expressed by the design drawing to the building component feature layers is provided, so that fine-grained feature layer automatic classification is realized, and a solid foundation is laid for simplification of a subsequent identification algorithm and introduction of a machine learning intelligent algorithm.

Description

Machine learning-oriented column large sample building drawing layer classification method and system

Technical Field

The invention belongs to the technical field of machine learning intelligent identification of building information models and building drawings, in particular to a layer classification method based on component characteristics, and particularly relates to a machine learning-oriented layer classification method and system for large-scale building drawings.

Background

With the development of computer software and hardware technology, the building information model BIM technology is applied more and more deeply in the building industry, and the BIM model is largely used to complete work such as calculation amount, volume, pricing and the like in each stage of budget, construction, settlement, operation and maintenance and the like of the building engineering, so that the informatization level of the building industry is greatly improved. It can be seen that the construction of the BIM model becomes particularly important. At present, the building construction drawing is delivered in a CAD drawing form in the building design industry, so the BIM model is usually realized in a mode of manual die turnover and automatic software die turnover.

The prior art such as the automatic modeling series parent-child patents of Guangzhou university, namely, the building automatic modeling method based on building drawings, the balcony automatic identification method based on building drawings, the door and window automatic identification method based on building drawings, the well automatic identification method and system based on building drawings, the column and wall automatic identification method based on building drawings, the stair automatic identification method based on building drawings, and the layer classification patent of Ningbo Ruifeng information technology Limited company, namely, the layer classification method for converting building drawings into three-dimensional BIM models, all provide solutions based on CAD drawings automatic mold turning. However, the existing CAD drawing automatic rollover technology faces great technical development pressure and challenges, and is embodied in that: the method has the advantages that the number of domestic building design units is large, no unified design standard exists, the habits of designers are different, delivered building CAD drawings are different finally, the identification effect is very poor, and a plurality of scenes which cannot be processed always appear; sometimes, scenes conflict with each other, and the automatic CAD rollover function is very difficult to implement.

The existing engineering drawing recognition method is based on characteristic engineering and rules for recognition, logic reasoning and deduction are carried out according to the found characteristics and rules, then a classic logic method is used for deduction and matching, and a best matching mode is selected through different strategies, which is also called as mode recognition, however, the recognition method has the following defects:

1. for simple graphics and image features, human comprehension is easy (such as parallel and intersecting), and when a business scene tends to be complex, multi-angle and multi-dimension combination among the features or more advanced features, human beings tend to choose to ignore such unintelligible features, which is not beneficial to abstract and identify at a higher level.

2. The extraction of the traditional feature engineering is usually based on the current professional features, and if another professional is made, on one hand, the user needs to learn from beginning to end, so that the learning cost is increased; in addition, the common characteristics between the professions are difficult to refine and fuse, so that the repetitive workload is large, and the universality is weak.

3. The recognition algorithm based on the rules is a set of completely self-consistent logical operation and symbolic operation, has strict logical reasoning, however, in reality, drawings are different, and after a set of rules are summarized, new scene discovery is not applicable any more, so that new rules need to be continuously searched to make up for the previous defects, on one hand, the software generalization capability is weak, and in addition, the maintenance cost of the software is increased day by day.

4. The feature engineering and rule searching have no uniform specification requirements and are seriously dependent on the literacy of the practitioner, which is difficult to ensure the quality of software and the development progress from the engineering perspective.

5. At present, an identification technology utilizes a feature engineering and a traditional machine learning method to train and predict, although the accuracy is improved to a certain degree, the accuracy is not completely separated from an artificial feature extraction range, and the quality of features is seriously depended, so that the generalization capability of the identification technology is not very strong.

Disclosure of Invention

Aiming at the defects of the background art, the invention aims to provide a column proof CAD (computer-aided design) construction drawing layer classification method facing to a machine learning intelligent identification algorithm, which can automatically classify column proof CAD construction drawings with finer granularity, so that the identification accuracy of the column proof CAD construction drawings is improved by applying the machine learning algorithm conveniently, and the efficiency of reconstructing a three-dimensional BIM (building information model) is improved.

In order to achieve the above object, the present invention provides a method for classifying column proof construction drawing layers for machine learning, which is used at a server side and comprises:

collecting various building digital drawing data, carrying out business function view region segmentation on the drawing, carrying out image preprocessing on the segmented drawing, carrying out enhancement processing on the image data, constructing a deep machine learning image semantic recognition network model, and carrying out new model training by adopting a transfer learning technology; and predicting the image by using the trained model, mapping and converting the predicted data to a geometric model coordinate system, and reconstructing the three-dimensional model of the recognized model type to finish drawing recognition.

For a server side, comprising: and importing the CAD construction drawing, preprocessing, dividing the CAD construction drawing into sub-drawings according to the graphic elements, establishing a corresponding attribute information mapping relation, analyzing the column large-sample characteristic layer, identifying and classifying, and finishing the classification of the column large-sample CAD construction drawing characteristic layer according to the table layer.

Preferably, the CAD construction drawing includes subgraphs of columns, walls, beams, plates, reinforcing bars, stairs and other various building components of different floors to form a subgraph set.

Preferably, the sub-figures have rectangular borders, and the sub-figures are geometrically separated from each other.

