A method for identifying construction drawing graphical elements

Abstract

The application discloses a kind of identification methods of construction drawing graphic component, comprising the following steps: S1: obtaining all original graph elements of target construction drawing;S2: the original graph element is constructed as minimization graphics, and the minimization graphics database is obtained;S3: a graph is taken from the minimization graphics database, and the target detection model is input into the graph, and the target detection result of the graph is obtained;S4: the preset rule and feature corresponding to target classification component are taken out from rule base and feature base, whether the target detection result satisfies preset rule and feature is judged;If satisfy, the target detection result is output, corresponding graph is removed from database, and the next graph is returned step S3 continues next identification task;If not satisfy, the graph is processed back into cage;S5: repeat steps S3-S4 until the minimization graphics database cannot detect target component.The application can more accurately and more quickly identify construction drawing graphic component.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of graphic component recognition, and in particular to a construction drawing graphic component recognition method. BACKGROUND

[0002] The construction drawing is an important drawing in engineering design, which describes the actual construction details and technical requirements of the building, and is a drawing that represents the overall layout of the project, the external shape of the building and structure, the internal arrangement, the structure, the internal and external decoration, the material method, and the requirements of equipment and construction.

[0003] The recognition technology refers to a technology of automatically recognizing and classifying the content in an image through computer vision, image processing, pattern recognition, and machine learning. The application of the recognition technology can automatically analyze various components and structures in the construction drawing, recognize the type, size, position, and other key information of the components and structures, so as to realize the automatic construction and design optimization of the construction drawing, and achieve the purposes of improving the design efficiency, avoiding design errors, and reducing the construction cost.

[0004] Some construction drawing graphic component recognition methods have been proposed in the prior art, which mainly include rule and feature-based recognition and machine learning-based recognition. Among them,

[0005] The rule and feature-based recognition method analyzes and extracts the geometric features, topological structure, connection relationship, and graphic features of the construction drawing, so as to realize the recognition of the construction drawing components, and automatically extract the key information such as the size and position of the components. The identification of components using this method needs to go through the following three steps: screening the target graph element from the mass of graph elements according to the graph element attributes; constructing the target graph element into a target graph according to the recognition rules; and classifying the target graph according to the rules and features to obtain the recognition result. Thus, the recognition method has the following main defects:

[0006] 1. Slow recognition speed: For most construction drawings, the number of graph elements contained is very large, and the number and complexity of components can be very large. Since the recognition of each component needs to complete the above three steps, the recognition method needs to consume a long time, resulting in slow recognition speed, which affects the efficiency and real-time performance of the actual application.

[0007] 2. Easily affected by drawing factors: Since the recognition method is based on the basic attributes of the graph elements and uses rules and features to recognize the components, when the drawing habits of the drafter do not conform to the preset rules (such as the attributes of the layers / color / line type do not conform to the preset rules), the attributes of the graph elements / graphs do not conform to the preset rules, which may cause the recognition to fail and affect the recognition effect. In fact, since the operation habits of the drafter cannot be constrained, this phenomenon exists in a large number of actual construction drawings.

[0008] The recognition method based on machine learning realizes automatic recognition of components by learning a large number of labeled construction drawing data, which can utilize algorithms such as deep learning to recognize various components in the construction drawing and classify them.

[0009] However, the recognition method based on machine learning also has the following main defects:

[0010] 1. High misjudgment rate: There is a lot of similarity in the expression of construction drawing graphic components, that is, many different types of components have very high similarity in graphic expression and even have completely consistent graphic characteristics, which brings great challenges to classification using machine learning.

[0011] 2. High dependence on data and labeling: The recognition method requires a large amount of labeled data to support model learning and optimization, so it has high dependence on construction drawing labeling. If the quality of the labeled data is not high or there is a lack of labeled data, it will affect the effect and application range of the recognition technology. In fact, since most construction drawings are non-public confidential documents, it is difficult to find a sufficient number of high-quality construction drawings for real production environment to perform data labeling.

[0012] 3. High requirement for hardware devices: The recognition method needs to process a large amount of images and data, so it has high requirements for hardware devices. If the hardware device configuration is insufficient, it will affect the running effect and speed of the recognition technology.

[0013] In summary, the current construction drawing graphic component recognition technology still has some defects and challenges, and needs to be further optimized and improved to improve the recognition accuracy and speed, reduce the dependence on labeled data and hardware devices, and realize more accurate and efficient construction drawing component and design optimization. SUMMARY

[0014] In view of the above problems, the present application aims to provide a construction drawing graphic component recognition method.

