Building Interior Structure Recognition Using BIM-Generated Training Images

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for checking construction status in multi-story buildings face challenges such as varying data accuracy due to skill levels, time-consuming manual data registration, and difficulty in reusing point cloud data, especially in large-scale facilities, and require a large number of annotated images for effective deep learning-based structure recognition.

Innovation Solution

A building inside structure recognition system that uses machine learning models generated from BIM data, including correct and virtually observed images, to automate structure recognition, utilizing mask and skeleton images, and enhancing images to improve accuracy, reducing the need for real image collection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If LiDAR measurement is performed at multiple portions of the construction site, then measurement coverage is improved, but data accuracy varies depending on the skill level of the measurer

Engineering Contradiction:

Improvemeasurement coverageVSAvoiddata accuracy

Core Design Contradiction:

Area of stationary objectVSMeasurement precision

Solution Approach 1:

The patent uses BIM (building information modeling) data as a digital copy of the construction site to generate training images. This virtual copy eliminates the need for physical LiDAR measurements and manual annotations, providing consistent and accurate structure information without depending on measurer skill levels.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces manual mechanical measurement processes with an automated machine learning system. The system automatically extracts structure information from BIM data and generates training images, eliminating human skill variability from the measurement process.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Manufacturing precision

If manual data registration and structure identification is performed, then data processing is completed, but it takes time and effort

Engineering Contradiction:

Improvedata processing completenessVSAvoidtime and effort

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The machine learning system performs self-service by automatically processing construction site images and identifying structures without human intervention. The system trains on BIM-generated data and autonomously completes the data processing and structure identification tasks that would otherwise require manual registration and analysis.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces manual data registration and structure identification processes with automated machine learning algorithms. The system automatically processes images, identifies structures, and extracts information, eliminating the time-consuming manual work while maintaining processing completeness.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If a large number of annotated images are collected for deep learning, then model accuracy is improved, but data collection becomes enormous and time-consuming

Engineering Contradiction:

Improvemodel accuracyVSAvoidamount of training data

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent creates virtual copies of construction site structures through BIM data rendering. These virtual images serve as training data and automatically include accurate structure annotations since they are generated from the BIM model itself, eliminating the need for manual annotation of large quantities of real images.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The system performs preliminary action by pre-processing BIM data to generate training images and annotations before actual construction site imaging. This advance preparation creates ready-to-use training data that would otherwise require extensive manual collection and annotation efforts.

Inventive Principle:

Principle #10Preliminary action

4Quantity of substance

If rendered images are used instead of real photographs for learning, then learning data can be generated, but production costs are high and annotation work becomes enormous

Engineering Contradiction:

Improvelearning data availabilityVSAvoidproduction cost

Core Design Contradiction:

Quantity of substanceVSEase of manufacture

Solution Approach 1:

The patent uses BIM data as a digital copy of the construction site to generate training images. This approach provides comprehensive learning data coverage while avoiding the high costs of real rendered images, as BIM data can be extracted from existing construction documents and models without requiring expensive rendering processes.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS20250384675A1Building inside structure recognition system and building inside structure recognition method

Publication Date: 2025.12.18 TOPCON CORPORATION

AI summary

Provided is a building inside structure recognition system for recognizing a structure in a building by using a machine learning model. A building inside structure recognition system according to the present invention comprises: a machine learning model generation device that generates a machine learning model by executing machine learning in which a correct image generated from building information modeling (BIM) data is set as correct data and a virtually observed image generated by rendering the BIM data is set as observation data; and a building inside structure recognition device that recognizes a structure in a building by using the machine learning model generated by the machine learning model generation device.