Preferably, the step of segmenting into subgraphs according to the graphic elements comprises the following specific steps:

step 101, traversing all graphic elements of the CAD architectural drawing, and finding out all rectangular elements;

102, finding out a maximum and separated rectangular graphic element set, wherein the rectangular elements are used as CAD sub-graph frames;

step 103, traversing the sub-picture frame, selecting CAD graphic elements in the range according to the frame range, storing the CAD graphic elements as a new CAD drawing file, and storing the original drawing layer classification information of the CAD drawing;

and step 104, repeating the steps S101-S103, and traversing to finish the automatic segmentation of the CAD drawing.

Preferably, the largest set of rectangular graphic elements refers to rectangular elements that are not included in other rectangles.

Preferably, the establishing of the mapping relationship of the corresponding attribute information to identify the usage of the subgraph includes the following steps:

s201, identifying all picture name information, identifying a drawing catalog table from a CAD design drawing, and obtaining the number, the picture number, the drawing name and other data of subgraphs in the CAD drawing;

s202, identifying the name information of the subgraph graph, traversing all the text graph element information in the subgraph, matching with the name of the drawing in the drawing catalog, and if matching, establishing the mapping relation between the subgraph graph and the name of the drawing;

s203, identifying the use of the sub-drawing, identifying the name information of the building component in the drawing name, and establishing the matching information between the drawing name and the building component.

Preferably, the column big sample member is divided into an in-situ column big sample and a table column big sample, and the table column big sample is divided into a first type and a second type and comprises the following expression information characteristics: name, elevation, section polygon, section dimension information, longitudinal bar arrangement position information, stirrup arrangement position information, longitudinal bar information, and stirrup information.

Preferably, the column macromolecular member comprises the following design element characteristics: the method comprises the following steps of table, in-situ big sample centralized labeling, column component name numbering, elevation, longitudinal bar centralized labeling, stirrup centralized labeling, section polygon, section size labeling, longitudinal bar arrangement point position set, longitudinal bar in-situ labeling, stirrup arrangement line set, stirrup splitting line set and stirrup in-situ labeling.

Preferably, the classification of the characteristic layers of the column master sample CAD construction drawing comprises:

s301, automatically classifying the table feature layers, and automatically identifying the table and the table header;

s302, automatically classifying the characteristic layers marked in the in-situ big sample set;

s303, automatically classifying the large sample characteristic layers of the two-type table columns;

s304, automatically classifying the large sample identification feature layers of the type I table columns;

s305, automatically classifying the size labeling feature map layers;

s306, automatically classifying the section polygon feature layers;

s307, automatically classifying feature map layers of the longitudinal bar arrangement point location set;

s308, automatically classifying the longitudinal bar in-situ labeling feature map layers;

s309, automatically classifying the feature map layers of the stirrup rib arrangement line set;

s310, automatically classifying the characteristic map layers of the stirrup splitting map set;

s311, automatically classifying the stirrup in-situ labeling feature map layer.

Preferably, the step S301 includes:

1.1, automatically identifying the form;

the geometric characteristics of the table comprise two groups of horizontal and vertical parallel lines, and the table has the following characteristics:

1) assuming a horizontal parallel line, taking the left end as a starting point and the right section as an end point; the parallel lines in the vertical direction have the upper end as the starting point and the lower end as the end point;

then the following reasoning is:

2) the lengths of the parallel lines in all horizontal directions are equal, and the lengths of the parallel lines in all vertical directions are equal;

3) all the starting points of the parallel lines in the horizontal direction are positioned on the same straight line, namely the leftmost parallel straight line of the horizontal line in the vertical direction; the terminal points of all parallel lines in the horizontal direction are positioned on the same straight line, namely the rightmost parallel straight line in the horizontal line in the vertical direction;

4) the starting points of all parallel lines in the vertical direction are all positioned on the same straight line, namely the uppermost parallel straight line of the horizontal line in the horizontal direction; the end points of all parallel lines in the vertical direction are positioned on the same straight line, namely the lowest parallel straight line of the horizontal line in the horizontal direction;

performing table identification according to the table features, identifying the number of rows and columns, and calculating each table unit according to the number of rows and columns, assuming Cell, the rectangular frame range of the table unit, assuming Rect, namely Rect (left, top, right, bottom);

if the automatic identification of the table fails, jumping to the step S302 of automatically classifying the marked feature layers in the in-situ big sample set;

1.2, identifying a header;

preferentially identifying the header according to the columns, if the header cannot be identified according to the columns, identifying the header according to a row mode, and the algorithm logic is the same;

1.2.1 traversing according to columns, and if traversing is finished, turning to the step 1.2.4;

1.2.2 traversing each cell according to rows in the current column, retrieving CAD graphic element information based on the rect range of the cell, if the current cell only retrieves one CAD text graphic element, recording the cell as a candidate header field, and simultaneously recording the graphic element information of all CAD graphic elements retrieved by the current cell for later use;

1.2.3 if all cells in the current list are candidate headers and header information in the current list meets the header information standard of the column full-page table, recording that the current list is a header list; adjusting to step 1.2.1;

1.2.3.1 if all cells in the current column are candidate headers and header information in the current column does not accord with the header information standard of the column full-page table, recording that the current column is a non-header column; adjusting to step 1.2.1;

1.2.3.2 if one cell in the current list is a non-candidate header, recording that the current list is a non-header list, and turning to the step 1.2.1;

1.2.4 checking all columns, and judging whether the head columns are identified or not and the number of the head columns; if the header is not identified, identifying the header by line;

1.2.5 if the header columns are identified, adding the table lines into the table line characteristic layer, and outputting the table information table for later use to finish automatic classification of the table lines;

if the table is successfully identified but the table header is failed to be identified, jumping to the step S303 to start the automatic classification of the large-sample characteristic layer of the two-type table column;

if the table is successfully identified and the table header is successfully identified, jumping to step S304 to start automatic classification of the column large sample identification feature layer of the table.

