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

Abstract

The invention discloses a machine learning-oriented column bulk sample architectural drawing layer classification method. The method is used for a server side and comprises the steps of importing a CAD architectural drawing and performing preprocessing, segmenting the CAD architectural drawing into sub-graphs according to graph elements and establishing a corresponding attribute information mapping relationship, analyzing column bulk sample feature graph layers and performing recognition classification, and finishing classification of the column bulk sample CAD architectural drawing feature graph layers according to table graph layers. According to the method, business modeling and analysis are carried out on the building components expressed by the CAD building drawings, business characteristics expressed by the design drawings are refined, and the concept of building component characteristic layers is further proposed to guide automatic classification of the layers; an automatic layer classification algorithm from business features expressed by a design drawing to building component feature layers is given, so that fine-grained feature layer automatic classification is realized,and a solid foundation is laid for simplifying a subsequent recognition algorithm and introducing a machine learning intelligent algorithm.

Description

Technical field [0001] The invention belongs to the technical field of machine learning intelligent recognition of building information models and architectural drawings, in particular to a layer classification method based on component characteristics, and in particular to a machine learning-oriented method and system for layer classification of column large-scale architectural drawings. Background technique [0002] With the development of computer software and hardware technology, the application of building information model BIM technology in the construction industry has become more and more in-depth. In the budget, construction, settlement, operation and maintenance of the construction project, a large number of BIM models are used to complete the calculation and volume. Work such as quantity and pricing has greatly improved the informatization level of the construction industry. This shows that the construction of BIM models has become particularly important. The current ...

AI Technical Summary

Problems solved by technology

[0005] 1. It is easier for humans to understand simple graphics and image features (such as parallel and intersecting), but when business scenarios become more complex, multi-angle and multi-dimensional combinations between features or more advanced features, humans cannot understand this. Features of understanding, often choose to ignore, which is not conducive to abstraction and identification at a higher level

[0006] 2. The extraction of traditional feature engineering is often based on the current professional features, and if you change to another major, on the one hand, you need to study it from beginning to end, which increases the learning cost; in addition, it is difficult to extract and integrate common features between majors. Leading to a large amount of repetitive work and weak universality

[0007] 3. The rule-based recognition algorithm is a set of completely self-consistent logical operations and symbolic operations, with rigorous logical reasoning. However, the drawings in reality vary widely. After summarizing a set of rules, it is no longer applicable when a new scene is discovered. , so it is necessary to constantly find new rules to make up for the previous shortcomings. On the one hand, the generalization ability of the software is weak, and the maintenance cost of the software is also increasing.

[0008] 4. There is no unified specification requirement for feature engineering and rule search, and it depends heavily on the quality of practitioners. It is difficult to guarantee the quality of software and the progress of research and development from an engineering perspective.

[0009] 5. At present, there are also recognition technologies that use feature engineering and traditional machine learning methods for training and prediction. Although the accuracy rate has been improved to a certain extent, it has not completely deviated from the range of artificially extracted features and relies heavily on the quality of features. Therefore, its generalization not very capable

Images