TL;DR
This paper introduces a novel system that adaptively visualizes Building Information Models (BIM) by highlighting relevant components for specific user roles. By leveraging saliency models tailored for architectural imagery and an ontological classification of BIM elements, the system dynamically adjusts visualization parameters to enhance user attention and improve information comprehension in complex construction projects.
What Problem Does This Paper Address?
Building Information Modeling (BIM) has revolutionized construction by integrating vast amounts of data into a 3D model. However, the sheer complexity and volume of information within BIM models pose significant challenges for adoption. Professionals from different disciplines (architects, structural engineers, method engineers, etc.) require specific subsets of this information, and current methods for managing this often involve manually hiding irrelevant data. This manual process is error-prone, time-consuming, and can lead to critical clashes being overlooked. The core problem this paper addresses is the difficulty users face in efficiently extracting and comprehending the information most relevant to their specific roles within a complex BIM environment.
Methods
The proposed system combines two key innovations: an ontological classification of BIM elements and an adapted saliency model for architectural visualization.
- Ontology-Based Classification: The system utilizes an ontology, derived from the knowledge of BIM managers, to classify BIM elements based on their relevance to different professional departments. This allows for automatic labeling of building elements.
- Adapted Saliency Model: Standard saliency models, often developed for natural images, are adapted to better suit architectural imagery. This adaptation specifically accounts for the strong presence of vanishing points and depth information, which are characteristic of architectural renderings. The model computes conspicuity maps for intensity, color, 3D orientations (incorporating vanishing points), and depth. These maps are then combined to generate a saliency map.
- Adaptive Visualization: Based on the saliency map and the user's profile, the system dynamically adjusts visualization parameters (hue, saturation, and value) of irrelevant pixels. The goal is to enhance the saliency of relevant objects without hiding other information, thereby improving visual attention and comprehension. This is achieved through a particle swarm optimization method that balances the saliency of relevant elements with the visibility of irrelevant ones.
Datasets
The paper describes experiments conducted using three real BIM models from Bouygues Construction projects in Paris, Nantes, and Lyon. These models were exported to the Unity3D engine for real-time visualization. The experiments involved generating enhanced color images for specific user profiles (structure, method, and plumbing engineers) based on these models. No specific public datasets are mentioned; the data is derived from proprietary construction projects.
Key Insights and Ideas
- Saliency for BIM is More Than Just Image Analysis: The paper demonstrates that saliency models, typically used for natural image analysis, can be effectively adapted and applied to the domain of architectural visualization within BIM.
- Architectural Imagery Has Unique Properties: Recognizing and leveraging the specific visual characteristics of architectural images (vanishing points, depth) is crucial for developing effective saliency models in this domain.
- Dynamic Color Adjustment Enhances Comprehension: Instead of hiding information, dynamically adjusting colors to emphasize relevant elements offers a more robust and less error-prone approach to managing information overload in BIM.
- User-Centric Visualization is Key: The system's ability to tailor visualizations to individual user roles is a critical factor in its effectiveness, directly addressing the problem of role-specific information needs.
Key Results
The experimental validation involved 21 construction experts (9 structure engineers, 8 method engineers, and 4 plumbing engineers). The results showed a strong preference for the saliency-enhanced visualizations:
- The saliency-enhanced image was chosen as more comfortable to work with in 73% of cases.
- The original image was chosen in only 9.9% of cases.
- 17.1% of answers were "indifferent," suggesting users recognized the system's advantage but were awaiting a full industrial BIM viewer implementation.
- A Wilcoxon's Signed-Rank test confirmed the statistical significance of these results (p-value = 0.0001).
Limitations and Future Work
- Limitation: A limitation identified during experiments was that for method engineers viewing images from a distance, the system sometimes struggled to render irrelevant elements clearly, occasionally turning them into "non-identifiable dark elements." This was attributed to the influence of the background on saliency.
- Future Work: The authors suggest extending the system with other visualization transformations, such as transparency and spatial position adjustments. They also propose improving the building ontology to include elements crucial for multiple departments and to incorporate priority information for complex objects. Further enhancements could involve adapting the visualization based on the virtual camera's position and orientation to define more appropriate enhancements.
Summary updated