Preferably, in step S302, the name, longitudinal bars, stirrups, dimension information, and other related information of the cylindrical macro sample are marked in the original CAD drawing set of the cylindrical macro sample, and include:

2.1 traversing all text information of the CAD drawing, and skipping for 2.5 if the traversing is finished;

2.2, if the current CAD text graphic element information is a multi-line text, skipping 2.3; otherwise, skipping 2.1;

2.3, if the current CAD multi-line text graphic primitive meets the marking information standards of column large sample name, size, longitudinal bar and stirrup, skipping 2.4;

2.3.1, if the current CAD multi-line text graphic primitive does not accord with the standard of the column big sample name, the size, the longitudinal bar and the stirrup marking information, skipping 2.1;

2.4 retrieving lead line primitives of the current CAD multi-line text primitives based on the current CAD multi-line text primitives, and if the lead line primitives are found, adding the CAD multi-line text primitives and the lead line primitives into the in-situ large sample set to mark the characteristic layer;

and 2.5, finishing automatic classification of the characteristic layers marked in the large in-situ column sample set.

Preferably, in the two-type table column sample in step S303, each cell completely contains the name, longitudinal bar, hoop bar, elevation information, and section information of the column sample, and the automatic classification of the layer includes:

3.1 traversing the table by rows, and skipping 3.4 if the traversal is finished;

3.2 traversing each cell of the current line, and if the cell is completed, skipping by 3.1;

3.3, marking the current cell as a section feature, traversing the CAD text graphic element of the current cell, and if the current cell is finished, skipping by 3.2;

3.3.1 if the character string value of the current text primitive conforms to the main component number/name semantic specification, marking the current line as the column component name number characteristic, adding CAD graphic elements of all column cells of the current line into the column component name number characteristic layer, and skipping 3.3;

3.3.2 if the character string value of the current text primitive conforms to the longitudinal bar semantic specification and the text primitive is connected with the filling section polygon through a lead, adding the current CAD graphic element into the feature layer of the column member longitudinal bar centralized labeling, and skipping by 3.3;

3.3.3, if the character string value of the current text primitive meets the stirrup semantic specification and simultaneously meets one of the following conditions, adding the current CAD graphic element into the column member stirrup set labeling feature layer, and skipping 3.3;

3.3.3.1 the text primitive is connected to the fill-section polygon by a lead;

3.3.3.2 the text graphic primitive is located at the middle and lower part of Rect of the current cell and an independent horizontal short straight line is arranged above the text graphic primitive;

3.3.4 if the character string value of the current text primitive contains related semantic information such as elevation, adding the current CAD graphic element into the column member elevation feature layer, and skipping 3.3;

and 3.4, completing automatic classification of the feature layers of the concentrated labeling type of the large samples of the two-type form columns.

Preferably, the column big sample in the first form in step S304 has a header cell of the column big sample name, the longitudinal bar, the hoop bar, the elevation information, and the section information field, and the value of each field is also located in different independent table cells.

Assuming that the list head is called list head (the row type processing logic is the same), the automatic classification of the layer includes:

4.1 traversing the table header, if the values of the header cells of the current row conform to the semantic specification, marking the current row as column member name numbering characteristics, adding CAD graphic elements of all column cells of the current row into a column member name numbering characteristic layer, and finishing the automatic classification of the layer;

4.2 traversing the table header, if the values of the table header cells of the current row meet the semantic specification, marking the current row as the longitudinal bar centralized labeling feature, adding the CAD graphic elements of all column cells of the current row into the column member longitudinal bar centralized labeling feature layer, and finishing the automatic classification of the layer;

4.3 traversing the table header, if the values of the header cells of the current row meet the semantic specification, marking the current row as the column member stirrup centralized labeling feature, adding the CAD graphic elements of all the column cells of the current row into the column member stirrup centralized labeling feature layer, and completing automatic classification of the layer;

4.4 traversing the table header, if the values of the table header cells of the current row conform to the semantic specification, marking the current row as column member elevation features, adding CAD graphic elements of all column cells of the current row into a column member elevation feature layer, and finishing automatic classification of the layer;

4.5, traversing the table header, if the values of the table header cells of the current row conform to the semantic specification, marking the current row as column member section polygon characteristics, adding CAD graphic elements of all column cells of the current row into a column member section polygon characteristic layer, and finishing the automatic classification of the layer.

Preferably, in the step S305, the size labeling feature recognition is performed, and the table column full-page proof and the in-situ column full-page proof have different automatic recognition and classification schemes.

