Building Drawing Equipment Extraction Using ML Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Extracting information from building engineering drawings is challenging due to their two-dimensional nature and lack of distinct foreground and background, making it difficult to automate the identification and classification of equipment and measurement points.
Innovation Solution
A neural-network based deep learning architecture is employed to automate data preprocessing, feature extraction, and classification, using machine learning models for image segmentation, text recognition, and classification to identify and classify equipment types, points, and their relationships from building engineering drawings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual methods are used to extract information from building engineering drawings, then information extraction accuracy can be maintained, but productivity is low and the process is time-consuming
Solution Approach 1:
The patent introduces an intermediary system consisting of deep learning models (image segmentation model, text recognition model, classification model) that act as a mediator between the building engineering drawings and the final extracted information. This intermediary automatically processes the drawings through multiple stages: segmenting the image into regions, recognizing text within those regions, and classifying equipment and measurement points, thereby achieving both high productivity and maintained accuracy without direct manual extraction.
2Productivity
If automated processing is implemented on building engineering drawings, then productivity increases, but measurement precision deteriorates due to the two-dimensional nature and lack of distinct foreground and background
Solution Approach 1:
The patent applies segmentation by dividing the building engineering drawing into multiple distinct regions using an image segmentation deep learning model. This segmentation separates different elements (equipment, measurement points, text, graphical representations) into identifiable regions, creating distinct foreground and background areas that enable accurate automated processing while maintaining high productivity through machine learning-based classification.
3Measurement precision
If complex deep learning architectures are used for accurate classification, then measurement precision improves, but device complexity increases
Solution Approach 1:
The complex classification task is segmented into multiple specialized deep learning models, each responsible for a specific function: image segmentation for region division, text recognition for extracting textual information, and classification for identifying equipment and measurement points. This segmentation of the overall system into specialized components achieves high classification accuracy while managing complexity through modular architecture.
Solution Approach 2:
The deep learning-based processing system is designed with universal applicability to handle various types of building engineering drawings and equipment classifications through a unified multi-stage architecture. The same framework processes different drawing types (HVAC, electrical, plumbing) and identifies diverse equipment categories, reducing the need for multiple specialized systems and managing overall complexity.
Data Source
AI summary
Techniques for determining equipment names and other information from engineering drawings are described. In an example embodiment, image data representing at least a portion of an engineering drawing for a building is received, the image data is processed using a first ML model to identify a first portion of the image data corresponding to a first region from a set of regions, the first portion of the image data is processed to recognize first text data, using at least a second ML model the first text data is determined to correspond to a first equipment type from a set of equipment types, wherein the set of equipment types relate to the equipment included in the building, and using the first equipment type and an ontology representing relationships between equipment types and equipment name, a first equipment name corresponding to the first text data is determined.