Idea
Near real-time TDOM generation platform for urban planners and surveyors enabling fast, accurate geospatial mapping from live images
Research Paper
Core Innovation
This paper introduces A-TDOM, which performs near real-time True Digital Orthophoto Map generation by optimizing 3D Gaussian Splatting on-the-fly as new images arrive. Unlike traditional offline pipelines, it updates camera poses and sparse point clouds dynamically and integrates new data into coarse regions, enabling immediate rendering with maintained accuracy and quality.
Market Size (TAM)
$2–10B TAM for geospatial mapping and urban planning tools; $1–3B SAM from city management and land surveying sectors. Driven by demand for real-time mapping and improved data accuracy.
Potential Customers & Pain Points
- Urban Planners Needing Up-to-Date Maps
- Land Surveyors Requiring Faster Orthophoto Generation
- City Management Teams Facing Delays in Geospatial Data
- GIS Professionals Struggling with Quality Degradation from Occlusions and Pose Errors
Business Model
Subscription-based SaaS platform with tiered pricing for different user scales and API access for integration with GIS software
Competitive Landscape
- Pix4D
- DroneDeploy
- Bentley Systems
Implementation Challenges
- Integration with existing GIS workflows
- Handling diverse and large-scale datasets
- Ensuring robustness in complex urban environments
Validation Strategy
- Pilot deployments with urban planning agencies
- Benchmarking against traditional TDOM pipelines on public datasets
- User feedback collection for iterative improvements
Research Paper Overview
A-TDOM: Active TDOM via On-the-Fly 3DGS
Summary
True Digital Orthophoto Map (TDOM) is essential for urban management, city planning, and land surveying but traditional methods are slow and prone to quality issues. This work presents A-TDOM, a near real-time TDOM generation method using On-the-Fly 3DGS optimization. It computes image poses and sparse point clouds as images are acquired, integrates new Gaussians into under-reconstructed areas, and uses orthogonal splatting for immediate rendering after each update. Experiments show A-TDOM can render TDOM in seconds per image with good quality and geometric accuracy.