5.1 automatic classification scheme of the dimension marking characteristic layer of the column large sample of the table;

5.1.1 traversing a cross section cell in the table, and skipping 5.1.3 if the traversal is finished;

5.1.2 traversing CAD graphic primitives in the cross-section cell to perform secondary grouping, wherein the first stage groups the CAD graphic primitives according to colors, namely, the graphic primitives with the same color are grouped into one group; based on the first-stage color grouping, performing second-stage grouping according to connectivity, namely dividing mutually communicated CAD graphic primitives into a group;

5.1.3 traversing the secondary line segment group, calling a 5.3 size marking feature recognition algorithm, recording the rect range of the size marking if the size marking primitive is recognized, and adding the rect range into the size marking feature layer. If not, skipping 5.1.1;

5.1.3, completing automatic classification of the large sample size marking characteristic map layer of the table column;

5.2 automatic classification scheme of the feature layer for in-situ column large sample size marking;

5.2.1 traversing the table by rows, and skipping 5.2.4 if the traversal is finished;

5.2.2 traversing each cell of the current row, and if the cell is completed, skipping by 5.2.1;

5.2.3 traversing CAD graphic primitives in the cross-section cell to perform secondary grouping, and if the secondary grouping is finished, skipping by 5.2.2;

5.2.3.1 the first stage groups CAD primitives by color, i.e., the same color primitives are grouped;

5.2.3.2 based on the first-stage color grouping, performing second-stage grouping according to connectivity, namely dividing mutually-communicated CAD graphic primitives into a group;

5.2.3.4 traversing the secondary line segment group, calling a 5.3 size marking feature recognition algorithm, if a size marking primitive is recognized, recording the rect range of the size marking, and adding the rect range into the size marking feature layer; skipping 5.2.2;

5.2.4, completing automatic classification of the in-situ column large sample size labeling feature layer;

5.3 size marking recognition algorithm;

5.3.1 analyzing a connected line segment set, if a group of parallel lines exist, a collinear connected straight line is intersected with the parallel lines, and a group of parallel segment straight lines exist at the intersection point, identifying a size marking characteristic, and recording the rect of the size marking characteristic; skipping 5.3.2;

5.3.2 traversing CAD graphic elements in the section cell and retrieving all digital text graphic elements. And searching the CAD digital text graphic element by using the rect range of the dimension marking characteristic, and identifying the CAD digital text graphic element as the dimension marking text characteristic if the CAD digital text graphic element is searched.

Preferably, the cross-sectional polygonal features, the table column large samples and the in-situ column large samples in step S306 have different automatic identifications and classifications, which specifically include:

6.1 automatically classifying the polygonal feature layers of the large sample section of the table column;

6.1.1 traversing the cross section cell in the table, and skipping 6.1.4 if the traversal is finished;

6.1.2 traversing the CAD graphic element in the current cell, and skipping 6.1.1 if the traversing is finished;

6.1.3 if the current CAD graphic primitive is a polygon, and each edge of the polygon is collinear with one of the parallel lines marked by the size, the polygon is a section polygon, and the section polygon is added into a section polygon characteristic layer; otherwise, skipping 6.1.2;

6.1.4, completing automatic classification of the section polygon feature layer;

6.2 automatically classifying the polygonal feature layers of the in-situ column large sample cross sections;

6.2.1 traversing the in-situ big sample set marking features, and skipping 6.2.4 if traversing is completed;

6.2.2 searching the lead wires connected with the current in-situ big sample set by labeling according to the geometric relation;

6.2.3 traversing the CAD polygon graphic primitive, calculating a polygon connected with the marking lead in the current in-situ big sample set, and if the polygon is found, adding the polygon into the section polygon feature layer; if not, skipping 6.2.1;

6.2.3.1 one end of the lead is crossed with the polygon, and the crossing point is located at the end point of the lead, then the lead is judged to be connected with the polygon.

6.2.4, completing automatic classification of the polygonal feature layer of the section of the in-situ column large sample.

Preferably, the step S307 includes:

7.1 traverse the cross-sectional polygon of the column spline. If the traversal is completed, jumping to 7.5;

7.2 screening a CAD drawing element set positioned in the current section polygon based on the current section polygon range;

7.3 traverse the set of CAD primitives as inside the current cross-sectional polygon. If the traversal is completed, jumping to 7.1;

and 7.4, if the current CAD graphic element is a circle, adding the current CAD graphic element into the feature map layer of the longitudinal bar distribution point set. Otherwise, jump 7.3.

7.5, completing automatic classification of the feature map layers of the longitudinal bar distribution point location set.

Preferably, the in-situ labeling of the longitudinal bar in step S308 is labeled on the longitudinal bar, and appears only in the CAD drawing of the column prototype of the form, and includes:

8.1, traversing a section cell in the table, and skipping 8.5 if the traversal is finished;

8.2 traversing CAD text graphic elements in the cross-section cell; if the traversal is completed, skipping 8.1;

and 8.3, if the steel bar information of the current CAD text graphic primitive meets the specification standard of the longitudinal bars, identifying the longitudinal bars as longitudinal bar labels, and skipping 8.4. Otherwise, skipping 8.2;

8.4 retrieving the corresponding longitudinal bar marking lead based on the current longitudinal bar marking, traversing the current longitudinal bar point set, and checking that the longitudinal bar marking lead is connected with the longitudinal bar circle; if the longitudinal bar in-situ labeling identification is successful, adding longitudinal bar labeling and longitudinal bar labeling lead wires into the longitudinal bar in-situ labeling characteristic image layer, and skipping by 8.2;

8.5, completing automatic classification of the longitudinal bar in-situ labeling feature map layers.

