Image source musa/stock.adobe.com.
|
This is the third of a six-part special feature covering data management, digital twins, reality capture, AI, IoT, and other technology for AEC designers, engineers, and owners. This issue focuses on using technology in reality capture applications. In future articles, we’ll be digging into other topics, such as real-time ray tracing and virtual reality. |
Data collection in the AEC industry has evolved significantly in the 21st century, shifting from largely manual processes to more automated, digitally based processes. Current data collection approaches rely heavily on reality capture — a process that collects real-world data to create digital representations of physical objects, environments, or spaces. The process uses technologies such as LiDAR (light detection and ranging), photogrammetry, and other techniques to gather information on projects in various stages, starting with pre-construction conditions, progressing through construction to finished conditions, and extending into operation.
Reality capture plays key roles in projects that employ digital twins, building information modeling (BIM), and other AEC applications that rely on accurate, high-resolution data. Throughout all stages, AEC professionals now deal with larger datasets than ever before, requiring high-performance computing systems and solutions.
Photogrammetry captures the structure’s meshes but misses the surroundings and background (left). Gaussian splatting includes the background, providing better visualization for developers and projects that benefit from full context (right). Photo credit: Ben Stocker, Skender. Click images to enlarge.
Current reality capture approaches often use LiDAR and photogrammetry technologies, which can be deployed on drones, conventional aircraft, satellites, or terrestrial devices. On AEC projects, ground-based or aircraft-mounted LiDAR systems take multiple measurements encompassing an area of interest, establishing large point clouds of data. LiDAR can capture highly accurate spatial data over large areas and map various materials. It is typically more expensive than photogrammetry due to the specialized hardware required. While it excels in geometric accuracy, LiDAR provides less texture information compared with photogrammetry, and it can produce inconsistent results with highly reflective surfaces such as water or glass.
Photogrammetry uses photographic images to extract detailed spatial information about physical objects. By analyzing objects from multiple viewpoints, photogrammetry can create highly detailed 3D models. This method is often more accessible and cost-effective than LiDAR, but not as accurate.
Conventional ground-based survey equipment such as theodolites, levels, and total stations are also used to collect data, sometimes in conjunction with previously mentioned techniques. And, in addition to the various forms of location data, sensors often provide critical data on AEC projects, including information related to motion, light, temperature, pressure, humidity, and other environmental conditions.
With the large datasets generated by reality capture, AEC teams need high-performance computing systems and solutions to process and manage data. Along with properly equipped workstations, technologies such as artificial intelligence (AI) and others can help transform massive raw datasets into actionable information for design, construction, and facility management.
AI is impacting countless aspects of daily life, and AEC reality capture is no exception. By aiding time-intensive tasks such as object identification, classification, and segmentation, AI can help accelerate 3D modeling of buildings and other facilities, enhancing the accuracy and efficiency of object analysis and categorization. With the help of AI, humans can be freed from tedious, mundane tasks and focus on more creative endeavors requiring critical thinking.
Lenovo’s ThinkStation P8 features a thermally advanced, rack-optimized chassis, the latest dual 4th Gen Intel® Xeon® scalable processors, up to 120 cores, and support for up to four NVIDIA RTXTM 6000 Ada Generation GPUs. Image source Lenovo.
Technologies such as NVIDIA® CUDA® can accelerate the processing of LiDAR point clouds and photogrammetric data. Introduced in 2006, CUDA is a parallel computing platform and application programming interface (API) that dramatically improves computing performance by leveraging the GPU for high-speed, parallel processing. This significantly reduces the time required for data conversion, visualization, and analysis. CUDA enhances reality capture at multiple stages, from on-device data collection to pre-processing and post-processing. On embedded devices — such as the NVIDIA Jetson™ family of modules — CUDA accelerates image processing and AI inference. For pre-processing on more powerful GPUs in workstations and datacenters, CUDA accelerates the transfer and transformation of raw data — such as LiDAR, point-cloud data, or photographic inputs — into structured formats. Once processed, it helps transform these massive datasets into detailed 3D meshes and images.
Radiance fields are an advancement in reality capture, significantly enhancing the quality and efficiency of 3D scene reconstruction. Using machine learning, radiance fields generate detailed and realistic 3D views from a vastly reduced number of 2D images compared with traditional photogrammetry. Radiance fields can interpolate between sparse data points, creating smooth, photorealistic scenes even from angles that weren’t originally captured.