[0015] The technical solution of the present application is as follows:

[0016] A construction drawing graphic component recognition method, comprising the following steps:

[0017] S1: obtaining all original graph elements of a target construction drawing;

[0018] S2: constructing the original graph elements into minimized graphics to obtain a minimized graphic database;

[0019] S3: taking one graphic from the minimized graphic database and inputting the graphic into a target detection model to obtain a target detection result of the graphic, the target detection result including classification information and position information of a target component;

[0020] S4: taking out preset rules and features corresponding to the target classification component from the rule library and the feature library, judging whether the target detection result meets the preset rules and features;

[0021] If yes, the target detection result is outputted, a target component is obtained, the graph of the target component is removed from the minimized graph database, and the next graph is taken in step S3 to continue the next recognition task.

[0022] If no, the graph is subjected to a back cage processing.

[0023] S5: repeating steps S3-S4 until no target component can be detected in the minimized graph database.

[0024] Preferably, in step S2, the original graph elements are constructed into the minimized graph according to the positional relationship.

[0025] Preferably, in step S3, the target detection model adopts a machine training model.

[0026] Preferably, in step S4, the back cage processing of the graph specifically includes the following sub-steps:

[0027] S41: taking out negative rules and negative features corresponding to the target classification component from the negative rule library and the negative feature library, judging whether the target detection result meets the negative rules and negative features;

[0028] If yes, it indicates that the target detection result is a false recognition, the graph is returned to the minimized graph database, and the next graph is taken in step S3 to continue the next recognition task.

[0029] If no, step S42 is entered.

[0030] S42: judging whether there is a graph adjacent to the target detection result in the minimized graph database.

[0031] If no, it indicates that the target detection result is a false recognition, the graph is returned to the minimized graph database, and the next graph is taken in step S3 to continue the next recognition task.

[0032] If yes, a graph adjacent to the target detection result is taken from the minimized graph database, the graph and the target detection result are recombined into a new graph, and step S3 is returned to continue the next recognition task with the new graph.

[0033] The present application has the following beneficial effects:

[0034] The construction drawing graphic component is constructed by using the minimized graph, and the performance is significantly higher than that of the traditional graphic element recognition method; compared with the recognition method based on machine learning, the recognition method does not completely depend on the classification result of the machine training model, but uses the classification result of the machine training model as a reference dimension, and can avoid the misrecognition problem caused by the similarity problem of the machine learning for the construction drawing graphic component. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0036] Figure 1 The figure is a flowchart of the construction drawing graphic component recognition method of the present application. DETAILED DESCRIPTION

[0037] The present application will be further described below in combination with the drawings and embodiments. It should be noted that the embodiments in the present application and the technical features in the embodiments can be combined with each other without conflict. It should be noted that unless otherwise specified, all technical and scientific terms used in the present application have the same meaning as generally understood by those skilled in the art to which the present application belongs. The present application discloses that the "including" or "containing" and similar words mean that the elements or objects before the words cover the elements or objects listed after the words and their equivalents, and do not exclude other elements or objects.

[0038] As shown in Figure 1 The present application provides a construction drawing graphic component recognition method, comprising the following steps:

[0039] S1: obtaining all original graphic elements of a target construction drawing.

[0040] S2: constructing the original graphic elements into minimized graphs to obtain a minimized graph database.

[0041] In one specific embodiment, when constructing the minimized graph, the original graphic elements are constructed into the minimized graph according to the positional relationship. In this embodiment, since only the positional relationship of the graphic elements is referred to when constructing the minimized graph, other drawing factors (such as the layer, color, etc. of the graphic elements) do not need to be referred to, so the recognition result is basically not affected by the drawing factors, which has great significance in practical application.

[0042] S3: taking one graph from the minimized graph database and inputting the graph into a target detection model to obtain a target detection result of the graph, the target detection result including classification information and position information of a target component.

[0043] In one specific embodiment, the target detection model adopts a machine training model. It should be noted that using a machine training model to identify a graph is prior art, and the specific structure of the machine training model is not described here.

[0044] S4: taking pre-set rules and features corresponding to the target classification component from a rule library and a feature library to determine whether the target detection result meets the pre-set rules and features;

[0045] If yes, the target detection result is outputted to obtain a target component, the graph of the target component is removed from the minimized graph database, and the next graph is taken to continue the next identification task in step S3;

[0046] If no, the graph is subjected to a back cage process.