Preferably, the step S309 includes:

9.1 traverse the cross-sectional polygon of the column spline. If the traversal is completed, jumping to 9.5;

9.2 screening a CAD drawing element set positioned in the current section polygon based on the current section polygon range;

9.3 traverse the set of CAD primitives as inside the current cross-sectional polygon. If the traversal is completed, jumping to 9.1;

9.4 if the current CAD graphic primitive is a polyline, adding the current CAD graphic primitive into the stirrup reinforcement line set characteristic layer. Otherwise, jumping to 9.3;

and 9.5, completing automatic classification of the stirrup rib arrangement line set characteristic map layer.

Preferably, the stirrup splitting diagram in step S310 appears only in the tabular pillar prototype CAD drawing, and includes:

10.1, traversing a cross section cell in the table, and skipping by 10.6 if the traversal is finished;

10.2, traversing the secondary grouping of the CAD graphic elements in the cross-section cell, and skipping by 10.1 if the traversing is finished;

10.3 if the current secondary line segment grouping is the size marking characteristic, skipping 10.2;

10.4 if the current secondary segment group is a cross-sectional polygon or inside the cross-sectional polygon, skipping by 10.2;

and 10.5 if the current secondary line segment group is a multi-segment line outside the section polygon and an arc exists, identifying the current secondary line segment group as a stirrup splitting map set characteristic, identifying the secondary line segment group as the stirrup splitting map, and adding the stirrup splitting map set characteristic map layer into the stirrup splitting map set characteristic map layer. Otherwise, skipping 10.2;

and 10.6, completing automatic classification of the stirrup splitting graph line set characteristic graph layers.

Preferably, the step S311 of in-situ labeling of the stirrup is performed on the stirrup splitting diagram, and only appears in the CAD drawing of the column prototype of the form, and includes:

11.1, traversing a section cell in the table, and skipping 11.5 if the traversal is finished;

11.2 traversing CAD text graphic elements in the cross-section cell; if the traversal is completed, jumping to 11.1;

11.3, if the steel bar information of the current CAD text graphic element meets the stirrup specification standard, identifying the steel bar as a stirrup mark, and skipping 11.4; otherwise, jumping to 11.2;

11.4 retrieving corresponding stirrup marking leads based on the current stirrup marking, traversing the secondary line segment grouping of the current cell, and checking the connection of the marking leads and the stirrup splitting diagram; if the in-situ stirrup marking identification is successful, adding the stirrup marking and the stirrup marking lead into the in-situ stirrup marking characteristic map layer, and skipping 11.2;

11.5, completing automatic classification of the stirrup in-situ labeling feature map layer.

The utility model provides a towards big appearance building drawing picture layer classification system of post of machine learning, includes:

the drawing preprocessing unit is used for importing CAD (computer-aided design) construction drawings and preprocessing the CAD construction drawings;

the drawing dividing unit is used for dividing the drawing into subgraphs according to the graphic elements, establishing a mapping relation of corresponding attribute information, traversing all the graphic elements of the CAD construction drawing and finding out all the rectangular elements; finding out a maximum and separated rectangular graphic element set, wherein the rectangular elements are used as CAD sub-graph frames; traversing the sub-picture frame, selecting CAD graphic elements in the frame range according to the frame range, storing the CAD graphic elements as a new CAD drawing file, and storing the original drawing layer classification information of the CAD drawing; traversing to finish automatic segmentation of the CAD drawing;

identifying all the picture name information, identifying a drawing directory table from the CAD design drawing, and obtaining the number, the picture number, the drawing name and other data of the subgraphs in the CAD drawing; identifying the name information of the subgraph graph, traversing all the text primitive information in the subgraph, matching the text primitive information with the drawing name in the drawing catalog, and if the text primitive information is matched with the drawing name in the drawing catalog, establishing a mapping relation between the subgraph and the drawing name; identifying the use of the sub-drawing, identifying the name information of the building component in the drawing name, and establishing the matching information between the drawing name and the building component;

the column big sample characteristic layer classification unit is used for analyzing the column big sample characteristic layer and identifying and classifying the column big sample characteristic layer;

and the drawing characteristic layer classification unit is used for finishing the classification of the column rough CAD building drawing characteristic layers according to the form layers.

Preferably, the step of classifying the pillar large sample feature layer includes:

the automatic classification module of the table characteristic map layer is used for carrying out automatic identification and header automatic identification of the table; an automatic classification module for characteristic layers is marked in the in-situ big sample set; the automatic classification module of the large sample characteristic map layer of the type II table column; the first type table column large sample identification feature layer automatic classification module; the automatic classification module of the size labeling characteristic map layer; the section polygon feature layer automatic classification module; the automatic classification module for the feature map layer of the longitudinal bar distribution point location set; the automatic classification module for the longitudinal bar in-situ labeling feature map layer; a stirrup distributing wire set characteristic image layer automatic module; the automatic classification module for the characteristic map layer of the stirrup splitting map set; and the automatic classifying module for the stirrup in-situ labeling characteristic map layer.