To advance reality capture, NVIDIA developed fVDB to facilitate the development of radiance field solutions and accelerate performance, translating data into AI-ready environments rendered in real time. Built on top of OpenVDB — the industry-standard library for simulating and rendering sparse volumetric data such as water, fire, smoke, and clouds — fVDB extends its capabilities to support the demands of neural rendering and large-scale 3D scene generation.
Using the Hexagon/Leica Cyclone 3DR solution, Coenradie, an engineering consultancy in The Netherlands, gathered point cloud data more efficiently than in the past, helping them save time and more accurately classify data for urban planning. Read the case study here. Image source: Leica.
According to Sean Young, Director of AECO, Geospatial, and AI Solutions Industry Marketing at NVIDIA, CUDA and GPU technology play key roles in reality capture workflows. “There’s an extensive amount of GPU work that accelerates reality capture for LiDAR, photogrammetry, and radiance fields. Most reality capture software takes advantage of CUDA,” said Young.
Computer workstation power becomes particularly critical in the preprocessing stage, according to Jon Clark, Solutions Architect for AEC/Product Design and Development at Lenovo. Workstations handling reality capture data are typically configured with multiple CPUs, multiple GPUs, and plenty of memory and storage. Regarding memory, video RAM (VRAM) is key, according to Clark.
“With reality capture, the performance bottlenecks are typically related to video memory bandwidth,” said Clark. “The more GPU VRAM you have, the bigger the scene that can be processed.”
Lenovo workstations equipped with NVIDIA professional graphics have proven well suited for AEC reality capture data. In particular, the combination of Lenovo ThinkStation and ThinkPad P Series workstations with NVIDIA RTX™ GPUs can help accelerate AI-aided reality capture throughout project lifecycles, ranging from initial data capture to drone-based construction monitoring and extending into operations and maintenance analysis.
Technology-aided reality capture is being demonstrated on numerous fronts. Esri, which develops geographic information system (GIS) software, has integrated AI with geospatial solutions to improve analysis of geospatial information. The integration, called GeoAI, enables more sophisticated decision-making and workflow acceleration.
Working in conjunction with NVIDIA GPUs and Lenovo Workstations, GeoAI can help analyze numerous scenarios and predict behaviors of complex systems such as natural resources under various conditions.
Hexagon, a provider of 3D scanning solutions and other reality capture devices, has integrated its reality capture sensors and digital reality platforms with AI, NVIDIA GPUs, and related technology. Hexagon recently upgraded AI classification in its Leica Cyclone 3DR software solution by leveraging NVIDIA CUDA. The combination of AI and reality capture technology helps users more efficiently classify point cloud data, with the latest version demonstrating a 20% faster point cloud workflow than its predecessor, according to Hexagon.
Hexagon also integrated NVIDIA Omniverse™ Cloud APIs with its scanning solutions, such as the Leica BLK2GO handheld 3D scanner. Leveraging Omniverse technology, scans are imported into Hexagon’s Reality Cloud Studio, where they can be aligned with CAD elements. The reality-based data is then brought into NVIDIA Omniverse, via OpenUSD universal data interchange, where it is merged with the platform’s physically based visualization and simulation capabilities in digital twins and other 3D models.
Reality capture has proven indispensable throughout lifecycles of AEC projects large and small. As the capabilities expand, so do the data volumes, requiring AEC professionals to adopt the right combination of hardware and software technologies to effectively capture, process, and manage project data.
GPU-accelerated technologies such as NVIDIA CUDA are playing critical roles in turning vast datasets into detailed, actionable insights. Workstations from Lenovo equipped with powerful NVIDIA RTX GPUs provide the performance needed to support reality capture workflows from field to finish. For enterprise-scale demands, Lenovo also offers rackable GPU-powered solutions that bring high-performance processing to centralized environments — ideal for collaboration, remote access, and large-scale data processing.
|
In our next article, we’ll explore how to handle remote workflows and data collaboration in AEC projects. We’ll explore the various hardware, software, and other solutions involved in effective collaboration. Stay tuned! |
↓↓↓Download the White Paper Series ↓↓↓
***
This article was sponsored by Lenovo and NVIDIA.
Searching for more information about Architecture, Infrastructure, and Construction?
Click here!