[0047] In one specific embodiment, the back cage process of the graph specifically includes the following sub-steps:

[0048] S41: taking negative rules and features corresponding to the target classification component from a negative rule library and a negative feature library to determine whether the target detection result meets the negative rules and features;

[0049] If yes, it indicates that the target detection result is a false recognition, the graph is returned to the minimized graph database, and the next graph is taken to continue the next identification task in step S3;

[0050] If no, step S42 is entered;

[0051] S42: detecting whether there is a graph adjacent to the target detection result in the minimized graph database;

[0052] If no, it indicates that the target detection result is a false recognition, the graph is returned to the minimized graph database, and the next graph is taken to continue the next identification task in step S3;

[0053] If yes, one graph adjacent to the target detection result is taken from the minimized graph database, which is combined with the target detection result to form a new graph, and the next identification task is continued with the new graph in step S3.

[0054] It should be noted that in the above embodiment, the rule base and feature base, the negative rule base and negative feature base are all knowledge bases established by empirical knowledge, and the extraction and determination of rules and features, negative rules and negative features can be completed by an inference engine, which is prior art, and the specific extraction and determination method will not be described here.

[0055] S5: repeating steps S3-S4 until the target component cannot be detected in the minimized graph database.

[0056] In the above embodiment, the present application constructs the construction drawing graph component by minimizing the graph database, that is, all the construction drawing graph components to be identified are composed by different minimized graphs, and the basic elements are graphs rather than traditional primitives. Since the number of minimized graphs in the drawing is much less than the number of primitives in the drawing (usually the amount of geometric data is reduced by several orders of magnitude), the performance of the present application is significantly higher than that of the traditional recognition method.

[0057] In addition, in the above embodiment, the present application uses the target detection model to pre-screen the target component, rather than relying on the classification result of the target detection model. That is, the present application does not completely rely on the classification result of the target detection model, but uses it as a reference dimension, and the classification result is determined by the expert system in a positive and negative direction and provides further processing method, so that the technical solution of the present application can avoid the misrecognition problem caused by the similarity problem of the construction drawing graph component in machine learning. At the same time, due to the intervention of the expert system, the demand for data sample size and annotation amount of machine learning is further reduced.

[0058] In summary, the present application can improve the accuracy and speed of construction drawing component recognition, and can make the recognition of construction drawing graph component no longer affected by the drawing factor, and solve the misrecognition problem caused by the similarity problem of the construction drawing graph component expression. Compared with the prior art, the present application has made significant progress.

[0059] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above with a preferred embodiment, it is not intended to limit the present application, and any person skilled in the art can make some changes or modifications to the above disclosed technical content without departing from the scope of the technical solution of the present application, and any simple modification, equivalent change and modification of the above embodiment according to the technical essence of the present application are still within the scope of the technical solution of the present application.

Claims

1. A method of identifying construction drawing graphical elements, characterized by, The method comprises the following steps: S1: obtaining all original graph elements of a target construction drawing; S2: constructing the original graph elements into minimized graphs to obtain a minimized graph database; S3: taking one graph from the minimized graph database and inputting the graph into a target detection model to obtain a target detection result of the graph, the target detection result comprising classification information and position information of a target component; S4: taking preset rules and features corresponding to the target classification component from a rule library and a feature library to determine whether the target detection result meets the preset rules and features; If yes, the target detection result is outputted to obtain the target component, the graph of the target component is removed from the minimized graph database, and the next graph is taken to continue the next recognition task; If no, the graph is subjected to cage processing; the cage processing of the graph comprises the following sub-steps: S41: taking negative rules and features corresponding to the target classification component from a negative rule library and a negative feature library to determine whether the target detection result meets the negative rules and features; If yes, it is indicated that the target detection result is misrecognized, the graph is returned to the minimized graph database, and the next graph is taken to continue the next recognition task; If no, step S42 is entered; S42: detecting whether there is a graph adjacent to the target detection result in the minimized graph database; If no, it is indicated that the target detection result is misrecognized, the graph is returned to the minimized graph database, and the next graph is taken to continue the next recognition task; If yes, a graph adjacent to the target detection result is taken from the minimized graph database, the graph and the target detection result are combined into a new graph, and the next recognition task is continued with the new graph in step S3; S5: repeating steps S3-S4 until no target component can be detected in the minimized graph database.

2. The construction drawing graphical element recognition method according to Claim 1, wherein In step S2, when the minimized graphs are constructed, the original graph elements are constructed into the minimized graphs according to position relationships.

3. The construction drawing graphical element recognition method according to Claim 1, wherein In step S3, the target detection model adopts a machine training model.

Images