Compared with the prior art, the invention has the following advantages:

1. the service characteristics of the column large-scale building components are divided more finely, the concept of the characteristic map layer is provided based on the characteristics, and the automatic classification process of the characteristic map layer is divided finely.

2. A detailed and feasible technical implementation scheme is provided for the characteristic map layer of the column large-sample drawing, and a better drawing data structuring basis is provided without performing a machine learning algorithm subsequently.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments of the present invention will be briefly described below, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

FIG. 1 is a schematic diagram of a process of classifying graphic layers of a column large sample construction drawing facing machine learning provided by the invention;

FIG. 2 is a schematic view of the automatic identification and classification process of the characteristic drawing layer of the column proof CAD drawing;

FIG. 3 is a schematic view of a large sample of the in situ column of the present invention;

FIG. 4 is a schematic view of a column of a table according to the first embodiment of the present invention;

FIG. 5 is a schematic diagram of a two-type table column according to the present invention.

Detailed Description

Features and exemplary embodiments of various aspects of the present invention will be described in detail below, and in order to make objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not to be construed as limiting the invention. It will be apparent to one skilled in the art that the present invention may be practiced without some of these specific details. The following description of the embodiments is merely intended to provide a better understanding of the present invention by illustrating examples of the present invention.

It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.

Currently, with the continuous development and maturity of new technologies such as cloud, big data, artificial intelligence, etc., it becomes feasible to use the technology of artificial intelligence machine learning to realize the intelligent rollover of CAD drawings, becoming a new development direction. In order to apply the artificial intelligence machine learning technology, the technology of carrying out structural processing on the unstructured drawing is the first step, namely the technology of extracting and structuring a single sample of a building component through careful layer classification.

As shown in fig. 1, the present embodiment provides a column proof construction drawing layer classification method for machine learning, which is used at a server side and includes: and importing the CAD construction drawing, preprocessing, dividing the CAD construction drawing into sub-drawings according to the graphic elements, establishing a corresponding attribute information mapping relation, analyzing the column large-sample characteristic layer, identifying and classifying, and finishing the classification of the column large-sample CAD construction drawing characteristic layer according to the table layer.

In some embodiments, the CAD construction drawing is imported and then BIM modeling is performed, and the CAD construction drawing is preprocessed.

CAD drawing outputted by the building design is provided in a form of a drawing set, namely, a dwg drawing file comprises sub-drawings of various building components such as columns, walls, beams, plates, reinforcing bars, stairs and the like of different floors. The CAD construction drawing comprises sub-drawings of columns, walls, beams, plates, plate bars, stairs and other various building components of different floors to form a sub-drawing set.

In some embodiments, the subgraphs have rectangular borders, and the subgraphs are geometrically separated from each other.

In some embodiments, the partitioning into subgraphs according to the graphical elements comprises the following specific steps:

In some embodiments, the largest set of rectangular graphic elements refers to rectangular elements that are not encompassed by other rectangles.

After the drawing is divided, the specific use of the divided subgraph needs to be further identified, namely the subgraph describes the design drawing of the building component, and the use of the subgraph is determined mainly by identifying the name of the subgraph.

In some embodiments, establishing the mapping relationship of the corresponding attribute information to identify the usage of the subgraph includes the following steps:

One sub-drawing corresponds to multiple purposes, for example, one sub-drawing simultaneously contains information of shear walls and column primitives, that is, the sub-drawing and the drawing purposes have a one-to-many relationship.

In an extreme case, under the condition that the drawing purpose is judged not well through the drawing name, the purpose of the sub-drawing is judged by the auxiliary drawing name through the identification of the primitive characteristic information.

In some embodiments, the pillar big sample member is divided into an in-situ pillar big sample (as shown in fig. 3) and a table pillar big sample, and the table pillar big sample is divided into a type (as shown in fig. 4) and a type (as shown in fig. 5), and includes the following expression information characteristics: name, elevation, section polygon, section dimension information, longitudinal bar arrangement position information, stirrup arrangement position information, longitudinal bar information, and stirrup information.

Through analysis of a large number of column big sample drawings, 7 characteristics expressed on the design drawings of the column big sample can be obtained, wherein some characteristics are necessary characteristics, and some characteristics are optional characteristics, and the details are shown in table 1.

From the CAD drawing design expression of the column-large building element, the information to be expressed in the drawing is shown in table 1.

TABLE 1

CAD drawing design elements corresponding to necessary attributes of large-scale building components of serial number columns

1 name table sample: form name field, name label

In-situ bulk sample preparation: centralized labeling

2 elevation table bulk sample: elevation field, elevation mark and floor mark of form

Table sample: form field

3 section polygon table large sample: table section field

In-situ bulk sample preparation: column frame

4 section size information table summary: dimension marking

In-situ bulk sample preparation: centralized labeling

5 circular point set inside polygon with longitudinal rib arrangement rib position information section

6 inside multistage line set of stirrup cloth muscle position information cross-section polygon

7 longitudinal bar reinforcing steel bar information table sample: longitudinal bar field, longitudinal bar centralized labeling and longitudinal bar in-situ labeling of table

In-situ bulk sample preparation: centralized labeling

8, a stirrup reinforcement information table is as follows: form stirrup field, stirrup centralized labeling, stirrup splitting diagram and stirrup in-situ labeling

