Alpine's Project Experience

Sports and Entertainment Venues

Major-League Arenas and Stadiums

We captured a professional sports stadium end to end in one week, interior through site, and have documented major-league arenas in top United States markets. Seating bowl, event floor, and back-of-house by ground scanner. Building envelope by ground scanner. The rest of the site by air, all registered together. Terrestrial laser scanning would have needed a multiple of that schedule and a larger crew for the same coverage.

Sightlines ran hundreds of feet across open volume. Seating geometry repeated until registration software could not tell one section from another. Access came between events rather than around them.

How we controlled error at this scale

• We captured each level and zone as a separate scan segment and registered the segments together, so a problem cost an hour instead of a day.
• We closed every loop, returning each segment to its start point so accumulated drift resolved at closure instead of propagating into the merged dataset.
• We tied ground, exterior, and aerial data to shared control, so the whole venue arrived in one coordinate system with no alignment work at delivery.

Delivered: Merged point clouds (E57, LAS, LAZ, RCP, PLY), 3D Gaussian Splat environments, textured and collision meshes, aerial orthomosaics and site surfaces, encrypted delivery to client-controlled storage.

Mission-Critical Facilities

Colocation Data Centers

We documented live colocation facilities under escort with no cage opened, no rack accessed, no change window, and no impact to critical load. White space and containment aisles, electrical rooms, generator yards, mechanical galleries.

The difficulty was never size. It was congestion and repetition. Overhead trays, busway, conduit, and piping occluded each other from any single vantage point, and every aisle presented the same geometry to the scanner. Both conditions produce quiet errors: gaps that stay invisible until a designer needs the routing, and misregistration that looks clean until someone measures it.

How we controlled error in congested space

• We walked critical aisles on opposing headings so stacked overhead systems were seen from both sides and trays did not shadow what ran above them.
• We registered to established control rather than scan geometry alone, in buildings engineered to look identical in every direction.
• We documented control for recurring programs, so a capture two years out registers to the same coordinate system as the first and change is measurable.
• Where photography was restricted, we scoped the capture to geometry only and confirmed the constraint in writing before the field date.

Delivered: Registered and georeferenced point clouds (E57, LAS, LAZ, RCP), as-built documentation of MEP and containment infrastructure, mesh-based digital twins for remote review, documented control for multi-site and multi-phase programs.

Design and Construction

Scan-to-BIM for Remodels and Renovations

We have delivered Scan-to-BIM from single tenant suites through full-facility documentation, including high-end residential and a professional sports stadium. On nearly every one the same pattern held: the existing drawings were wrong or missing, nothing above the ceiling was documented, and the design team needed a model it could build against.

High-end residential added a constraint the commercial work did not have. The houses were occupied, the finishes were unforgiving, and the owner was present. We worked on the household's schedule, moved nothing, and marked nothing.

How we kept the models honest

• We set level of development before modeling began. That single decision is what separated a fixed-price model from an open-ended one.
• We reviewed scan data for coverage and registration quality before it reached the modeling team, because a gap found mid-model becomes a site revisit.
• We checked finished models back against the point cloud before delivery. A model that no longer matches its source data is worse than none, because it will be trusted.
• We defined revision rounds in the scope of work, which is how renovation modeling stayed on budget.

Project types delivered

• Tenant improvement and interior fit-out documentation
• Multi-story office and mixed-use renovation as-builts
• High-end and custom residential documentation
• Large-venue and stadium interior documentation
• Industrial and warehouse facility modeling
• MEP-dense mechanical space coordination
• Existing-condition documentation for preservation and adaptive reuse

Delivered: Revit models (RVT) for architectural, structural, and MEPF scopes, IFC exports, Navisworks federated models (NWC, NWD), 2D floor plans, sections, and elevations (DWG, PDF). Modeling is produced through Alpine's certified production partner under our scope control and quality review.

Aerial and Site-Scale

Large-Area Aerial Mapping and Terrain

We have flown full sites and scanned the buildings on them within the same engagement, then delivered both as one registered dataset instead of two files the client had to reconcile. Large parcels, campuses, corridors, rooftops, facades, and open terrain, including sites with no cell coverage and no established control.

How we made the aerial data land correctly

• We set an on-site base station and tied the flight to the project coordinate system, so the data dropped into the design workflow without a transformation step.
• We flew nadir and oblique passes on large structures. A vertical-only flight documents roofs and leaves facades to be guessed at.
• We used multi-return LiDAR and classified the returns, so the terrain model reflected the ground that would be built on rather than the canopy above it.
• We registered aerial and ground data to shared control rather than aligning them visually afterward. Visual alignment looks correct and measures wrong.

Delivered: Aerial point clouds (LAS), orthomosaics (GeoTIFF), digital surface and terrain models, contour maps, classified terrain data, photogrammetric meshes (OBJ), high-resolution still imagery.

AI, Robotics, and Simulation

Simulation Environments and Training Data

We have captured real operating environments as LiDAR-fused 3D Gaussian Splats paired with registered point clouds from the same pass. Commercial interiors, warehouses and industrial floors, hospitality and food service spaces, offices, and large public venues.

On these projects we handled the site side of the work: securing access, executing NDAs, coordinating escorts, and capturing without disrupting the businesses operating around us.

What made the data usable downstream

• We captured appearance and geometry in one pass, one coordinate system, so the splat and the point cloud loaded together without manual alignment or rescaling.
• We shipped full metadata with every capture: date, sensor, positioning method, coordinate system. Undocumented data is unusable data six months later.
• We have never licensed, resold, or reused a client-commissioned capture as training data. Licensed environment captures run under separate agreements, and the two categories never mix.

Delivered: 3D Gaussian Splat models in PLY, OpenUSD (.usd) for NVIDIA Isaac Sim and Omniverse pipelines, 3D Tiles for browser streaming, collision meshes, registered georeferenced point clouds (E57, LAS, LAZ) with full capture metadata.

Media and Virtual Production

Environment Capture for Film and Production

We have captured locations for production use under single-visit access, with nothing on site permitted to move. The captures support remote scouting, camera and lighting planning against real geometry, and bringing real locations onto LED walls and real-time sets.

What made these captures difficult

• Glass, polished floors, reflective fixtures, and blank walls break camera-only reconstruction. We let LiDAR hold the geometry while the splat carried lighting, reflection, and transparency.
• Access was granted once, so we planned coverage before arriving and confirmed it on site before leaving.
• Unreleased location and production information stays confidential. Nothing captured for production work appears in our marketing, including this page.

Delivered: 3D Gaussian Splat environments, textured meshes (OBJ), collision geometry for real-time engines, registered point clouds for measurement and layout, web-shareable review links.