Deep Neural Network for Construction Drawing Element Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems fail to accurately and efficiently interpret construction engineering drawings due to complexity, overlapping elements, and the need to recognize both text and non-text matter, leading to errors and inefficiencies in extracting construction elements for construction processes.
Innovation Solution
A deep neural network-based system that uses a network of neural networks to detect and classify construction elements, including text and symbols, and provides an interactive visualization for error correction, enabling the generation of accurate lists of construction elements from recognized elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional OCR solutions are used to recognize text in construction engineering drawings, then text recognition can be performed, but the quality and accuracy of recognition deteriorates due to the complexity of the drawings, overlapping elements, and surrounding graphical elements
Solution Approach 1:
The patent divides the complex construction drawing into multiple regions and processes each region separately using a grid-based approach. The drawing is segmented into cells, and text recognition is performed on individual cells rather than the entire complex drawing at once, which improves accuracy by reducing interference from overlapping elements and graphical elements.
Solution Approach 2:
The patent introduces an intermediary processing step that detects and removes graphical elements (lines, symbols, hatching) before performing text recognition. This intermediary step cleans the image by eliminating interfering graphical elements that would otherwise degrade OCR accuracy in complex construction drawings.
2Reliability
If manual methods are used to generate lists of construction elements from drawings, then accuracy can be maintained through expert knowledge, but the process becomes time-consuming and inefficient
Solution Approach 1:
The system enables automated self-service extraction of construction elements by using machine learning models trained on construction drawing data. The system automatically detects, classifies, and extracts construction elements (walls, doors, windows, furniture, etc.) without requiring manual intervention, thereby maintaining high accuracy through trained algorithms while dramatically improving processing speed and productivity.
Solution Approach 2:
The patent transforms the construction drawing from its original visual format into structured data parameters by detecting geometric features, text annotations, and symbolic representations. This parameter transformation enables automated classification and extraction of construction elements, bridging the gap between visual interpretation and structured data output with high accuracy and efficiency.
3Device complexity
If a single neural network is used for object detection, then the system remains simple, but the ability to accurately detect and classify multiple types of construction elements (ducts, pipes, equipment) deteriorates
Solution Approach 1:
The patent employs a multi-network architecture where different neural networks are specialized for detecting different types of construction elements. Separate networks are trained for detecting ducts, pipes, equipment, and other elements, with each network optimized for its specific detection task. This segmentation of detection functions improves accuracy for each element type while maintaining manageable system complexity through modular design.
4Extent of automation
If AI-assisted OCR is applied to construction engineering drawings, then automated processing is achieved, but challenges arise from overlapping elements, incomplete drawings, inconsistent text labeling, and symbols incorrectly recognized as text
Solution Approach 1:
The patent implements preprocessing steps that detect and flag potential problems before text recognition occurs. The system identifies overlapping elements, incomplete drawings, and ambiguous regions in advance, and applies corrective measures or adjusts recognition parameters beforehand to prevent errors. This prior cushioning approach mitigates the impact of drawing quality issues on interpretation accuracy.
Solution Approach 2:
The system incorporates feedback mechanisms where detection results are validated and corrected iteratively. The neural networks process the drawing, initial results are generated, then feedback from result validation (checking for consistency, completeness, and合理性) leads to refined processing. This feedback loop improves reliability by identifying and correcting errors from overlapping elements, incomplete drawings, and symbol misrecognition.
Data Source
AI summary
A deep neural network-based system for detection and recognition of construction elements in construction engineering drawings. A first neural network performs object detection and recognition to detect and recognize ducts and fittings. A second neural network performs text detection to detect text. A third neural network performs text recognition. Pixel tracing of the construction engineering drawing is performed using the bounding box for detected text as a reference point. The recognized text is filtered to output system types and system sizes from the recognized text. Each system type and size instance is correlated to a detected duct, pipe or fitting instance to determine object coordinates, system type, item type, size and length. A list of construction elements is generated for all detected duct, pipe and fitting instances. An interactive visualization of the construction engineering drawing based on the list of construction elements and/or a costing may be provided.