In-situ bulk sample preparation: centralized labeling

The design elements of the cylindrical large sample CAD drawing can be summarized into 13 graphic features, as shown in Table 2:

TABLE 2

Serial number characteristic description schematic diagram

1 Table form general sample essential characteristic, in-situ general sample does not have the characteristic

2, the in-situ hand sample is marked with the necessary characteristics of the in-situ hand sample, and the name, longitudinal bar, stirrup and size information are marked in a centralized way

Essential characteristics of 3-column member name number table

4 essential characteristics of elevation

Necessary characteristics of 5-section polygon

6 section size marking essential characteristics

7 the longitudinal rib distribution point position is integrated with the necessary characteristics, and the red circle marks the longitudinal rib distribution point position

8 the stirrup laying line has the necessary characteristics, and the red line marks the longitudinal bar laying position

9 optional feature of stirrup splitting diagram line set table, red cutting head pointing to

Optional characteristics of large sample of 10 longitudinal bar centralized labeled table

Optional characteristics of 11 longitudinal bar in-situ labeling table large sample

Essential characteristics of a large sample of 12-stirrup centralized labeling table

Optional features of 13 stirrup in-situ labeling table large sample

The characteristic division is carried out on the column large sample design drawing in the table above, and the CAD graphic elements are classified according to the characteristic design layer which can be designed according to the characteristics. The feature layer is shown in table 3.

TABLE 3

Sequence number feature layer naming

Table 1 COLUMN _ DETAIL _ TABLE

2 in-situ labeling of the Large sample set COLUMN _ DETAIL _ YUANWEI _ ZJBZ

3 COLUMN Member NAME number COLUMN _ DETAIL _ NAME

4 ELEVATION COLUMN _ DETAIL _ ELEVATION

5 vertical Bar set notation COLUMN _ DETAIL _ POINTBAR _ JZBZ

6 stirrup set label COLUMN _ DETAIL _ LINEBAR _ JZBZ

7 Cross-section POLYGON COLUMN _ DETAIL _ POLYGON

8 Cross-sectional dimension notation COLUMN _ DETAIL _ SIZE _ LABEL

COLUMN _ DETAIL _ POINTBAR _ SET is located to 9 longitudinal muscle-laying point

The 10 longitudinal bar in situ notation COLUMN _ DETAIL _ POINTBAR _ YWBZ

11 stirrup-laying wire SET COLUMN _ DETAIL _ LINEBAR _ SET

12-stirrup splitting diagram line set COLUMN _ DETAIL _ LINEBAR _ RF

13 stirrup in situ labeling COLUMN _ DETAIL _ LINEBAR _ YWBZ

In some embodiments, the cylindrical macromolecular member comprises the following design element features: the method comprises the following steps of table, in-situ big sample centralized labeling, column component name numbering, elevation, longitudinal bar centralized labeling, stirrup centralized labeling, section polygon, section size labeling, longitudinal bar arrangement point position set, longitudinal bar in-situ labeling, stirrup arrangement line set, stirrup splitting line set and stirrup in-situ labeling.

As shown in fig. 2, in some embodiments, the classifying of the feature layer of the pillar-macro CAD drawing includes:

s305, automatically classifying the size labeling feature map layers;

s306, automatically classifying the section polygon feature layers;

s311, automatically classifying the stirrup in-situ labeling feature map layer.

In some embodiments, step S301 comprises:

1.1, automatically identifying the form;

then the following reasoning is:

1.2, identifying a header;

In some embodiments, in step S302, the original column proof CAD drawing set is labeled with the name, longitudinal bars, stirrups, dimension information, and other related information of the column proof, including:

In some embodiments, in the step S303, the two-type table column sample includes a column sample, each cell completely includes a name, a longitudinal bar, a hoop bar, elevation information, and section information of the column sample, and the automatic classification of the layer includes:

3.3.3.1 the text primitive is connected to the fill-section polygon by a lead;

In some embodiments, the type one table column big sample in step S304 has the column big sample name, longitudinal bar, hoop bar, elevation information, section information field header cells, and the values of the fields are also located in different independent table cells.

Assuming that the list head is called list head according to column arrangement, the automatic classification of the layer comprises the following steps:

In some embodiments, the size-labeling feature recognition in step S305, the table column prototype and the in-situ column prototype have different automatic recognition and classification schemes.

5.3 size marking recognition algorithm;

In some embodiments, the automatic identification and classification of the cross-sectional polygonal features, the table column large samples and the in-situ column large samples in step S306 specifically include:

In some embodiments, step S307 comprises:

In some embodiments, the in-situ labeling of the longitudinal bars in step S308 is performed on the longitudinal bars, and only appears in the CAD drawings of the column prototype of the form, and includes:

In some embodiments, step S309 comprises:

In some embodiments, the stirrup splitting diagram in step S310, which appears only in the tabular pillar-proof CAD drawings, includes:

In some embodiments, the in-situ labeling of the stirrup in step S311 is performed on the stirrup splitting diagram, and only appears in the column prototype CAD drawing of the table, and includes:

The invention provides an embodiment of a column proof building drawing layer classification system facing machine learning, which comprises the following steps:

In some embodiments, the pillar large sample feature layer classification unit includes:

Furthermore, a computer-readable storage medium is provided, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method.

Furthermore, a server may be provided, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the above method when executing the program.

Compared with the prior art, the invention has the following advantages:

For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functionality of the units may be implemented in one or more software and/or hardware when implementing the present application.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only an example of the present application and is not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A column full-page proof building drawing layer classification method facing machine learning is used for a server side and comprises the following steps: and importing the CAD construction drawing, preprocessing, dividing the CAD construction drawing into sub-drawings according to the graphic elements, establishing a corresponding attribute information mapping relation, analyzing the column large-sample characteristic layer, identifying and classifying, and finishing the classification of the column large-sample CAD construction drawing characteristic layer according to the table layer.

2. The machine-learning-oriented layer classification method for column large-scale building drawings according to claim 1, wherein the CAD building drawing comprises a subgraph set formed by subgraphs of columns, walls, beams, plates, reinforcing bars, stairs and other various building components on different floors.

3. The machine-learning-oriented column large-scale construction drawing layer classification method according to claim 2, wherein the subgraphs are rectangular frames, and the subgraphs are separated from each other geometrically.

4. The machine-learning-oriented column large sample construction drawing layer classification method according to claim 1, 2 or 3, wherein the step of segmenting into subgraphs according to graphic elements comprises the following specific steps:

5. The machine-learning-oriented column large sample construction drawing layer classification method according to claim 4, wherein the largest rectangular graphic element set refers to rectangular elements which are not included by other rectangles.

6. The machine-learning-oriented image layer classification method for the post-large-scale construction drawings of the claims 1 or 2, wherein the establishing of the mapping relation of the corresponding attribute information needs to identify the sub-image application, and the method comprises the following steps:

7. The machine-learning-oriented column proof construction drawing layer classification method according to claim 1, wherein the column proof components are divided into an in-situ column proof and a table column proof, and the table column proof is divided into a type I and a type II and comprises the following expression information characteristics: name, elevation, section polygon, section dimension information, longitudinal bar arrangement position information, stirrup arrangement position information, longitudinal bar information, and stirrup information.

8. The machine-learning-oriented column proof construction drawing layer classification method according to claim 1, wherein the column proof member comprises the following design element features: the method comprises the following steps of table, in-situ big sample centralized labeling, column component name numbering, elevation, longitudinal bar centralized labeling, stirrup centralized labeling, section polygon, section size labeling, longitudinal bar arrangement point position set, longitudinal bar in-situ labeling, stirrup arrangement line set, stirrup splitting line set and stirrup in-situ labeling.

9. The machine-learning-oriented column proof construction drawing layer classification method according to claim 1 or 7, characterized in that the column proof CAD construction drawing feature layer classification comprises:

s305, automatically classifying the size labeling feature map layers;

s306, automatically classifying the section polygon feature layers;

s311, automatically classifying the stirrup in-situ labeling feature map layer.

10. The machine learning-oriented column proof construction drawing layer classification method according to claim 9, wherein the step S301 includes:

1.1, automatically identifying the form;

then the following reasoning is:

1.2, identifying a header;

11. The machine-learning-oriented layer classification method for column large sample construction drawings according to claim 9, wherein in step S302, names, longitudinal bars, stirrups, dimension information and other related information of column large samples are marked in a raw column large sample CAD drawing set, and the method includes:

12. The machine-learning-oriented column large sample construction drawing layer classification method according to claim 9, wherein in the step S303, the column large sample of the two types of tables includes a name, a longitudinal bar, a hoop bar, elevation information and section information of the column large sample, each cell is complete, and the automatic classification of the layer includes:

3.3.3.1 the text primitive is connected to the fill-section polygon by a lead;

13. The machine-learning-oriented column big sample construction drawing layer classification method according to claim 9, wherein the type one column big sample in step S304 has column big sample name, longitudinal bar, hoop bar, elevation information, section information field header cells, and values of each field are also located in different independent table cells.

14. The machine-learning-oriented column proof construction drawing layer classification method according to claim 9, wherein in the step S305, the size marking feature recognition is performed, and the table column proof and the in-situ column proof have different automatic recognition and classification schemes.

5.3 size marking recognition algorithm;

15. The machine-learning-oriented column proof construction drawing layer classification method according to claim 9, wherein the section polygon feature, the table column proof and the in-situ column proof in step S306 have different automatic identifications and classifications, specifically comprising:

16. The machine-learning-oriented column large sample construction drawing layer classification method according to claim 9, wherein the step S307 includes:

17. The machine-learning-oriented column proof construction drawing layer classification method according to claim 9, wherein the in-situ labeling of the longitudinal bars in step S308 is performed on the longitudinal bars and only appears in the table column proof CAD drawing, and comprises:

18. The machine-learning-oriented column proof construction drawing layer classification method according to claim 9, wherein the step S309 includes:

19. The machine-learning-oriented column proof construction drawing layer classification method according to claim 9, wherein the stirrup split map in the step S310 appears only in the table column proof CAD drawing, and comprises:

20. The machine-learning-oriented column proof construction drawing layer classification method according to claim 9, wherein the stirrup in-situ labeling in step S311 is labeled on a stirrup splitting diagram and only appears in a table column proof CAD drawing, and comprises:

21. The utility model provides a towards big appearance building drawing picture layer classification system of post of machine learning which characterized in that the system includes:

22. The system according to claim 21, wherein the histogram feature layer classification unit comprises: